{"global_step": 1, "acc_step": 0, "speed/wps": 1496.377842199158, "speed/FLOPS": 23502670250371.367, "speed/curr_iter_time": 1.8452, "speed/data_load_time": 0.2354, "optim/grad_norm": 3.262824058532715, "optim/lr": 0.0, "optim/total_tokens": 262144, "memory/max_active_gib": 55.36746549606323, "memory/max_active_pct": 69.95169336198055, "memory/max_reserved_gib": 59.234375, "memory/max_reserved_pct": 74.83717738142056, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 110671, "loss/out": 11.935291290283203, "created_at": "2025-01-14T12:58:45.072557+00:00"} {"global_step": 2, "acc_step": 0, "speed/wps": 12951.64715816192, "speed/FLOPS": 203423416047165.47, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.1464505195617676, "optim/lr": 6.000000000000001e-07, "optim/total_tokens": 524288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358963, "loss/out": 11.955284118652344, "created_at": "2025-01-14T12:58:50.141564+00:00"} {"global_step": 3, "acc_step": 0, "speed/wps": 12920.277870622034, "speed/FLOPS": 202930718280433.28, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0011, "optim/grad_norm": 3.3608880043029785, "optim/lr": 1.2000000000000002e-06, "optim/total_tokens": 786432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 11.893197059631348, "created_at": "2025-01-14T12:58:55.220332+00:00"} {"global_step": 4, "acc_step": 0, "speed/wps": 12930.382209247628, "speed/FLOPS": 203089420803364.53, "speed/curr_iter_time": 1.2781, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.2139861583709717, "optim/lr": 1.8e-06, "optim/total_tokens": 1048576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 11.843280792236328, "created_at": "2025-01-14T12:59:00.296904+00:00"} {"global_step": 5, "acc_step": 0, "speed/wps": 12912.157550915412, "speed/FLOPS": 202803177500952.2, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0014, "optim/grad_norm": 3.5547306537628174, "optim/lr": 2.4000000000000003e-06, "optim/total_tokens": 1310720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 383474, "loss/out": 11.674278259277344, "created_at": "2025-01-14T12:59:05.379356+00:00"} {"global_step": 6, "acc_step": 0, "speed/wps": 12889.049657971003, "speed/FLOPS": 202440236288687.8, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.001, "optim/grad_norm": 3.1573727130889893, "optim/lr": 3e-06, "optim/total_tokens": 1572864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 11.583027839660645, "created_at": "2025-01-14T12:59:10.474983+00:00"} {"global_step": 7, "acc_step": 0, "speed/wps": 12866.400916788803, "speed/FLOPS": 202084506685789.56, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.2509303092956543, "optim/lr": 3.6e-06, "optim/total_tokens": 1835008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326201, "loss/out": 11.340741157531738, "created_at": "2025-01-14T12:59:15.573174+00:00"} {"global_step": 8, "acc_step": 0, "speed/wps": 12855.167129274454, "speed/FLOPS": 201908064616035.0, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.20878267288208, "optim/lr": 4.2000000000000004e-06, "optim/total_tokens": 2097152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 11.161297798156738, "created_at": "2025-01-14T12:59:20.675289+00:00"} {"global_step": 9, "acc_step": 0, "speed/wps": 12852.860559351333, "speed/FLOPS": 201871836765833.16, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0018, "optim/grad_norm": 3.0258114337921143, "optim/lr": 4.800000000000001e-06, "optim/total_tokens": 2359296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298121, "loss/out": 10.974069595336914, "created_at": "2025-01-14T12:59:25.779583+00:00"} {"global_step": 10, "acc_step": 0, "speed/wps": 12867.34207820943, "speed/FLOPS": 202099288919191.84, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0013, "optim/grad_norm": 2.839926242828369, "optim/lr": 5.4e-06, "optim/total_tokens": 2621440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 10.696242332458496, "created_at": "2025-01-14T12:59:30.877517+00:00"} {"global_step": 11, "acc_step": 0, "speed/wps": 12854.374234381967, "speed/FLOPS": 201895611112196.72, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.6126978397369385, "optim/lr": 6e-06, "optim/total_tokens": 2883584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 404430, "loss/out": 10.486185073852539, "created_at": "2025-01-14T12:59:35.981785+00:00"} {"global_step": 12, "acc_step": 0, "speed/wps": 12866.77705193082, "speed/FLOPS": 202090414405058.78, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.2158823013305664, "optim/lr": 6.6e-06, "optim/total_tokens": 3145728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 10.36819076538086, "created_at": "2025-01-14T12:59:41.081419+00:00"} {"global_step": 13, "acc_step": 0, "speed/wps": 12877.275307551705, "speed/FLOPS": 202255303935699.75, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.454559326171875, "optim/lr": 7.2e-06, "optim/total_tokens": 3407872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436096, "loss/out": 10.175821304321289, "created_at": "2025-01-14T12:59:46.176994+00:00"} {"global_step": 14, "acc_step": 0, "speed/wps": 12874.804669204112, "speed/FLOPS": 202216499165437.94, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3689119815826416, "optim/lr": 7.8e-06, "optim/total_tokens": 3670016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 10.058097839355469, "created_at": "2025-01-14T12:59:51.272694+00:00"} {"global_step": 15, "acc_step": 0, "speed/wps": 12881.291639492283, "speed/FLOPS": 202318385947844.97, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.001, "optim/grad_norm": 2.0582387447357178, "optim/lr": 8.400000000000001e-06, "optim/total_tokens": 3932160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374658, "loss/out": 9.912544250488281, "created_at": "2025-01-14T12:59:56.366124+00:00"} {"global_step": 16, "acc_step": 0, "speed/wps": 12886.017915935934, "speed/FLOPS": 202392618613976.9, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0017, "optim/grad_norm": 1.819486379623413, "optim/lr": 9e-06, "optim/total_tokens": 4194304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466913, "loss/out": 9.839317321777344, "created_at": "2025-01-14T13:00:01.457197+00:00"} {"global_step": 17, "acc_step": 0, "speed/wps": 12891.886497134457, "speed/FLOPS": 202484792745975.38, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0011, "optim/grad_norm": 1.7943283319473267, "optim/lr": 9.600000000000001e-06, "optim/total_tokens": 4456448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299119, "loss/out": 9.807682991027832, "created_at": "2025-01-14T13:00:06.546195+00:00"} {"global_step": 18, "acc_step": 0, "speed/wps": 12905.729372885915, "speed/FLOPS": 202702214131756.66, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.6287389993667603, "optim/lr": 1.0199999999999999e-05, "optim/total_tokens": 4718592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503389, "loss/out": 9.693235397338867, "created_at": "2025-01-14T13:00:11.632842+00:00"} {"global_step": 19, "acc_step": 0, "speed/wps": 12912.13978601934, "speed/FLOPS": 202802898478847.84, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6596145629882812, "optim/lr": 1.08e-05, "optim/total_tokens": 4980736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501012, "loss/out": 9.60323429107666, "created_at": "2025-01-14T13:00:16.715284+00:00"} {"global_step": 20, "acc_step": 0, "speed/wps": 12915.537343137557, "speed/FLOPS": 202856261782122.2, "speed/curr_iter_time": 1.2774, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.4108303785324097, "optim/lr": 1.1400000000000001e-05, "optim/total_tokens": 5242880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468765, "loss/out": 9.572436332702637, "created_at": "2025-01-14T13:00:21.794491+00:00"} {"global_step": 21, "acc_step": 0, "speed/wps": 12910.379699678844, "speed/FLOPS": 202775253904258.28, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.531826376914978, "optim/lr": 1.2e-05, "optim/total_tokens": 5505024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402800, "loss/out": 9.507811546325684, "created_at": "2025-01-14T13:00:26.876485+00:00"} {"global_step": 22, "acc_step": 0, "speed/wps": 12908.605271885048, "speed/FLOPS": 202747384077437.4, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.494186520576477, "optim/lr": 1.26e-05, "optim/total_tokens": 5767168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 9.371390342712402, "created_at": "2025-01-14T13:00:31.958941+00:00"} {"global_step": 23, "acc_step": 0, "speed/wps": 12922.711304070024, "speed/FLOPS": 202968938696621.94, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2946768999099731, "optim/lr": 1.32e-05, "optim/total_tokens": 6029312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487515, "loss/out": 9.428951263427734, "created_at": "2025-01-14T13:00:37.034535+00:00"} {"global_step": 24, "acc_step": 0, "speed/wps": 12918.133261489093, "speed/FLOPS": 202897034246998.7, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0005, "optim/grad_norm": 1.543128252029419, "optim/lr": 1.38e-05, "optim/total_tokens": 6291456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364643, "loss/out": 9.26694107055664, "created_at": "2025-01-14T13:00:42.115390+00:00"} {"global_step": 25, "acc_step": 0, "speed/wps": 12781.836779749907, "speed/FLOPS": 200756310710292.16, "speed/curr_iter_time": 1.2778, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3205899000167847, "optim/lr": 1.44e-05, "optim/total_tokens": 6553600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 118942, "loss/out": 9.403535842895508, "created_at": "2025-01-14T13:00:47.251402+00:00"} {"global_step": 26, "acc_step": 0, "speed/wps": 7589.89870783862, "speed/FLOPS": 119209788820375.52, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.001, "optim/grad_norm": 1.316868782043457, "optim/lr": 1.5e-05, "optim/total_tokens": 6815744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 86859, "loss/out": 9.290573120117188, "created_at": "2025-01-14T13:00:55.923952+00:00"} {"global_step": 27, "acc_step": 0, "speed/wps": 12961.276565362397, "speed/FLOPS": 203574659119441.1, "speed/curr_iter_time": 1.2722, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.402248740196228, "optim/lr": 1.56e-05, "optim/total_tokens": 7077888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313070, "loss/out": 9.238706588745117, "created_at": "2025-01-14T13:01:00.985012+00:00"} {"global_step": 28, "acc_step": 0, "speed/wps": 12979.020937744384, "speed/FLOPS": 203853358871023.03, "speed/curr_iter_time": 1.2721, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.2717516422271729, "optim/lr": 1.62e-05, "optim/total_tokens": 7340032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358652, "loss/out": 9.171380996704102, "created_at": "2025-01-14T13:01:06.043765+00:00"} {"global_step": 29, "acc_step": 0, "speed/wps": 12963.763444024888, "speed/FLOPS": 203613718966164.88, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3491601943969727, "optim/lr": 1.6800000000000002e-05, "optim/total_tokens": 7602176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497589, "loss/out": 9.13576602935791, "created_at": "2025-01-14T13:01:11.106047+00:00"} {"global_step": 30, "acc_step": 0, "speed/wps": 12955.057868166594, "speed/FLOPS": 203476985934595.56, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3647795915603638, "optim/lr": 1.74e-05, "optim/total_tokens": 7864320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486988, "loss/out": 9.101506233215332, "created_at": "2025-01-14T13:01:16.168807+00:00"} {"global_step": 31, "acc_step": 0, "speed/wps": 12946.572937972385, "speed/FLOPS": 203343718446379.6, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3578506708145142, "optim/lr": 1.8e-05, "optim/total_tokens": 8126464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487515, "loss/out": 9.14272403717041, "created_at": "2025-01-14T13:01:21.235121+00:00"} {"global_step": 32, "acc_step": 0, "speed/wps": 12954.643408461625, "speed/FLOPS": 203470476275401.7, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3465275764465332, "optim/lr": 1.8599999999999998e-05, "optim/total_tokens": 8388608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495975, "loss/out": 9.09735107421875, "created_at": "2025-01-14T13:01:26.300996+00:00"} {"global_step": 33, "acc_step": 0, "speed/wps": 12957.506077425824, "speed/FLOPS": 203515438425203.8, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4689499139785767, "optim/lr": 1.9200000000000003e-05, "optim/total_tokens": 8650752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499609, "loss/out": 8.989468574523926, "created_at": "2025-01-14T13:01:31.364431+00:00"} {"global_step": 34, "acc_step": 0, "speed/wps": 12946.957643761025, "speed/FLOPS": 203349760779431.3, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3527109622955322, "optim/lr": 1.98e-05, "optim/total_tokens": 8912896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332341, "loss/out": 9.006418228149414, "created_at": "2025-01-14T13:01:36.431281+00:00"} {"global_step": 35, "acc_step": 0, "speed/wps": 12958.578561866747, "speed/FLOPS": 203532283267096.56, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4754406213760376, "optim/lr": 2.0399999999999998e-05, "optim/total_tokens": 9175040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364358, "loss/out": 8.936878204345703, "created_at": "2025-01-14T13:01:41.492680+00:00"} {"global_step": 36, "acc_step": 0, "speed/wps": 12961.347601295485, "speed/FLOPS": 203575774836383.84, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3339836597442627, "optim/lr": 2.1000000000000002e-05, "optim/total_tokens": 9437184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480094, "loss/out": 8.861126899719238, "created_at": "2025-01-14T13:01:46.555588+00:00"} {"global_step": 37, "acc_step": 0, "speed/wps": 12955.161449671315, "speed/FLOPS": 203478612824463.1, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0005, "optim/grad_norm": 1.4650343656539917, "optim/lr": 2.16e-05, "optim/total_tokens": 9699328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486175, "loss/out": 8.882704734802246, "created_at": "2025-01-14T13:01:51.620779+00:00"} {"global_step": 38, "acc_step": 0, "speed/wps": 12961.860785332623, "speed/FLOPS": 203583835096875.84, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.375938892364502, "optim/lr": 2.22e-05, "optim/total_tokens": 9961472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493362, "loss/out": 8.899469375610352, "created_at": "2025-01-14T13:01:56.683729+00:00"} {"global_step": 39, "acc_step": 0, "speed/wps": 12964.889340952492, "speed/FLOPS": 203631402724557.4, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3738009929656982, "optim/lr": 2.2800000000000002e-05, "optim/total_tokens": 10223616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 8.75922966003418, "created_at": "2025-01-14T13:02:01.746722+00:00"} {"global_step": 40, "acc_step": 0, "speed/wps": 12959.837154307903, "speed/FLOPS": 203552051190872.2, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3017040491104126, "optim/lr": 2.34e-05, "optim/total_tokens": 10485760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314573, "loss/out": 8.83224868774414, "created_at": "2025-01-14T13:02:06.810201+00:00"} {"global_step": 41, "acc_step": 0, "speed/wps": 12961.644060280278, "speed/FLOPS": 203580431132117.34, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2920515537261963, "optim/lr": 2.4e-05, "optim/total_tokens": 10747904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462126, "loss/out": 8.70849609375, "created_at": "2025-01-14T13:02:11.873161+00:00"} {"global_step": 42, "acc_step": 0, "speed/wps": 12943.88768744219, "speed/FLOPS": 203301542896881.97, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.408813238143921, "optim/lr": 2.46e-05, "optim/total_tokens": 11010048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499609, "loss/out": 8.658443450927734, "created_at": "2025-01-14T13:02:16.942542+00:00"} {"global_step": 43, "acc_step": 0, "speed/wps": 12958.98154135159, "speed/FLOPS": 203538612613657.88, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1215037107467651, "optim/lr": 2.52e-05, "optim/total_tokens": 11272192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481431, "loss/out": 8.665678024291992, "created_at": "2025-01-14T13:02:22.008246+00:00"} {"global_step": 44, "acc_step": 0, "speed/wps": 12956.161473573799, "speed/FLOPS": 203494319574028.44, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3439092636108398, "optim/lr": 2.58e-05, "optim/total_tokens": 11534336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 8.571155548095703, "created_at": "2025-01-14T13:02:27.070396+00:00"} {"global_step": 45, "acc_step": 0, "speed/wps": 12955.082167181352, "speed/FLOPS": 203477367584012.66, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3159161806106567, "optim/lr": 2.64e-05, "optim/total_tokens": 11796480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 8.493266105651855, "created_at": "2025-01-14T13:02:32.134117+00:00"} {"global_step": 46, "acc_step": 0, "speed/wps": 12943.002938131758, "speed/FLOPS": 203287646693188.7, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0594969987869263, "optim/lr": 2.7e-05, "optim/total_tokens": 12058624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 8.478338241577148, "created_at": "2025-01-14T13:02:37.201507+00:00"} {"global_step": 47, "acc_step": 0, "speed/wps": 12959.98415531687, "speed/FLOPS": 203554360043718.56, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1404788494110107, "optim/lr": 2.76e-05, "optim/total_tokens": 12320768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315181, "loss/out": 8.480193138122559, "created_at": "2025-01-14T13:02:42.265360+00:00"} {"global_step": 48, "acc_step": 0, "speed/wps": 12955.716967112607, "speed/FLOPS": 203487337989239.53, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1626590490341187, "optim/lr": 2.82e-05, "optim/total_tokens": 12582912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475987, "loss/out": 8.31144905090332, "created_at": "2025-01-14T13:02:47.327547+00:00"} {"global_step": 49, "acc_step": 0, "speed/wps": 12953.353453442292, "speed/FLOPS": 203450215759239.12, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.319288730621338, "optim/lr": 2.88e-05, "optim/total_tokens": 12845056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 8.40835189819336, "created_at": "2025-01-14T13:02:52.393094+00:00"} {"global_step": 50, "acc_step": 0, "speed/wps": 12950.675288209291, "speed/FLOPS": 203408151494070.16, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9107226133346558, "optim/lr": 2.94e-05, "optim/total_tokens": 13107200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 8.283149719238281, "created_at": "2025-01-14T13:02:57.458630+00:00"} {"global_step": 51, "acc_step": 0, "speed/wps": 12951.860767044695, "speed/FLOPS": 203426771068198.8, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3931138515472412, "optim/lr": 3e-05, "optim/total_tokens": 13369344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493533, "loss/out": 8.278526306152344, "created_at": "2025-01-14T13:03:02.522405+00:00"} {"global_step": 52, "acc_step": 0, "speed/wps": 12946.44246405235, "speed/FLOPS": 203341669174175.47, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3321785926818848, "optim/lr": 3.0600000000000005e-05, "optim/total_tokens": 13631488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325437, "loss/out": 8.214118003845215, "created_at": "2025-01-14T13:03:07.588051+00:00"} {"global_step": 53, "acc_step": 0, "speed/wps": 12943.433761589402, "speed/FLOPS": 203294413367604.44, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.970695436000824, "optim/lr": 3.12e-05, "optim/total_tokens": 13893632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 373449, "loss/out": 8.146632194519043, "created_at": "2025-01-14T13:03:12.655471+00:00"} {"global_step": 54, "acc_step": 0, "speed/wps": 12935.808307707519, "speed/FLOPS": 203174645135917.78, "speed/curr_iter_time": 1.2785, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2688077688217163, "optim/lr": 3.18e-05, "optim/total_tokens": 14155776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490175, "loss/out": 8.061421394348145, "created_at": "2025-01-14T13:03:17.725458+00:00"} {"global_step": 55, "acc_step": 0, "speed/wps": 12936.56530435016, "speed/FLOPS": 203186534808412.84, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3696190118789673, "optim/lr": 3.24e-05, "optim/total_tokens": 14417920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477198, "loss/out": 8.073787689208984, "created_at": "2025-01-14T13:03:22.797941+00:00"} {"global_step": 56, "acc_step": 0, "speed/wps": 12930.65452614041, "speed/FLOPS": 203093697914368.3, "speed/curr_iter_time": 1.2783, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0927879810333252, "optim/lr": 3.2999999999999996e-05, "optim/total_tokens": 14680064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478008, "loss/out": 8.081428527832031, "created_at": "2025-01-14T13:03:27.870274+00:00"} {"global_step": 57, "acc_step": 0, "speed/wps": 12933.031091242177, "speed/FLOPS": 203131025135034.25, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8684508800506592, "optim/lr": 3.3600000000000004e-05, "optim/total_tokens": 14942208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287866, "loss/out": 8.011983871459961, "created_at": "2025-01-14T13:03:32.944648+00:00"} {"global_step": 58, "acc_step": 0, "speed/wps": 12933.919493794125, "speed/FLOPS": 203144978717905.6, "speed/curr_iter_time": 1.2782, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5089294910430908, "optim/lr": 3.4200000000000005e-05, "optim/total_tokens": 15204352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484611, "loss/out": 7.964122772216797, "created_at": "2025-01-14T13:03:38.016887+00:00"} {"global_step": 59, "acc_step": 0, "speed/wps": 12927.37543963129, "speed/FLOPS": 203042195354805.53, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5625263452529907, "optim/lr": 3.48e-05, "optim/total_tokens": 15466496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497825, "loss/out": 8.013906478881836, "created_at": "2025-01-14T13:03:43.092088+00:00"} {"global_step": 60, "acc_step": 0, "speed/wps": 12921.984537813561, "speed/FLOPS": 202957523833882.72, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5262891054153442, "optim/lr": 3.54e-05, "optim/total_tokens": 15728640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479574, "loss/out": 7.883946418762207, "created_at": "2025-01-14T13:03:48.167684+00:00"} {"global_step": 61, "acc_step": 0, "speed/wps": 12927.116744825002, "speed/FLOPS": 203038132197388.84, "speed/curr_iter_time": 1.2784, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2710356712341309, "optim/lr": 3.6e-05, "optim/total_tokens": 15990784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473004, "loss/out": 7.817775726318359, "created_at": "2025-01-14T13:03:53.243428+00:00"} {"global_step": 62, "acc_step": 0, "speed/wps": 12932.390709647254, "speed/FLOPS": 203120967062109.44, "speed/curr_iter_time": 1.2782, "speed/data_load_time": 0.0005, "optim/grad_norm": 1.5327122211456299, "optim/lr": 3.66e-05, "optim/total_tokens": 16252928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 388985, "loss/out": 8.017232894897461, "created_at": "2025-01-14T13:03:58.314940+00:00"} {"global_step": 63, "acc_step": 0, "speed/wps": 12931.873827204074, "speed/FLOPS": 203112848713068.38, "speed/curr_iter_time": 1.2782, "speed/data_load_time": 0.0005, "optim/grad_norm": 1.4360063076019287, "optim/lr": 3.7199999999999996e-05, "optim/total_tokens": 16515072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491159, "loss/out": 7.841887474060059, "created_at": "2025-01-14T13:04:03.390398+00:00"} {"global_step": 64, "acc_step": 0, "speed/wps": 12922.15378206404, "speed/FLOPS": 202960182047402.47, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8337390422821045, "optim/lr": 3.7800000000000004e-05, "optim/total_tokens": 16777216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487515, "loss/out": 7.728095054626465, "created_at": "2025-01-14T13:04:08.469036+00:00"} {"global_step": 65, "acc_step": 0, "speed/wps": 12924.243976894762, "speed/FLOPS": 202993011429447.62, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4064942598342896, "optim/lr": 3.8400000000000005e-05, "optim/total_tokens": 17039360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296653, "loss/out": 7.70404052734375, "created_at": "2025-01-14T13:04:13.546484+00:00"} {"global_step": 66, "acc_step": 0, "speed/wps": 12923.648795009989, "speed/FLOPS": 202983663280081.66, "speed/curr_iter_time": 1.2784, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3978575468063354, "optim/lr": 3.9e-05, "optim/total_tokens": 17301504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490411, "loss/out": 7.655619144439697, "created_at": "2025-01-14T13:04:18.621080+00:00"} {"global_step": 67, "acc_step": 0, "speed/wps": 12922.526826540598, "speed/FLOPS": 202966041223523.62, "speed/curr_iter_time": 1.2779, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8992419242858887, "optim/lr": 3.96e-05, "optim/total_tokens": 17563648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491222, "loss/out": 7.668068885803223, "created_at": "2025-01-14T13:04:23.698191+00:00"} {"global_step": 68, "acc_step": 0, "speed/wps": 12920.195230792742, "speed/FLOPS": 202929420308355.25, "speed/curr_iter_time": 1.279, "speed/data_load_time": 0.0024, "optim/grad_norm": 1.453015923500061, "optim/lr": 4.02e-05, "optim/total_tokens": 17825792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326344, "loss/out": 7.647052764892578, "created_at": "2025-01-14T13:04:28.775289+00:00"} {"global_step": 69, "acc_step": 0, "speed/wps": 12909.825614350855, "speed/FLOPS": 202766551232789.2, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2380461692810059, "optim/lr": 4.0799999999999996e-05, "optim/total_tokens": 18087936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496190, "loss/out": 7.634332180023193, "created_at": "2025-01-14T13:04:33.858333+00:00"} {"global_step": 70, "acc_step": 0, "speed/wps": 12901.522181031036, "speed/FLOPS": 202636134402388.34, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.5686107873916626, "optim/lr": 4.1400000000000003e-05, "optim/total_tokens": 18350080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494118, "loss/out": 7.666420936584473, "created_at": "2025-01-14T13:04:38.945658+00:00"} {"global_step": 71, "acc_step": 0, "speed/wps": 12905.606180119566, "speed/FLOPS": 202700279220076.56, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3826687335968018, "optim/lr": 4.2000000000000004e-05, "optim/total_tokens": 18612224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313679, "loss/out": 7.566954612731934, "created_at": "2025-01-14T13:04:44.033094+00:00"} {"global_step": 72, "acc_step": 0, "speed/wps": 12913.299710296484, "speed/FLOPS": 202821116683523.38, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3169530630111694, "optim/lr": 4.2600000000000005e-05, "optim/total_tokens": 18874368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481190, "loss/out": 7.52208948135376, "created_at": "2025-01-14T13:04:49.115852+00:00"} {"global_step": 73, "acc_step": 0, "speed/wps": 12901.056791475947, "speed/FLOPS": 202628824819913.4, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3877025842666626, "optim/lr": 4.32e-05, "optim/total_tokens": 19136512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467117, "loss/out": 7.646997451782227, "created_at": "2025-01-14T13:04:54.201974+00:00"} {"global_step": 74, "acc_step": 0, "speed/wps": 12918.51518088539, "speed/FLOPS": 202903032815929.78, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0005, "optim/grad_norm": 1.345410943031311, "optim/lr": 4.38e-05, "optim/total_tokens": 19398656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469548, "loss/out": 7.531191825866699, "created_at": "2025-01-14T13:04:59.283138+00:00"} {"global_step": 75, "acc_step": 0, "speed/wps": 12893.156891447727, "speed/FLOPS": 202504746034370.0, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3888537883758545, "optim/lr": 4.44e-05, "optim/total_tokens": 19660800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488035, "loss/out": 7.451564788818359, "created_at": "2025-01-14T13:05:04.370743+00:00"} {"global_step": 76, "acc_step": 0, "speed/wps": 12908.161020145068, "speed/FLOPS": 202740406493395.16, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4801114797592163, "optim/lr": 4.4999999999999996e-05, "optim/total_tokens": 19922944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500702, "loss/out": 7.397028923034668, "created_at": "2025-01-14T13:05:09.456722+00:00"} {"global_step": 77, "acc_step": 0, "speed/wps": 12888.23548312694, "speed/FLOPS": 202427448553968.3, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4419937133789062, "optim/lr": 4.5600000000000004e-05, "optim/total_tokens": 20185088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 414557, "loss/out": 7.49977970123291, "created_at": "2025-01-14T13:05:14.545426+00:00"} {"global_step": 78, "acc_step": 0, "speed/wps": 12905.727791777257, "speed/FLOPS": 202702189298272.53, "speed/curr_iter_time": 1.2783, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.294921636581421, "optim/lr": 4.6200000000000005e-05, "optim/total_tokens": 20447232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477481, "loss/out": 7.432483673095703, "created_at": "2025-01-14T13:05:19.629726+00:00"} {"global_step": 79, "acc_step": 0, "speed/wps": 12911.504143862312, "speed/FLOPS": 202792914845307.88, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5228075981140137, "optim/lr": 4.68e-05, "optim/total_tokens": 20709376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480621, "loss/out": 7.389868259429932, "created_at": "2025-01-14T13:05:24.711583+00:00"} {"global_step": 80, "acc_step": 0, "speed/wps": 12899.61485219557, "speed/FLOPS": 202606177182082.88, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3931857347488403, "optim/lr": 4.740000000000001e-05, "optim/total_tokens": 20971520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483045, "loss/out": 7.496644973754883, "created_at": "2025-01-14T13:05:29.795912+00:00"} {"global_step": 81, "acc_step": 0, "speed/wps": 12901.280270397796, "speed/FLOPS": 202632334863473.28, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9746453166007996, "optim/lr": 4.8e-05, "optim/total_tokens": 21233664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489891, "loss/out": 7.26727294921875, "created_at": "2025-01-14T13:05:34.879637+00:00"} {"global_step": 82, "acc_step": 0, "speed/wps": 12901.888308302374, "speed/FLOPS": 202641884934296.3, "speed/curr_iter_time": 1.2789, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.474733591079712, "optim/lr": 4.8599999999999995e-05, "optim/total_tokens": 21495808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 343329, "loss/out": 7.3300251960754395, "created_at": "2025-01-14T13:05:39.966016+00:00"} {"global_step": 83, "acc_step": 0, "speed/wps": 12897.725150514896, "speed/FLOPS": 202576496820465.7, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3114824295043945, "optim/lr": 4.92e-05, "optim/total_tokens": 21757952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480385, "loss/out": 7.313185691833496, "created_at": "2025-01-14T13:05:45.051926+00:00"} {"global_step": 84, "acc_step": 0, "speed/wps": 12908.446135821736, "speed/FLOPS": 202744884626887.22, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8774034976959229, "optim/lr": 4.9800000000000004e-05, "optim/total_tokens": 22020096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481715, "loss/out": 7.324306488037109, "created_at": "2025-01-14T13:05:50.134699+00:00"} {"global_step": 85, "acc_step": 0, "speed/wps": 12895.544656407348, "speed/FLOPS": 202542249164193.75, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.392155647277832, "optim/lr": 5.04e-05, "optim/total_tokens": 22282240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367853, "loss/out": 7.376789093017578, "created_at": "2025-01-14T13:05:55.225051+00:00"} {"global_step": 86, "acc_step": 0, "speed/wps": 12900.208718682701, "speed/FLOPS": 202615504671321.34, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2234783172607422, "optim/lr": 5.1000000000000006e-05, "optim/total_tokens": 22544384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501464, "loss/out": 7.257665634155273, "created_at": "2025-01-14T13:06:00.311565+00:00"} {"global_step": 87, "acc_step": 0, "speed/wps": 12893.886187981376, "speed/FLOPS": 202516200638589.4, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4785438776016235, "optim/lr": 5.16e-05, "optim/total_tokens": 22806528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495739, "loss/out": 7.254939079284668, "created_at": "2025-01-14T13:06:05.398799+00:00"} {"global_step": 88, "acc_step": 0, "speed/wps": 12891.830240366815, "speed/FLOPS": 202483909156134.66, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3727775812149048, "optim/lr": 5.2199999999999995e-05, "optim/total_tokens": 23068672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483045, "loss/out": 7.18730354309082, "created_at": "2025-01-14T13:06:10.486326+00:00"} {"global_step": 89, "acc_step": 0, "speed/wps": 12893.149750158198, "speed/FLOPS": 202504633870604.78, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3722891807556152, "optim/lr": 5.28e-05, "optim/total_tokens": 23330816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498115, "loss/out": 7.213588237762451, "created_at": "2025-01-14T13:06:15.576067+00:00"} {"global_step": 90, "acc_step": 0, "speed/wps": 12897.083305981878, "speed/FLOPS": 202566415770087.8, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3750687837600708, "optim/lr": 5.3400000000000004e-05, "optim/total_tokens": 23592960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300465, "loss/out": 7.211299419403076, "created_at": "2025-01-14T13:06:20.664811+00:00"} {"global_step": 91, "acc_step": 0, "speed/wps": 12898.93283209069, "speed/FLOPS": 202595465119141.9, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5399205684661865, "optim/lr": 5.4e-05, "optim/total_tokens": 23855104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490113, "loss/out": 7.181225776672363, "created_at": "2025-01-14T13:06:25.749332+00:00"} {"global_step": 92, "acc_step": 0, "speed/wps": 12893.817155912766, "speed/FLOPS": 202515116395091.88, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4071744680404663, "optim/lr": 5.4600000000000006e-05, "optim/total_tokens": 24117248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466077, "loss/out": 7.253214359283447, "created_at": "2025-01-14T13:06:30.837422+00:00"} {"global_step": 93, "acc_step": 0, "speed/wps": 12885.614938166476, "speed/FLOPS": 202386289294358.1, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0898702144622803, "optim/lr": 5.52e-05, "optim/total_tokens": 24379392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479338, "loss/out": 7.150873184204102, "created_at": "2025-01-14T13:06:35.932216+00:00"} {"global_step": 94, "acc_step": 0, "speed/wps": 12889.878201551166, "speed/FLOPS": 202453249704152.47, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9067450761795044, "optim/lr": 5.5799999999999994e-05, "optim/total_tokens": 24641536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496480, "loss/out": 7.283794403076172, "created_at": "2025-01-14T13:06:41.020047+00:00"} {"global_step": 95, "acc_step": 0, "speed/wps": 12896.106315455198, "speed/FLOPS": 202551070791340.62, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.607743740081787, "optim/lr": 5.64e-05, "optim/total_tokens": 24903680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 392428, "loss/out": 7.2050981521606445, "created_at": "2025-01-14T13:06:46.109051+00:00"} {"global_step": 96, "acc_step": 0, "speed/wps": 12887.339611193072, "speed/FLOPS": 202413377654189.88, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4514870643615723, "optim/lr": 5.7e-05, "optim/total_tokens": 25165824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483518, "loss/out": 7.216063499450684, "created_at": "2025-01-14T13:06:51.205928+00:00"} {"global_step": 97, "acc_step": 0, "speed/wps": 12891.994607208615, "speed/FLOPS": 202486490763248.8, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4434500932693481, "optim/lr": 5.76e-05, "optim/total_tokens": 25427968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301342, "loss/out": 7.168144226074219, "created_at": "2025-01-14T13:06:56.295467+00:00"} {"global_step": 98, "acc_step": 0, "speed/wps": 12892.054750761734, "speed/FLOPS": 202487435400396.47, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2073103189468384, "optim/lr": 5.8200000000000005e-05, "optim/total_tokens": 25690112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502556, "loss/out": 7.089415073394775, "created_at": "2025-01-14T13:07:01.385026+00:00"} {"global_step": 99, "acc_step": 0, "speed/wps": 12881.190652970416, "speed/FLOPS": 202316799815749.0, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4158687591552734, "optim/lr": 5.88e-05, "optim/total_tokens": 25952256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496999, "loss/out": 7.197681903839111, "created_at": "2025-01-14T13:07:06.476627+00:00"} {"global_step": 100, "acc_step": 0, "speed/wps": 12882.04064634688, "speed/FLOPS": 202330150129739.7, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3112713098526, "optim/lr": 5.940000000000001e-05, "optim/total_tokens": 26214400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361148, "loss/out": 7.067574501037598, "created_at": "2025-01-14T13:07:11.568671+00:00"} {"global_step": 101, "acc_step": 0, "speed/wps": 12879.530266239359, "speed/FLOPS": 202290721160527.3, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.056966781616211, "optim/lr": 6e-05, "optim/total_tokens": 26476544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471114, "loss/out": 7.115772247314453, "created_at": "2025-01-14T13:07:16.666819+00:00"} {"global_step": 102, "acc_step": 0, "speed/wps": 12876.23295237627, "speed/FLOPS": 202238932315323.4, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3173903226852417, "optim/lr": 6.0599999999999996e-05, "optim/total_tokens": 26738688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319909, "loss/out": 7.161823272705078, "created_at": "2025-01-14T13:07:21.761118+00:00"} {"global_step": 103, "acc_step": 0, "speed/wps": 12879.613876536445, "speed/FLOPS": 202292034375135.62, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8688938021659851, "optim/lr": 6.120000000000001e-05, "optim/total_tokens": 27000832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484611, "loss/out": 6.945498943328857, "created_at": "2025-01-14T13:07:26.853080+00:00"} {"global_step": 104, "acc_step": 0, "speed/wps": 12885.807016058998, "speed/FLOPS": 202389306141602.84, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9920310974121094, "optim/lr": 6.18e-05, "optim/total_tokens": 27262976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332103, "loss/out": 7.035680294036865, "created_at": "2025-01-14T13:07:31.947679+00:00"} {"global_step": 105, "acc_step": 0, "speed/wps": 12870.383215136428, "speed/FLOPS": 202147054153590.38, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0198208093643188, "optim/lr": 6.24e-05, "optim/total_tokens": 27525120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489068, "loss/out": 6.919334411621094, "created_at": "2025-01-14T13:07:37.043467+00:00"} {"global_step": 106, "acc_step": 0, "speed/wps": 12877.264050542466, "speed/FLOPS": 202255127128900.9, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.314629077911377, "optim/lr": 6.3e-05, "optim/total_tokens": 27787264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484902, "loss/out": 7.109885215759277, "created_at": "2025-01-14T13:07:42.136568+00:00"} {"global_step": 107, "acc_step": 0, "speed/wps": 12874.233892647931, "speed/FLOPS": 202207534335293.94, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8089399933815002, "optim/lr": 6.36e-05, "optim/total_tokens": 28049408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483754, "loss/out": 7.031657695770264, "created_at": "2025-01-14T13:07:47.232324+00:00"} {"global_step": 108, "acc_step": 0, "speed/wps": 12867.402443719893, "speed/FLOPS": 202100237042484.53, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3719568252563477, "optim/lr": 6.42e-05, "optim/total_tokens": 28311552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469257, "loss/out": 7.081680774688721, "created_at": "2025-01-14T13:07:52.329077+00:00"} {"global_step": 109, "acc_step": 0, "speed/wps": 12872.998356689535, "speed/FLOPS": 202188128545263.0, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3290098905563354, "optim/lr": 6.48e-05, "optim/total_tokens": 28573696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498328, "loss/out": 6.930727958679199, "created_at": "2025-01-14T13:07:57.424235+00:00"} {"global_step": 110, "acc_step": 0, "speed/wps": 12872.070502322566, "speed/FLOPS": 202173555317424.25, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3228459358215332, "optim/lr": 6.54e-05, "optim/total_tokens": 28835840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469257, "loss/out": 6.944452285766602, "created_at": "2025-01-14T13:08:02.519110+00:00"} {"global_step": 111, "acc_step": 0, "speed/wps": 12870.42284690108, "speed/FLOPS": 202147676624914.0, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4572021961212158, "optim/lr": 6.599999999999999e-05, "optim/total_tokens": 29097984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467927, "loss/out": 6.933067321777344, "created_at": "2025-01-14T13:08:07.615036+00:00"} {"global_step": 112, "acc_step": 0, "speed/wps": 12869.71633596224, "speed/FLOPS": 202136579899771.72, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3807687759399414, "optim/lr": 6.66e-05, "optim/total_tokens": 29360128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476195, "loss/out": 7.022397518157959, "created_at": "2025-01-14T13:08:12.717261+00:00"} {"global_step": 113, "acc_step": 0, "speed/wps": 12872.76643000952, "speed/FLOPS": 202184485818052.72, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9437456727027893, "optim/lr": 6.720000000000001e-05, "optim/total_tokens": 29622272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356884, "loss/out": 6.827639102935791, "created_at": "2025-01-14T13:08:17.812518+00:00"} {"global_step": 114, "acc_step": 0, "speed/wps": 12877.577741294104, "speed/FLOPS": 202260054073211.62, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.535727620124817, "optim/lr": 6.78e-05, "optim/total_tokens": 29884416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481758, "loss/out": 7.221077919006348, "created_at": "2025-01-14T13:08:22.909044+00:00"} {"global_step": 115, "acc_step": 0, "speed/wps": 12868.97443338992, "speed/FLOPS": 202124927300392.5, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.234896183013916, "optim/lr": 6.840000000000001e-05, "optim/total_tokens": 30146560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306094, "loss/out": 6.950098514556885, "created_at": "2025-01-14T13:08:28.006181+00:00"} {"global_step": 116, "acc_step": 0, "speed/wps": 12865.247877947679, "speed/FLOPS": 202066396626348.28, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3357961177825928, "optim/lr": 6.9e-05, "optim/total_tokens": 30408704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475631, "loss/out": 7.055054664611816, "created_at": "2025-01-14T13:08:33.105398+00:00"} {"global_step": 117, "acc_step": 0, "speed/wps": 12875.156508839942, "speed/FLOPS": 202222025290396.97, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1842031478881836, "optim/lr": 6.96e-05, "optim/total_tokens": 30670848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335320, "loss/out": 6.910458564758301, "created_at": "2025-01-14T13:08:38.201406+00:00"} {"global_step": 118, "acc_step": 0, "speed/wps": 12873.758408609416, "speed/FLOPS": 202200066205085.5, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0385299921035767, "optim/lr": 7.02e-05, "optim/total_tokens": 30932992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479574, "loss/out": 6.881363391876221, "created_at": "2025-01-14T13:08:43.299991+00:00"} {"global_step": 119, "acc_step": 0, "speed/wps": 12869.556312001881, "speed/FLOPS": 202134066503578.34, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1050199270248413, "optim/lr": 7.08e-05, "optim/total_tokens": 31195136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333446, "loss/out": 6.899198532104492, "created_at": "2025-01-14T13:08:48.396256+00:00"} {"global_step": 120, "acc_step": 0, "speed/wps": 12874.047810036447, "speed/FLOPS": 202204611652176.28, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8284894824028015, "optim/lr": 7.14e-05, "optim/total_tokens": 31457280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479338, "loss/out": 6.896834373474121, "created_at": "2025-01-14T13:08:53.498935+00:00"} {"global_step": 121, "acc_step": 0, "speed/wps": 12870.835762920973, "speed/FLOPS": 202154162038412.5, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9612516164779663, "optim/lr": 7.2e-05, "optim/total_tokens": 31719424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361148, "loss/out": 6.805120944976807, "created_at": "2025-01-14T13:08:58.594444+00:00"} {"global_step": 122, "acc_step": 0, "speed/wps": 12869.743938308862, "speed/FLOPS": 202137013432555.03, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0811676979064941, "optim/lr": 7.26e-05, "optim/total_tokens": 31981568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488318, "loss/out": 6.774514198303223, "created_at": "2025-01-14T13:09:03.694449+00:00"} {"global_step": 123, "acc_step": 0, "speed/wps": 12872.718296974648, "speed/FLOPS": 202183729822598.3, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8134856224060059, "optim/lr": 7.32e-05, "optim/total_tokens": 32243712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357700, "loss/out": 6.8499298095703125, "created_at": "2025-01-14T13:09:08.791476+00:00"} {"global_step": 124, "acc_step": 0, "speed/wps": 12871.601942336532, "speed/FLOPS": 202166195938974.7, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3550691604614258, "optim/lr": 7.38e-05, "optim/total_tokens": 32505856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484091, "loss/out": 6.788802623748779, "created_at": "2025-01-14T13:09:13.888113+00:00"} {"global_step": 125, "acc_step": 0, "speed/wps": 12861.119277984868, "speed/FLOPS": 202001551290641.47, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7756983637809753, "optim/lr": 7.439999999999999e-05, "optim/total_tokens": 32768000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324857, "loss/out": 6.785540580749512, "created_at": "2025-01-14T13:09:18.988778+00:00"} {"global_step": 126, "acc_step": 0, "speed/wps": 12872.227702003873, "speed/FLOPS": 202176024354434.16, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.018935203552246, "optim/lr": 7.500000000000001e-05, "optim/total_tokens": 33030144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485138, "loss/out": 6.729659080505371, "created_at": "2025-01-14T13:09:24.084207+00:00"} {"global_step": 127, "acc_step": 0, "speed/wps": 12875.699868766427, "speed/FLOPS": 202230559504697.22, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8372176289558411, "optim/lr": 7.560000000000001e-05, "optim/total_tokens": 33292288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487456, "loss/out": 6.8167524337768555, "created_at": "2025-01-14T13:09:29.178571+00:00"} {"global_step": 128, "acc_step": 0, "speed/wps": 12869.54312199476, "speed/FLOPS": 202133859336391.53, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1639796495437622, "optim/lr": 7.62e-05, "optim/total_tokens": 33554432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467961, "loss/out": 6.758498668670654, "created_at": "2025-01-14T13:09:34.277253+00:00"} {"global_step": 129, "acc_step": 0, "speed/wps": 12859.347392419948, "speed/FLOPS": 201973721393018.5, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.057863712310791, "optim/lr": 7.680000000000001e-05, "optim/total_tokens": 33816576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484611, "loss/out": 6.825688362121582, "created_at": "2025-01-14T13:09:39.380929+00:00"} {"global_step": 130, "acc_step": 0, "speed/wps": 12866.961529246386, "speed/FLOPS": 202093311874798.12, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7421848177909851, "optim/lr": 7.74e-05, "optim/total_tokens": 34078720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487515, "loss/out": 6.627368927001953, "created_at": "2025-01-14T13:09:44.487086+00:00"} {"global_step": 131, "acc_step": 0, "speed/wps": 12863.442247124583, "speed/FLOPS": 202038036713074.8, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8946439027786255, "optim/lr": 7.8e-05, "optim/total_tokens": 34340864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479054, "loss/out": 6.663169860839844, "created_at": "2025-01-14T13:09:49.586626+00:00"} {"global_step": 132, "acc_step": 0, "speed/wps": 12866.814812285276, "speed/FLOPS": 202091007483313.7, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.165084958076477, "optim/lr": 7.86e-05, "optim/total_tokens": 34603008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500702, "loss/out": 6.73529577255249, "created_at": "2025-01-14T13:09:54.683657+00:00"} {"global_step": 133, "acc_step": 0, "speed/wps": 12859.321935495094, "speed/FLOPS": 201973321557032.2, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0041314363479614, "optim/lr": 7.92e-05, "optim/total_tokens": 34865152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490411, "loss/out": 6.682906150817871, "created_at": "2025-01-14T13:09:59.786482+00:00"} {"global_step": 134, "acc_step": 0, "speed/wps": 12871.363203063918, "speed/FLOPS": 202162446210636.16, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4895459413528442, "optim/lr": 7.98e-05, "optim/total_tokens": 35127296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499373, "loss/out": 6.64188814163208, "created_at": "2025-01-14T13:10:04.881607+00:00"} {"global_step": 135, "acc_step": 0, "speed/wps": 12865.907714586749, "speed/FLOPS": 202076760267475.44, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0746266841888428, "optim/lr": 8.04e-05, "optim/total_tokens": 35389440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474821, "loss/out": 6.741091251373291, "created_at": "2025-01-14T13:10:09.980069+00:00"} {"global_step": 136, "acc_step": 0, "speed/wps": 12861.978254365553, "speed/FLOPS": 202015042695057.38, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4066874980926514, "optim/lr": 8.1e-05, "optim/total_tokens": 35651584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502030, "loss/out": 6.653233528137207, "created_at": "2025-01-14T13:10:15.079352+00:00"} {"global_step": 137, "acc_step": 0, "speed/wps": 12865.667568142375, "speed/FLOPS": 202072988437570.75, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4366662502288818, "optim/lr": 8.159999999999999e-05, "optim/total_tokens": 35913728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468737, "loss/out": 6.66969108581543, "created_at": "2025-01-14T13:10:20.180212+00:00"} {"global_step": 138, "acc_step": 0, "speed/wps": 12868.36554502918, "speed/FLOPS": 202115363871986.78, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0561178922653198, "optim/lr": 8.22e-05, "optim/total_tokens": 36175872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499090, "loss/out": 6.707050323486328, "created_at": "2025-01-14T13:10:25.279949+00:00"} {"global_step": 139, "acc_step": 0, "speed/wps": 12863.465063224137, "speed/FLOPS": 202038395071270.9, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.4200072288513184, "optim/lr": 8.280000000000001e-05, "optim/total_tokens": 36438016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495151, "loss/out": 6.783594131469727, "created_at": "2025-01-14T13:10:30.378316+00:00"} {"global_step": 140, "acc_step": 0, "speed/wps": 12867.475263782295, "speed/FLOPS": 202101380781629.8, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3186663389205933, "optim/lr": 8.34e-05, "optim/total_tokens": 36700160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427589, "loss/out": 6.627696990966797, "created_at": "2025-01-14T13:10:35.477549+00:00"} {"global_step": 141, "acc_step": 0, "speed/wps": 12866.287690727791, "speed/FLOPS": 202082728314912.44, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3030517101287842, "optim/lr": 8.400000000000001e-05, "optim/total_tokens": 36962304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497589, "loss/out": 6.745237350463867, "created_at": "2025-01-14T13:10:40.578838+00:00"} {"global_step": 142, "acc_step": 0, "speed/wps": 12864.764682207619, "speed/FLOPS": 202058807373270.12, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4015449285507202, "optim/lr": 8.46e-05, "optim/total_tokens": 37224448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475904, "loss/out": 6.687929153442383, "created_at": "2025-01-14T13:10:45.681072+00:00"} {"global_step": 143, "acc_step": 0, "speed/wps": 12862.520119304805, "speed/FLOPS": 202023553428528.78, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0549798011779785, "optim/lr": 8.520000000000001e-05, "optim/total_tokens": 37486592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492777, "loss/out": 6.490950107574463, "created_at": "2025-01-14T13:10:50.779978+00:00"} {"global_step": 144, "acc_step": 0, "speed/wps": 12862.740776802255, "speed/FLOPS": 202027019157741.4, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4877595901489258, "optim/lr": 8.58e-05, "optim/total_tokens": 37748736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479338, "loss/out": 6.604954242706299, "created_at": "2025-01-14T13:10:55.885825+00:00"} {"global_step": 145, "acc_step": 0, "speed/wps": 12864.340975293053, "speed/FLOPS": 202052152473942.56, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3386386632919312, "optim/lr": 8.64e-05, "optim/total_tokens": 38010880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482471, "loss/out": 6.730251789093018, "created_at": "2025-01-14T13:11:00.987791+00:00"} {"global_step": 146, "acc_step": 0, "speed/wps": 12867.90479633271, "speed/FLOPS": 202108127180573.66, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2293115854263306, "optim/lr": 8.7e-05, "optim/total_tokens": 38273024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479574, "loss/out": 6.643328666687012, "created_at": "2025-01-14T13:11:06.086209+00:00"} {"global_step": 147, "acc_step": 0, "speed/wps": 12865.531578520468, "speed/FLOPS": 202070852533689.38, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8890581727027893, "optim/lr": 8.76e-05, "optim/total_tokens": 38535168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478008, "loss/out": 6.627523899078369, "created_at": "2025-01-14T13:11:11.185094+00:00"} {"global_step": 148, "acc_step": 0, "speed/wps": 12865.13393833598, "speed/FLOPS": 202064607048177.12, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0163570642471313, "optim/lr": 8.82e-05, "optim/total_tokens": 38797312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498564, "loss/out": 6.619181156158447, "created_at": "2025-01-14T13:11:16.283114+00:00"} {"global_step": 149, "acc_step": 0, "speed/wps": 12861.183482445904, "speed/FLOPS": 202002559709928.22, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.962559163570404, "optim/lr": 8.88e-05, "optim/total_tokens": 39059456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478487, "loss/out": 6.4872283935546875, "created_at": "2025-01-14T13:11:21.392621+00:00"} {"global_step": 150, "acc_step": 0, "speed/wps": 12864.561438917322, "speed/FLOPS": 202055615158110.47, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8599451780319214, "optim/lr": 8.94e-05, "optim/total_tokens": 39321600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502792, "loss/out": 6.5910234451293945, "created_at": "2025-01-14T13:11:26.490624+00:00"} {"global_step": 151, "acc_step": 0, "speed/wps": 12862.737386391007, "speed/FLOPS": 202026965906673.84, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.878116250038147, "optim/lr": 8.999999999999999e-05, "optim/total_tokens": 39583744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490411, "loss/out": 6.446880340576172, "created_at": "2025-01-14T13:11:31.590828+00:00"} {"global_step": 152, "acc_step": 0, "speed/wps": 12862.236067221062, "speed/FLOPS": 202019092000225.12, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9335630536079407, "optim/lr": 9.06e-05, "optim/total_tokens": 39845888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355439, "loss/out": 6.626241683959961, "created_at": "2025-01-14T13:11:36.692905+00:00"} {"global_step": 153, "acc_step": 0, "speed/wps": 12868.927846683322, "speed/FLOPS": 202124195592148.53, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6971123218536377, "optim/lr": 9.120000000000001e-05, "optim/total_tokens": 40108032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471672, "loss/out": 6.607149124145508, "created_at": "2025-01-14T13:11:41.792003+00:00"} {"global_step": 154, "acc_step": 0, "speed/wps": 12862.205111641311, "speed/FLOPS": 202018605800307.72, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9943373203277588, "optim/lr": 9.18e-05, "optim/total_tokens": 40370176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357066, "loss/out": 6.552493095397949, "created_at": "2025-01-14T13:11:46.896324+00:00"} {"global_step": 155, "acc_step": 0, "speed/wps": 12860.578870790576, "speed/FLOPS": 201993063453057.8, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9649426937103271, "optim/lr": 9.240000000000001e-05, "optim/total_tokens": 40632320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500182, "loss/out": 6.54964542388916, "created_at": "2025-01-14T13:11:51.996675+00:00"} {"global_step": 156, "acc_step": 0, "speed/wps": 12858.402801938402, "speed/FLOPS": 201958885301502.47, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6764341592788696, "optim/lr": 9.3e-05, "optim/total_tokens": 40894464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472720, "loss/out": 6.469317436218262, "created_at": "2025-01-14T13:11:57.099950+00:00"} {"global_step": 157, "acc_step": 0, "speed/wps": 12858.797147689253, "speed/FLOPS": 201965079043407.44, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9821031093597412, "optim/lr": 9.36e-05, "optim/total_tokens": 41156608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501747, "loss/out": 6.664370059967041, "created_at": "2025-01-14T13:12:02.200789+00:00"} {"global_step": 158, "acc_step": 0, "speed/wps": 12854.490960968456, "speed/FLOPS": 201897444463636.7, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.8579787015914917, "optim/lr": 9.42e-05, "optim/total_tokens": 41418752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 6.588953971862793, "created_at": "2025-01-14T13:12:07.307353+00:00"} {"global_step": 159, "acc_step": 0, "speed/wps": 12851.425155391753, "speed/FLOPS": 201849291774196.22, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8167021870613098, "optim/lr": 9.480000000000001e-05, "optim/total_tokens": 41680896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481380, "loss/out": 6.529260635375977, "created_at": "2025-01-14T13:12:12.413204+00:00"} {"global_step": 160, "acc_step": 0, "speed/wps": 12865.053754679397, "speed/FLOPS": 202063347653666.6, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8619881272315979, "optim/lr": 9.54e-05, "optim/total_tokens": 41943040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503601, "loss/out": 6.551255702972412, "created_at": "2025-01-14T13:12:17.510921+00:00"} {"global_step": 161, "acc_step": 0, "speed/wps": 12860.57851588063, "speed/FLOPS": 201993057878709.44, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6816393733024597, "optim/lr": 9.6e-05, "optim/total_tokens": 42205184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323562, "loss/out": 6.546786308288574, "created_at": "2025-01-14T13:12:22.611871+00:00"} {"global_step": 162, "acc_step": 0, "speed/wps": 12859.998887116735, "speed/FLOPS": 201983954012479.06, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.6955951452255249, "optim/lr": 9.66e-05, "optim/total_tokens": 42467328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490113, "loss/out": 6.4652605056762695, "created_at": "2025-01-14T13:12:27.712055+00:00"} {"global_step": 163, "acc_step": 0, "speed/wps": 12861.051458987498, "speed/FLOPS": 202000486100094.72, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8276886940002441, "optim/lr": 9.719999999999999e-05, "optim/total_tokens": 42729472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330760, "loss/out": 6.460460186004639, "created_at": "2025-01-14T13:12:32.813195+00:00"} {"global_step": 164, "acc_step": 0, "speed/wps": 12855.184432451668, "speed/FLOPS": 201908336386210.2, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9237439036369324, "optim/lr": 9.779999999999999e-05, "optim/total_tokens": 42991616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503885, "loss/out": 6.526254653930664, "created_at": "2025-01-14T13:12:37.915135+00:00"} {"global_step": 165, "acc_step": 0, "speed/wps": 12858.55057243859, "speed/FLOPS": 201961206240265.84, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.059998631477356, "optim/lr": 9.84e-05, "optim/total_tokens": 43253760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500938, "loss/out": 6.518193244934082, "created_at": "2025-01-14T13:12:43.022176+00:00"} {"global_step": 166, "acc_step": 0, "speed/wps": 12857.111799577124, "speed/FLOPS": 201938608335395.44, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6621858477592468, "optim/lr": 9.900000000000001e-05, "optim/total_tokens": 43515904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501295, "loss/out": 6.531671047210693, "created_at": "2025-01-14T13:12:48.123028+00:00"} {"global_step": 167, "acc_step": 0, "speed/wps": 12860.765751225379, "speed/FLOPS": 201995998667087.03, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9150899052619934, "optim/lr": 9.960000000000001e-05, "optim/total_tokens": 43778048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483281, "loss/out": 6.426274299621582, "created_at": "2025-01-14T13:12:53.223562+00:00"} {"global_step": 168, "acc_step": 0, "speed/wps": 12851.346154094275, "speed/FLOPS": 201848050950260.16, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2042263746261597, "optim/lr": 0.0001002, "optim/total_tokens": 44040192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492777, "loss/out": 6.47704553604126, "created_at": "2025-01-14T13:12:58.331609+00:00"} {"global_step": 169, "acc_step": 0, "speed/wps": 12868.665814215938, "speed/FLOPS": 202120080012177.9, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6311513185501099, "optim/lr": 0.0001008, "optim/total_tokens": 44302336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466913, "loss/out": 6.429878234863281, "created_at": "2025-01-14T13:13:03.430877+00:00"} {"global_step": 170, "acc_step": 0, "speed/wps": 12851.752946046507, "speed/FLOPS": 201854440176860.7, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0015, "optim/grad_norm": 1.3966560363769531, "optim/lr": 0.0001014, "optim/total_tokens": 44564480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492777, "loss/out": 6.553410530090332, "created_at": "2025-01-14T13:13:08.540079+00:00"} {"global_step": 171, "acc_step": 0, "speed/wps": 12857.133709949418, "speed/FLOPS": 201938952467900.34, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9051603078842163, "optim/lr": 0.00010200000000000001, "optim/total_tokens": 44826624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304917, "loss/out": 6.500462055206299, "created_at": "2025-01-14T13:13:13.641651+00:00"} {"global_step": 172, "acc_step": 0, "speed/wps": 12857.391885393943, "speed/FLOPS": 201943007468026.84, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9989539980888367, "optim/lr": 0.0001026, "optim/total_tokens": 45088768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482761, "loss/out": 6.4284820556640625, "created_at": "2025-01-14T13:13:18.745768+00:00"} {"global_step": 173, "acc_step": 0, "speed/wps": 12860.909476737013, "speed/FLOPS": 201998256073746.78, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.9464795589447021, "optim/lr": 0.0001032, "optim/total_tokens": 45350912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359040, "loss/out": 6.444293022155762, "created_at": "2025-01-14T13:13:23.848014+00:00"} {"global_step": 174, "acc_step": 0, "speed/wps": 12866.808532993524, "speed/FLOPS": 202090908858408.06, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.999801754951477, "optim/lr": 0.0001038, "optim/total_tokens": 45613056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499155, "loss/out": 6.497164726257324, "created_at": "2025-01-14T13:13:28.947603+00:00"} {"global_step": 175, "acc_step": 0, "speed/wps": 12863.399951844216, "speed/FLOPS": 202037372407576.84, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8014741539955139, "optim/lr": 0.00010439999999999999, "optim/total_tokens": 45875200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369430, "loss/out": 6.428458213806152, "created_at": "2025-01-14T13:13:34.047187+00:00"} {"global_step": 176, "acc_step": 0, "speed/wps": 12864.041032534182, "speed/FLOPS": 202047441460749.94, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9850582480430603, "optim/lr": 0.00010500000000000002, "optim/total_tokens": 46137344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497305, "loss/out": 6.403379440307617, "created_at": "2025-01-14T13:13:39.145687+00:00"} {"global_step": 177, "acc_step": 0, "speed/wps": 12847.66560870542, "speed/FLOPS": 201790242927328.8, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7382384538650513, "optim/lr": 0.0001056, "optim/total_tokens": 46399488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336239, "loss/out": 6.39181661605835, "created_at": "2025-01-14T13:13:44.253014+00:00"} {"global_step": 178, "acc_step": 0, "speed/wps": 12865.070171890651, "speed/FLOPS": 202063605508528.97, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7417484521865845, "optim/lr": 0.0001062, "optim/total_tokens": 46661632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502556, "loss/out": 6.451013088226318, "created_at": "2025-01-14T13:13:49.355207+00:00"} {"global_step": 179, "acc_step": 0, "speed/wps": 12853.206376094447, "speed/FLOPS": 201877268292985.78, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9209765195846558, "optim/lr": 0.00010680000000000001, "optim/total_tokens": 46923776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467150, "loss/out": 6.512592315673828, "created_at": "2025-01-14T13:13:54.460139+00:00"} {"global_step": 180, "acc_step": 0, "speed/wps": 12869.369959762034, "speed/FLOPS": 202131139585576.75, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9147020578384399, "optim/lr": 0.0001074, "optim/total_tokens": 47185920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491159, "loss/out": 6.385218143463135, "created_at": "2025-01-14T13:13:59.559180+00:00"} {"global_step": 181, "acc_step": 0, "speed/wps": 12860.914379919215, "speed/FLOPS": 201998333084960.97, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7410407662391663, "optim/lr": 0.000108, "optim/total_tokens": 47448064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324619, "loss/out": 6.42431640625, "created_at": "2025-01-14T13:14:04.659626+00:00"} {"global_step": 182, "acc_step": 0, "speed/wps": 12858.529035530259, "speed/FLOPS": 201960867973525.72, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.639997661113739, "optim/lr": 0.00010860000000000001, "optim/total_tokens": 47710208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491442, "loss/out": 6.440360069274902, "created_at": "2025-01-14T13:14:09.760285+00:00"} {"global_step": 183, "acc_step": 0, "speed/wps": 12853.375337807394, "speed/FLOPS": 201879922068865.3, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6367738842964172, "optim/lr": 0.00010920000000000001, "optim/total_tokens": 47972352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316156, "loss/out": 6.398746967315674, "created_at": "2025-01-14T13:14:14.862878+00:00"} {"global_step": 184, "acc_step": 0, "speed/wps": 12852.700874472699, "speed/FLOPS": 201869328695384.2, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8695307374000549, "optim/lr": 0.0001098, "optim/total_tokens": 48234496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484091, "loss/out": 6.359612941741943, "created_at": "2025-01-14T13:14:19.970233+00:00"} {"global_step": 185, "acc_step": 0, "speed/wps": 12860.169042766567, "speed/FLOPS": 201986626540778.03, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0248782634735107, "optim/lr": 0.0001104, "optim/total_tokens": 48496640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352913, "loss/out": 6.289403915405273, "created_at": "2025-01-14T13:14:25.073465+00:00"} {"global_step": 186, "acc_step": 0, "speed/wps": 12863.188103398197, "speed/FLOPS": 202034045036621.84, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8797543048858643, "optim/lr": 0.000111, "optim/total_tokens": 48758784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493598, "loss/out": 6.442431449890137, "created_at": "2025-01-14T13:14:30.175934+00:00"} {"global_step": 187, "acc_step": 0, "speed/wps": 12856.263126698192, "speed/FLOPS": 201925278761633.06, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7346510291099548, "optim/lr": 0.00011159999999999999, "optim/total_tokens": 49020928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490175, "loss/out": 6.355038642883301, "created_at": "2025-01-14T13:14:35.279624+00:00"} {"global_step": 188, "acc_step": 0, "speed/wps": 12853.430724164442, "speed/FLOPS": 201880791987711.6, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.001, "optim/grad_norm": 0.9529773592948914, "optim/lr": 0.00011220000000000002, "optim/total_tokens": 49283072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491685, "loss/out": 6.3907389640808105, "created_at": "2025-01-14T13:14:40.384715+00:00"} {"global_step": 189, "acc_step": 0, "speed/wps": 12862.518250175273, "speed/FLOPS": 202023524071281.03, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8813312649726868, "optim/lr": 0.0001128, "optim/total_tokens": 49545216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499155, "loss/out": 6.3236589431762695, "created_at": "2025-01-14T13:14:45.485966+00:00"} {"global_step": 190, "acc_step": 0, "speed/wps": 12858.6500599815, "speed/FLOPS": 201962768828837.5, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9257749915122986, "optim/lr": 0.0001134, "optim/total_tokens": 49807360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491442, "loss/out": 6.325042724609375, "created_at": "2025-01-14T13:14:50.586167+00:00"} {"global_step": 191, "acc_step": 0, "speed/wps": 12858.823742167702, "speed/FLOPS": 201965496746236.12, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2187652587890625, "optim/lr": 0.000114, "optim/total_tokens": 50069504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369811, "loss/out": 6.344945430755615, "created_at": "2025-01-14T13:14:55.688830+00:00"} {"global_step": 192, "acc_step": 0, "speed/wps": 12858.736407046139, "speed/FLOPS": 201964125028140.94, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9919604659080505, "optim/lr": 0.0001146, "optim/total_tokens": 50331648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490985, "loss/out": 6.4616241455078125, "created_at": "2025-01-14T13:15:00.792097+00:00"} {"global_step": 193, "acc_step": 0, "speed/wps": 12856.321316623984, "speed/FLOPS": 201926192714378.97, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7947763204574585, "optim/lr": 0.0001152, "optim/total_tokens": 50593792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322458, "loss/out": 6.355794429779053, "created_at": "2025-01-14T13:15:05.895953+00:00"} {"global_step": 194, "acc_step": 0, "speed/wps": 12852.606096464864, "speed/FLOPS": 201867840076532.5, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8891825675964355, "optim/lr": 0.00011580000000000001, "optim/total_tokens": 50855936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500702, "loss/out": 6.274432182312012, "created_at": "2025-01-14T13:15:11.005069+00:00"} {"global_step": 195, "acc_step": 0, "speed/wps": 12858.989667598224, "speed/FLOPS": 201968102833127.9, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7590185403823853, "optim/lr": 0.00011640000000000001, "optim/total_tokens": 51118080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475631, "loss/out": 6.4034528732299805, "created_at": "2025-01-14T13:15:16.107583+00:00"} {"global_step": 196, "acc_step": 0, "speed/wps": 12864.842079896769, "speed/FLOPS": 202060023010333.78, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7445764541625977, "optim/lr": 0.000117, "optim/total_tokens": 51380224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498328, "loss/out": 6.34002161026001, "created_at": "2025-01-14T13:15:21.205937+00:00"} {"global_step": 197, "acc_step": 0, "speed/wps": 12853.232564775142, "speed/FLOPS": 201877679622203.1, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8201125264167786, "optim/lr": 0.0001176, "optim/total_tokens": 51642368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472681, "loss/out": 6.338621139526367, "created_at": "2025-01-14T13:15:26.308461+00:00"} {"global_step": 198, "acc_step": 0, "speed/wps": 12855.188070164095, "speed/FLOPS": 201908393521482.6, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5797470211982727, "optim/lr": 0.0001182, "optim/total_tokens": 51904512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485138, "loss/out": 6.359138488769531, "created_at": "2025-01-14T13:15:31.413037+00:00"} {"global_step": 199, "acc_step": 0, "speed/wps": 12855.000313817585, "speed/FLOPS": 201905444550056.4, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9750800728797913, "optim/lr": 0.00011880000000000001, "optim/total_tokens": 52166656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495954, "loss/out": 6.266099452972412, "created_at": "2025-01-14T13:15:36.514928+00:00"} {"global_step": 200, "acc_step": 0, "speed/wps": 12859.468058627912, "speed/FLOPS": 201975616621627.3, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9907289147377014, "optim/lr": 0.0001194, "optim/total_tokens": 52428800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 6.3834991455078125, "created_at": "2025-01-14T13:15:41.615903+00:00"} {"global_step": 201, "acc_step": 0, "speed/wps": 12855.13597756699, "speed/FLOPS": 201907575335662.34, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5869675278663635, "optim/lr": 0.00012, "optim/total_tokens": 52690944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488549, "loss/out": 6.388124465942383, "created_at": "2025-01-14T13:15:46.719421+00:00"} {"global_step": 202, "acc_step": 0, "speed/wps": 12854.38921311867, "speed/FLOPS": 201895846373839.56, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0005, "optim/grad_norm": 1.3994063138961792, "optim/lr": 0.0001206, "optim/total_tokens": 52953088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321933, "loss/out": 6.321395397186279, "created_at": "2025-01-14T13:15:51.821575+00:00"} {"global_step": 203, "acc_step": 0, "speed/wps": 12853.143488480993, "speed/FLOPS": 201876280556599.84, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7970197796821594, "optim/lr": 0.00012119999999999999, "optim/total_tokens": 53215232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497808, "loss/out": 6.287402629852295, "created_at": "2025-01-14T13:15:56.927264+00:00"} {"global_step": 204, "acc_step": 0, "speed/wps": 12855.609631335203, "speed/FLOPS": 201915014718960.6, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6499330401420593, "optim/lr": 0.00012179999999999999, "optim/total_tokens": 53477376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478528, "loss/out": 6.255059242248535, "created_at": "2025-01-14T13:16:02.031716+00:00"} {"global_step": 205, "acc_step": 0, "speed/wps": 12856.347139834797, "speed/FLOPS": 201926598303389.62, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7855697870254517, "optim/lr": 0.00012240000000000002, "optim/total_tokens": 53739520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487798, "loss/out": 6.321211338043213, "created_at": "2025-01-14T13:16:07.133237+00:00"} {"global_step": 206, "acc_step": 0, "speed/wps": 12856.303681549267, "speed/FLOPS": 201925915731297.4, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6286308169364929, "optim/lr": 0.000123, "optim/total_tokens": 54001664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495954, "loss/out": 6.303275108337402, "created_at": "2025-01-14T13:16:12.237513+00:00"} {"global_step": 207, "acc_step": 0, "speed/wps": 12846.449062243608, "speed/FLOPS": 201771135393436.06, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6162287592887878, "optim/lr": 0.0001236, "optim/total_tokens": 54263808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494342, "loss/out": 6.256097316741943, "created_at": "2025-01-14T13:16:17.342863+00:00"} {"global_step": 208, "acc_step": 0, "speed/wps": 12863.362867574999, "speed/FLOPS": 202036789948169.66, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6104185581207275, "optim/lr": 0.0001242, "optim/total_tokens": 54525952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480385, "loss/out": 6.274533271789551, "created_at": "2025-01-14T13:16:22.441292+00:00"} {"global_step": 209, "acc_step": 0, "speed/wps": 12857.771967175528, "speed/FLOPS": 201948977174693.72, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5323169827461243, "optim/lr": 0.0001248, "optim/total_tokens": 54788096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497235, "loss/out": 6.2833404541015625, "created_at": "2025-01-14T13:16:27.546587+00:00"} {"global_step": 210, "acc_step": 0, "speed/wps": 12845.508586262395, "speed/FLOPS": 201756363925801.66, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6386939287185669, "optim/lr": 0.00012539999999999999, "optim/total_tokens": 55050240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300932, "loss/out": 6.314824104309082, "created_at": "2025-01-14T13:16:32.660367+00:00"} {"global_step": 211, "acc_step": 0, "speed/wps": 12859.281843334013, "speed/FLOPS": 201972691854550.0, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8191348910331726, "optim/lr": 0.000126, "optim/total_tokens": 55312384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478244, "loss/out": 6.333139896392822, "created_at": "2025-01-14T13:16:37.760592+00:00"} {"global_step": 212, "acc_step": 0, "speed/wps": 12848.288205636863, "speed/FLOPS": 201800021667675.0, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7866783142089844, "optim/lr": 0.00012660000000000001, "optim/total_tokens": 55574528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 381520, "loss/out": 6.241124153137207, "created_at": "2025-01-14T13:16:42.864901+00:00"} {"global_step": 213, "acc_step": 0, "speed/wps": 12857.687601809803, "speed/FLOPS": 201947652100694.84, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.48436424136161804, "optim/lr": 0.0001272, "optim/total_tokens": 55836672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491685, "loss/out": 6.227824687957764, "created_at": "2025-01-14T13:16:47.971563+00:00"} {"global_step": 214, "acc_step": 0, "speed/wps": 12840.21749225234, "speed/FLOPS": 201673260023648.7, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8873641490936279, "optim/lr": 0.0001278, "optim/total_tokens": 56098816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494106, "loss/out": 6.259526252746582, "created_at": "2025-01-14T13:16:53.087427+00:00"} {"global_step": 215, "acc_step": 0, "speed/wps": 12855.960725607334, "speed/FLOPS": 201920529136958.5, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8515705466270447, "optim/lr": 0.0001284, "optim/total_tokens": 56360960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472161, "loss/out": 6.1248250007629395, "created_at": "2025-01-14T13:16:58.191058+00:00"} {"global_step": 216, "acc_step": 0, "speed/wps": 12855.711555863645, "speed/FLOPS": 201916615583738.78, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8277077674865723, "optim/lr": 0.000129, "optim/total_tokens": 56623104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498854, "loss/out": 6.189246654510498, "created_at": "2025-01-14T13:17:03.294697+00:00"} {"global_step": 217, "acc_step": 0, "speed/wps": 12862.005468540883, "speed/FLOPS": 202015470131077.12, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9166450500488281, "optim/lr": 0.0001296, "optim/total_tokens": 56885248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476151, "loss/out": 6.21877384185791, "created_at": "2025-01-14T13:17:08.399520+00:00"} {"global_step": 218, "acc_step": 0, "speed/wps": 12841.368593506944, "speed/FLOPS": 201691339650631.12, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6895045042037964, "optim/lr": 0.0001302, "optim/total_tokens": 57147392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476387, "loss/out": 6.17383337020874, "created_at": "2025-01-14T13:17:13.511211+00:00"} {"global_step": 219, "acc_step": 0, "speed/wps": 12857.629122295652, "speed/FLOPS": 201946733599565.84, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5883886814117432, "optim/lr": 0.0001308, "optim/total_tokens": 57409536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479574, "loss/out": 6.244125843048096, "created_at": "2025-01-14T13:17:18.617301+00:00"} {"global_step": 220, "acc_step": 0, "speed/wps": 12851.95498675647, "speed/FLOPS": 201857613503843.66, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8458524346351624, "optim/lr": 0.0001314, "optim/total_tokens": 57671680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497235, "loss/out": 6.245402812957764, "created_at": "2025-01-14T13:17:23.724690+00:00"} {"global_step": 221, "acc_step": 0, "speed/wps": 12858.237375658115, "speed/FLOPS": 201956287054451.03, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7617703080177307, "optim/lr": 0.00013199999999999998, "optim/total_tokens": 57933824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322458, "loss/out": 6.1732096672058105, "created_at": "2025-01-14T13:17:28.825024+00:00"} {"global_step": 222, "acc_step": 0, "speed/wps": 12861.140480382986, "speed/FLOPS": 202001884303439.06, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6762282252311707, "optim/lr": 0.00013260000000000002, "optim/total_tokens": 58195968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491159, "loss/out": 6.264778137207031, "created_at": "2025-01-14T13:17:33.924628+00:00"} {"global_step": 223, "acc_step": 0, "speed/wps": 12854.827371288247, "speed/FLOPS": 201902728249985.22, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9015085101127625, "optim/lr": 0.0001332, "optim/total_tokens": 58458112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310301, "loss/out": 6.218998908996582, "created_at": "2025-01-14T13:17:39.028023+00:00"} {"global_step": 224, "acc_step": 0, "speed/wps": 12858.79062949759, "speed/FLOPS": 201964976666250.4, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9561952352523804, "optim/lr": 0.0001338, "optim/total_tokens": 58720256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481951, "loss/out": 6.280904769897461, "created_at": "2025-01-14T13:17:44.131667+00:00"} {"global_step": 225, "acc_step": 0, "speed/wps": 12852.806011290531, "speed/FLOPS": 201870980013581.38, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7103268504142761, "optim/lr": 0.00013440000000000001, "optim/total_tokens": 58982400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323655, "loss/out": 6.194652557373047, "created_at": "2025-01-14T13:17:49.257248+00:00"} {"global_step": 226, "acc_step": 0, "speed/wps": 12852.973248716315, "speed/FLOPS": 201873606707161.6, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3398207426071167, "optim/lr": 0.000135, "optim/total_tokens": 59244544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483518, "loss/out": 6.189067840576172, "created_at": "2025-01-14T13:17:54.359722+00:00"} {"global_step": 227, "acc_step": 0, "speed/wps": 12857.720713498045, "speed/FLOPS": 201948172165258.94, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6333998441696167, "optim/lr": 0.0001356, "optim/total_tokens": 59506688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494402, "loss/out": 6.433516502380371, "created_at": "2025-01-14T13:17:59.460251+00:00"} {"global_step": 228, "acc_step": 0, "speed/wps": 12857.271065512567, "speed/FLOPS": 201941109825765.9, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7332465648651123, "optim/lr": 0.0001362, "optim/total_tokens": 59768832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487272, "loss/out": 6.181964874267578, "created_at": "2025-01-14T13:18:04.569183+00:00"} {"global_step": 229, "acc_step": 0, "speed/wps": 12854.284536453995, "speed/FLOPS": 201894202282980.06, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9002785682678223, "optim/lr": 0.00013680000000000002, "optim/total_tokens": 60030976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500182, "loss/out": 6.346478462219238, "created_at": "2025-01-14T13:18:09.672294+00:00"} {"global_step": 230, "acc_step": 0, "speed/wps": 12848.650665487736, "speed/FLOPS": 201805714597704.97, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.602932333946228, "optim/lr": 0.0001374, "optim/total_tokens": 60293120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287620, "loss/out": 6.175745487213135, "created_at": "2025-01-14T13:18:14.780175+00:00"} {"global_step": 231, "acc_step": 0, "speed/wps": 12851.30757853142, "speed/FLOPS": 201847445068037.12, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6409474611282349, "optim/lr": 0.000138, "optim/total_tokens": 60555264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496999, "loss/out": 6.170928478240967, "created_at": "2025-01-14T13:18:19.888069+00:00"} {"global_step": 232, "acc_step": 0, "speed/wps": 12856.096134824003, "speed/FLOPS": 201922655924778.06, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5931607484817505, "optim/lr": 0.0001386, "optim/total_tokens": 60817408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479338, "loss/out": 6.206583023071289, "created_at": "2025-01-14T13:18:24.995078+00:00"} {"global_step": 233, "acc_step": 0, "speed/wps": 12856.178381837422, "speed/FLOPS": 201923947727143.06, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5251399278640747, "optim/lr": 0.0001392, "optim/total_tokens": 61079552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500938, "loss/out": 6.157829284667969, "created_at": "2025-01-14T13:18:30.098974+00:00"} {"global_step": 234, "acc_step": 0, "speed/wps": 12860.750153556159, "speed/FLOPS": 201995753684258.47, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5146976709365845, "optim/lr": 0.0001398, "optim/total_tokens": 61341696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481188, "loss/out": 6.1707987785339355, "created_at": "2025-01-14T13:18:35.198999+00:00"} {"global_step": 235, "acc_step": 0, "speed/wps": 12854.638136702364, "speed/FLOPS": 201899756060778.34, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5900346040725708, "optim/lr": 0.0001404, "optim/total_tokens": 61603840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495954, "loss/out": 6.124685287475586, "created_at": "2025-01-14T13:18:40.302434+00:00"} {"global_step": 236, "acc_step": 0, "speed/wps": 12857.581840075634, "speed/FLOPS": 201945990967327.88, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5561619400978088, "optim/lr": 0.000141, "optim/total_tokens": 61865984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313386, "loss/out": 6.226252555847168, "created_at": "2025-01-14T13:18:45.409423+00:00"} {"global_step": 237, "acc_step": 0, "speed/wps": 12856.213578534884, "speed/FLOPS": 201924500539642.0, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5342633724212646, "optim/lr": 0.0001416, "optim/total_tokens": 62128128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481951, "loss/out": 6.146413803100586, "created_at": "2025-01-14T13:18:50.511540+00:00"} {"global_step": 238, "acc_step": 0, "speed/wps": 12858.071016191487, "speed/FLOPS": 201953674150425.78, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5920140147209167, "optim/lr": 0.0001422, "optim/total_tokens": 62390272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368907, "loss/out": 6.09031867980957, "created_at": "2025-01-14T13:18:55.613188+00:00"} {"global_step": 239, "acc_step": 0, "speed/wps": 12857.403011367702, "speed/FLOPS": 201943182216733.4, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6920345425605774, "optim/lr": 0.0001428, "optim/total_tokens": 62652416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480385, "loss/out": 6.175528526306152, "created_at": "2025-01-14T13:19:00.716624+00:00"} {"global_step": 240, "acc_step": 0, "speed/wps": 12860.768689307752, "speed/FLOPS": 201996044813708.03, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5641527771949768, "optim/lr": 0.00014340000000000002, "optim/total_tokens": 62914560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479331, "loss/out": 6.180868625640869, "created_at": "2025-01-14T13:19:05.822345+00:00"} {"global_step": 241, "acc_step": 0, "speed/wps": 12853.53622509672, "speed/FLOPS": 201882449024826.47, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4754968285560608, "optim/lr": 0.000144, "optim/total_tokens": 63176704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485366, "loss/out": 6.190311908721924, "created_at": "2025-01-14T13:19:10.924716+00:00"} {"global_step": 242, "acc_step": 0, "speed/wps": 12863.272919534853, "speed/FLOPS": 202035377190597.6, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5433962941169739, "optim/lr": 0.0001446, "optim/total_tokens": 63438848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478528, "loss/out": 6.2208251953125, "created_at": "2025-01-14T13:19:16.023130+00:00"} {"global_step": 243, "acc_step": 0, "speed/wps": 12859.848387727056, "speed/FLOPS": 201981590212756.3, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.45960694551467896, "optim/lr": 0.0001452, "optim/total_tokens": 63700992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501228, "loss/out": 6.16837215423584, "created_at": "2025-01-14T13:19:21.125303+00:00"} {"global_step": 244, "acc_step": 0, "speed/wps": 12859.407462044339, "speed/FLOPS": 201974664869013.88, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6318148970603943, "optim/lr": 0.0001458, "optim/total_tokens": 63963136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294835, "loss/out": 6.192949295043945, "created_at": "2025-01-14T13:19:26.228318+00:00"} {"global_step": 245, "acc_step": 0, "speed/wps": 12855.62552889998, "speed/FLOPS": 201915264412060.97, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7576618790626526, "optim/lr": 0.0001464, "optim/total_tokens": 64225280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490923, "loss/out": 6.101706504821777, "created_at": "2025-01-14T13:19:31.329671+00:00"} {"global_step": 246, "acc_step": 0, "speed/wps": 12856.803579975785, "speed/FLOPS": 201933767323018.3, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0005, "optim/grad_norm": 1.1288641691207886, "optim/lr": 0.000147, "optim/total_tokens": 64487424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325144, "loss/out": 6.122736930847168, "created_at": "2025-01-14T13:19:36.433439+00:00"} {"global_step": 247, "acc_step": 0, "speed/wps": 12854.631063849245, "speed/FLOPS": 201899644971901.0, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.132357120513916, "optim/lr": 0.0001476, "optim/total_tokens": 64749568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484320, "loss/out": 6.134947299957275, "created_at": "2025-01-14T13:19:41.540539+00:00"} {"global_step": 248, "acc_step": 0, "speed/wps": 12856.05512797474, "speed/FLOPS": 201922011855861.06, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.149116039276123, "optim/lr": 0.0001482, "optim/total_tokens": 65011712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350115, "loss/out": 6.091949462890625, "created_at": "2025-01-14T13:19:46.644947+00:00"} {"global_step": 249, "acc_step": 0, "speed/wps": 12854.625303659544, "speed/FLOPS": 201899554500206.4, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9775282144546509, "optim/lr": 0.00014879999999999998, "optim/total_tokens": 65273856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499682, "loss/out": 6.260706901550293, "created_at": "2025-01-14T13:19:51.754669+00:00"} {"global_step": 250, "acc_step": 0, "speed/wps": 12856.895734934493, "speed/FLOPS": 201935214743279.0, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3296104669570923, "optim/lr": 0.0001494, "optim/total_tokens": 65536000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481715, "loss/out": 6.198958396911621, "created_at": "2025-01-14T13:19:56.856757+00:00"} {"global_step": 251, "acc_step": 0, "speed/wps": 12853.355083672752, "speed/FLOPS": 201879603949848.66, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9611312747001648, "optim/lr": 0.00015000000000000001, "optim/total_tokens": 65798144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486411, "loss/out": 6.178905487060547, "created_at": "2025-01-14T13:20:01.964266+00:00"} {"global_step": 252, "acc_step": 0, "speed/wps": 12857.298231581617, "speed/FLOPS": 201941536506210.44, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7981414198875427, "optim/lr": 0.0001506, "optim/total_tokens": 66060288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487798, "loss/out": 6.204502105712891, "created_at": "2025-01-14T13:20:07.070819+00:00"} {"global_step": 253, "acc_step": 0, "speed/wps": 12853.25782621237, "speed/FLOPS": 201878076387787.62, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.698196530342102, "optim/lr": 0.00015120000000000002, "optim/total_tokens": 66322432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486988, "loss/out": 6.054280757904053, "created_at": "2025-01-14T13:20:12.177101+00:00"} {"global_step": 254, "acc_step": 0, "speed/wps": 12846.750739272591, "speed/FLOPS": 201775873645723.9, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6168485879898071, "optim/lr": 0.0001518, "optim/total_tokens": 66584576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 6.130191802978516, "created_at": "2025-01-14T13:20:17.284551+00:00"} {"global_step": 255, "acc_step": 0, "speed/wps": 12855.888620635236, "speed/FLOPS": 201919396629289.1, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6779220104217529, "optim/lr": 0.0001524, "optim/total_tokens": 66846720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331148, "loss/out": 6.1100568771362305, "created_at": "2025-01-14T13:20:22.390124+00:00"} {"global_step": 256, "acc_step": 0, "speed/wps": 12851.658158547101, "speed/FLOPS": 201852951408930.84, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7192084193229675, "optim/lr": 0.000153, "optim/total_tokens": 67108864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482471, "loss/out": 6.107982635498047, "created_at": "2025-01-14T13:20:27.493889+00:00"} {"global_step": 257, "acc_step": 0, "speed/wps": 12852.306426024068, "speed/FLOPS": 201863133340469.47, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8695621490478516, "optim/lr": 0.00015360000000000002, "optim/total_tokens": 67371008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318185, "loss/out": 6.072858810424805, "created_at": "2025-01-14T13:20:32.596986+00:00"} {"global_step": 258, "acc_step": 0, "speed/wps": 12855.307242198074, "speed/FLOPS": 201910265282035.78, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9716563820838928, "optim/lr": 0.0001542, "optim/total_tokens": 67633152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497525, "loss/out": 6.150859355926514, "created_at": "2025-01-14T13:20:37.699837+00:00"} {"global_step": 259, "acc_step": 0, "speed/wps": 12854.995469155987, "speed/FLOPS": 201905368457988.75, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7617890238761902, "optim/lr": 0.0001548, "optim/total_tokens": 67895296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490695, "loss/out": 5.993987083435059, "created_at": "2025-01-14T13:20:42.804234+00:00"} {"global_step": 260, "acc_step": 0, "speed/wps": 12850.130255133254, "speed/FLOPS": 201828953586257.06, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.001, "optim/grad_norm": 0.5909282565116882, "optim/lr": 0.0001554, "optim/total_tokens": 68157440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496190, "loss/out": 6.203211307525635, "created_at": "2025-01-14T13:20:47.907831+00:00"} {"global_step": 261, "acc_step": 0, "speed/wps": 12853.947072129233, "speed/FLOPS": 201888901942035.1, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6412302255630493, "optim/lr": 0.000156, "optim/total_tokens": 68419584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320345, "loss/out": 6.1290059089660645, "created_at": "2025-01-14T13:20:53.012641+00:00"} {"global_step": 262, "acc_step": 0, "speed/wps": 12854.165538352194, "speed/FLOPS": 201892333254270.75, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5690025091171265, "optim/lr": 0.0001566, "optim/total_tokens": 68681728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474052, "loss/out": 6.102691650390625, "created_at": "2025-01-14T13:20:58.114980+00:00"} {"global_step": 263, "acc_step": 0, "speed/wps": 12854.745775591597, "speed/FLOPS": 201901446677445.1, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6207069158554077, "optim/lr": 0.0001572, "optim/total_tokens": 68943872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305203, "loss/out": 6.1505656242370605, "created_at": "2025-01-14T13:21:03.218470+00:00"} {"global_step": 264, "acc_step": 0, "speed/wps": 12856.361850702475, "speed/FLOPS": 201926829357781.34, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5390504002571106, "optim/lr": 0.00015780000000000001, "optim/total_tokens": 69206016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470861, "loss/out": 6.076080322265625, "created_at": "2025-01-14T13:21:08.322195+00:00"} {"global_step": 265, "acc_step": 0, "speed/wps": 12852.918063499985, "speed/FLOPS": 201872739947506.72, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.603278398513794, "optim/lr": 0.0001584, "optim/total_tokens": 69468160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352511, "loss/out": 6.101034641265869, "created_at": "2025-01-14T13:21:13.428079+00:00"} {"global_step": 266, "acc_step": 0, "speed/wps": 12846.669548621561, "speed/FLOPS": 201774598434981.84, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.001, "optim/grad_norm": 0.6223445534706116, "optim/lr": 0.000159, "optim/total_tokens": 69730304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496999, "loss/out": 6.286012649536133, "created_at": "2025-01-14T13:21:18.536207+00:00"} {"global_step": 267, "acc_step": 0, "speed/wps": 12845.49708457844, "speed/FLOPS": 201756183276050.2, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4980131983757019, "optim/lr": 0.0001596, "optim/total_tokens": 69992448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483565, "loss/out": 6.086792469024658, "created_at": "2025-01-14T13:21:23.647564+00:00"} {"global_step": 268, "acc_step": 0, "speed/wps": 12853.155435835923, "speed/FLOPS": 201876468206226.38, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5295194387435913, "optim/lr": 0.00016020000000000002, "optim/total_tokens": 70254592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472444, "loss/out": 6.089341640472412, "created_at": "2025-01-14T13:21:28.753202+00:00"} {"global_step": 269, "acc_step": 0, "speed/wps": 12857.954475967936, "speed/FLOPS": 201951843726071.84, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6028825640678406, "optim/lr": 0.0001608, "optim/total_tokens": 70516736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496480, "loss/out": 6.133152008056641, "created_at": "2025-01-14T13:21:33.854999+00:00"} {"global_step": 270, "acc_step": 0, "speed/wps": 12854.004888891825, "speed/FLOPS": 201889810033740.16, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.49541765451431274, "optim/lr": 0.0001614, "optim/total_tokens": 70778880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498044, "loss/out": 6.111475944519043, "created_at": "2025-01-14T13:21:38.957566+00:00"} {"global_step": 271, "acc_step": 0, "speed/wps": 12851.208028550396, "speed/FLOPS": 201845881498789.16, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5529731512069702, "optim/lr": 0.000162, "optim/total_tokens": 71041024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 395627, "loss/out": 6.276895046234131, "created_at": "2025-01-14T13:21:44.090805+00:00"} {"global_step": 272, "acc_step": 0, "speed/wps": 12855.490007472623, "speed/FLOPS": 201913135861818.38, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.523460328578949, "optim/lr": 0.0001626, "optim/total_tokens": 71303168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470624, "loss/out": 6.1139984130859375, "created_at": "2025-01-14T13:21:49.192399+00:00"} {"global_step": 273, "acc_step": 0, "speed/wps": 12855.623812083217, "speed/FLOPS": 201915237447094.56, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5692545771598816, "optim/lr": 0.00016319999999999998, "optim/total_tokens": 71565312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489068, "loss/out": 5.992871284484863, "created_at": "2025-01-14T13:21:54.299125+00:00"} {"global_step": 274, "acc_step": 0, "speed/wps": 12859.082518906129, "speed/FLOPS": 201969561190509.62, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7329165935516357, "optim/lr": 0.00016380000000000002, "optim/total_tokens": 71827456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299294, "loss/out": 6.15283203125, "created_at": "2025-01-14T13:21:59.405076+00:00"} {"global_step": 275, "acc_step": 0, "speed/wps": 12855.945315743873, "speed/FLOPS": 201920287103877.47, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8433670997619629, "optim/lr": 0.0001644, "optim/total_tokens": 72089600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475867, "loss/out": 6.059289455413818, "created_at": "2025-01-14T13:22:04.507851+00:00"} {"global_step": 276, "acc_step": 0, "speed/wps": 12850.511925880515, "speed/FLOPS": 201834948249813.38, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7359510064125061, "optim/lr": 0.000165, "optim/total_tokens": 72351744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441734, "loss/out": 6.132664203643799, "created_at": "2025-01-14T13:22:09.614358+00:00"} {"global_step": 277, "acc_step": 0, "speed/wps": 12848.567205689553, "speed/FLOPS": 201804403746888.72, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5792891383171082, "optim/lr": 0.00016560000000000001, "optim/total_tokens": 72613888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485892, "loss/out": 6.039268493652344, "created_at": "2025-01-14T13:22:14.720707+00:00"} {"global_step": 278, "acc_step": 0, "speed/wps": 12851.106752766456, "speed/FLOPS": 201844290823434.88, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5746562480926514, "optim/lr": 0.0001662, "optim/total_tokens": 72876032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492835, "loss/out": 6.063000679016113, "created_at": "2025-01-14T13:22:19.825895+00:00"} {"global_step": 279, "acc_step": 0, "speed/wps": 12858.335974913489, "speed/FLOPS": 201957835691246.25, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.578770637512207, "optim/lr": 0.0001668, "optim/total_tokens": 73138176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482471, "loss/out": 5.953780174255371, "created_at": "2025-01-14T13:22:24.929751+00:00"} {"global_step": 280, "acc_step": 0, "speed/wps": 12851.836020046003, "speed/FLOPS": 201855744968178.4, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5245673656463623, "optim/lr": 0.0001674, "optim/total_tokens": 73400320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496999, "loss/out": 6.111576557159424, "created_at": "2025-01-14T13:22:30.033990+00:00"} {"global_step": 281, "acc_step": 0, "speed/wps": 12855.548796328183, "speed/FLOPS": 201914059221579.28, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4831485152244568, "optim/lr": 0.00016800000000000002, "optim/total_tokens": 73662464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330235, "loss/out": 5.982255935668945, "created_at": "2025-01-14T13:22:35.138449+00:00"} {"global_step": 282, "acc_step": 0, "speed/wps": 12846.200385356085, "speed/FLOPS": 201767229581199.66, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.6153652667999268, "optim/lr": 0.0001686, "optim/total_tokens": 73924608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469784, "loss/out": 5.989099502563477, "created_at": "2025-01-14T13:22:40.248198+00:00"} {"global_step": 283, "acc_step": 0, "speed/wps": 12859.28816697988, "speed/FLOPS": 201972791176097.97, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6277735233306885, "optim/lr": 0.0001692, "optim/total_tokens": 74186752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328599, "loss/out": 6.0596818923950195, "created_at": "2025-01-14T13:22:45.348892+00:00"} {"global_step": 284, "acc_step": 0, "speed/wps": 12846.969992930683, "speed/FLOPS": 201779317325710.34, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6381641626358032, "optim/lr": 0.0001698, "optim/total_tokens": 74448896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487982, "loss/out": 6.03579044342041, "created_at": "2025-01-14T13:22:50.453832+00:00"} {"global_step": 285, "acc_step": 0, "speed/wps": 12854.076571223202, "speed/FLOPS": 201890935903256.28, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6190909743309021, "optim/lr": 0.00017040000000000002, "optim/total_tokens": 74711040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473254, "loss/out": 6.052833080291748, "created_at": "2025-01-14T13:22:55.558299+00:00"} {"global_step": 286, "acc_step": 0, "speed/wps": 12851.069084552748, "speed/FLOPS": 201843699192377.03, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.560383677482605, "optim/lr": 0.000171, "optim/total_tokens": 74973184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500702, "loss/out": 6.044416427612305, "created_at": "2025-01-14T13:23:00.665156+00:00"} {"global_step": 287, "acc_step": 0, "speed/wps": 12855.088081964504, "speed/FLOPS": 201906823069409.84, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.5498281121253967, "optim/lr": 0.0001716, "optim/total_tokens": 75235328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299874, "loss/out": 5.976693153381348, "created_at": "2025-01-14T13:23:05.773158+00:00"} {"global_step": 288, "acc_step": 0, "speed/wps": 12850.68932853923, "speed/FLOPS": 201837734602345.28, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5011969208717346, "optim/lr": 0.0001722, "optim/total_tokens": 75497472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476715, "loss/out": 5.949962615966797, "created_at": "2025-01-14T13:23:10.878499+00:00"} {"global_step": 289, "acc_step": 0, "speed/wps": 12846.120463333138, "speed/FLOPS": 201765974296004.9, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5219794511795044, "optim/lr": 0.0001728, "optim/total_tokens": 75759616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432127, "loss/out": 6.012976169586182, "created_at": "2025-01-14T13:23:15.987445+00:00"} {"global_step": 290, "acc_step": 0, "speed/wps": 12857.486562501663, "speed/FLOPS": 201944494502103.56, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5450097322463989, "optim/lr": 0.00017339999999999999, "optim/total_tokens": 76021760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499090, "loss/out": 6.1443400382995605, "created_at": "2025-01-14T13:23:21.088336+00:00"} {"global_step": 291, "acc_step": 0, "speed/wps": 12856.331249880628, "speed/FLOPS": 201926348729824.3, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5787420272827148, "optim/lr": 0.000174, "optim/total_tokens": 76283904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485422, "loss/out": 6.145047664642334, "created_at": "2025-01-14T13:23:26.189742+00:00"} {"global_step": 292, "acc_step": 0, "speed/wps": 12856.195537414165, "speed/FLOPS": 201924217179050.03, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6144761443138123, "optim/lr": 0.00017460000000000002, "optim/total_tokens": 76546048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490923, "loss/out": 6.025234222412109, "created_at": "2025-01-14T13:23:31.291218+00:00"} {"global_step": 293, "acc_step": 0, "speed/wps": 12855.471235641067, "speed/FLOPS": 201912841024408.6, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7352548241615295, "optim/lr": 0.0001752, "optim/total_tokens": 76808192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484328, "loss/out": 6.0890960693359375, "created_at": "2025-01-14T13:23:36.400717+00:00"} {"global_step": 294, "acc_step": 0, "speed/wps": 12851.273868523336, "speed/FLOPS": 201846915606037.78, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7712165713310242, "optim/lr": 0.0001758, "optim/total_tokens": 77070336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476387, "loss/out": 6.0462727546691895, "created_at": "2025-01-14T13:23:41.506913+00:00"} {"global_step": 295, "acc_step": 0, "speed/wps": 12854.208024171525, "speed/FLOPS": 201893000552444.97, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8034048676490784, "optim/lr": 0.0001764, "optim/total_tokens": 77332480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492251, "loss/out": 6.0416669845581055, "created_at": "2025-01-14T13:23:46.618074+00:00"} {"global_step": 296, "acc_step": 0, "speed/wps": 12856.687614735116, "speed/FLOPS": 201931945929560.66, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6040025353431702, "optim/lr": 0.000177, "optim/total_tokens": 77594624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462654, "loss/out": 6.055532455444336, "created_at": "2025-01-14T13:23:51.722699+00:00"} {"global_step": 297, "acc_step": 0, "speed/wps": 12856.809111913935, "speed/FLOPS": 201933854209708.62, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7092100977897644, "optim/lr": 0.0001776, "optim/total_tokens": 77856768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499899, "loss/out": 6.007391452789307, "created_at": "2025-01-14T13:23:56.827632+00:00"} {"global_step": 298, "acc_step": 0, "speed/wps": 12849.380862380074, "speed/FLOPS": 201817183343294.8, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9182506203651428, "optim/lr": 0.0001782, "optim/total_tokens": 78118912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471398, "loss/out": 6.039802551269531, "created_at": "2025-01-14T13:24:01.931781+00:00"} {"global_step": 299, "acc_step": 0, "speed/wps": 12860.591896704233, "speed/FLOPS": 201993268042931.28, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8332247734069824, "optim/lr": 0.0001788, "optim/total_tokens": 78381056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488266, "loss/out": 6.091609954833984, "created_at": "2025-01-14T13:24:07.034709+00:00"} {"global_step": 300, "acc_step": 0, "speed/wps": 12846.578365740777, "speed/FLOPS": 201773166282540.72, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.650550127029419, "optim/lr": 0.0001794, "optim/total_tokens": 78643200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327304, "loss/out": 6.10545539855957, "created_at": "2025-01-14T13:24:12.141041+00:00"} {"global_step": 301, "acc_step": 0, "speed/wps": 12846.44533097554, "speed/FLOPS": 201771076788743.75, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.725060224533081, "optim/lr": 0.00017999999999999998, "optim/total_tokens": 78905344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468737, "loss/out": 6.0522613525390625, "created_at": "2025-01-14T13:24:17.247560+00:00"} {"global_step": 302, "acc_step": 0, "speed/wps": 12852.263473912968, "speed/FLOPS": 201862458718542.25, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.6585635542869568, "optim/lr": 0.0001806, "optim/total_tokens": 79167488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350408, "loss/out": 6.109155178070068, "created_at": "2025-01-14T13:24:22.351402+00:00"} {"global_step": 303, "acc_step": 0, "speed/wps": 12847.058378548485, "speed/FLOPS": 201780705543292.5, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6974268555641174, "optim/lr": 0.0001812, "optim/total_tokens": 79429632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489891, "loss/out": 6.089820861816406, "created_at": "2025-01-14T13:24:27.464814+00:00"} {"global_step": 304, "acc_step": 0, "speed/wps": 12848.172325927773, "speed/FLOPS": 201798201617608.3, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.535826563835144, "optim/lr": 0.0001818, "optim/total_tokens": 79691776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489891, "loss/out": 5.985036849975586, "created_at": "2025-01-14T13:24:32.573959+00:00"} {"global_step": 305, "acc_step": 0, "speed/wps": 12858.239424458427, "speed/FLOPS": 201956319233675.28, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.542055606842041, "optim/lr": 0.00018240000000000002, "optim/total_tokens": 79953920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 6.042738437652588, "created_at": "2025-01-14T13:24:37.680493+00:00"} {"global_step": 306, "acc_step": 0, "speed/wps": 12850.441872012005, "speed/FLOPS": 201833847957544.3, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.740977942943573, "optim/lr": 0.000183, "optim/total_tokens": 80216064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487220, "loss/out": 6.039855003356934, "created_at": "2025-01-14T13:24:42.785737+00:00"} {"global_step": 307, "acc_step": 0, "speed/wps": 12849.752696903488, "speed/FLOPS": 201823023515439.22, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7555776238441467, "optim/lr": 0.0001836, "optim/total_tokens": 80478208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304391, "loss/out": 6.067751884460449, "created_at": "2025-01-14T13:24:47.889741+00:00"} {"global_step": 308, "acc_step": 0, "speed/wps": 12851.68666459512, "speed/FLOPS": 201853399135586.38, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5549353361129761, "optim/lr": 0.0001842, "optim/total_tokens": 80740352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469548, "loss/out": 6.008078098297119, "created_at": "2025-01-14T13:24:53.001226+00:00"} {"global_step": 309, "acc_step": 0, "speed/wps": 12858.99968252268, "speed/FLOPS": 201968260131278.44, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5500532388687134, "optim/lr": 0.00018480000000000002, "optim/total_tokens": 81002496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318405, "loss/out": 5.966831207275391, "created_at": "2025-01-14T13:24:58.102637+00:00"} {"global_step": 310, "acc_step": 0, "speed/wps": 12852.590168807357, "speed/FLOPS": 201867589910784.47, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6548507213592529, "optim/lr": 0.0001854, "optim/total_tokens": 81264640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499899, "loss/out": 5.971786022186279, "created_at": "2025-01-14T13:25:03.205147+00:00"} {"global_step": 311, "acc_step": 0, "speed/wps": 12853.698076697721, "speed/FLOPS": 201884991126627.62, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7222872376441956, "optim/lr": 0.000186, "optim/total_tokens": 81526784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473815, "loss/out": 5.951832294464111, "created_at": "2025-01-14T13:25:08.307713+00:00"} {"global_step": 312, "acc_step": 0, "speed/wps": 12848.029565244025, "speed/FLOPS": 201795959364896.3, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6730996370315552, "optim/lr": 0.0001866, "optim/total_tokens": 81788928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502030, "loss/out": 5.957481384277344, "created_at": "2025-01-14T13:25:13.412487+00:00"} {"global_step": 313, "acc_step": 0, "speed/wps": 12850.315445157512, "speed/FLOPS": 201831862250065.78, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6639809012413025, "optim/lr": 0.0001872, "optim/total_tokens": 82051072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325144, "loss/out": 6.009283065795898, "created_at": "2025-01-14T13:25:18.519738+00:00"} {"global_step": 314, "acc_step": 0, "speed/wps": 12850.698490389017, "speed/FLOPS": 201837878501785.88, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5697249174118042, "optim/lr": 0.0001878, "optim/total_tokens": 82313216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482473, "loss/out": 5.97756290435791, "created_at": "2025-01-14T13:25:23.626048+00:00"} {"global_step": 315, "acc_step": 0, "speed/wps": 12849.14999004332, "speed/FLOPS": 201813557175994.25, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5226783156394958, "optim/lr": 0.0001884, "optim/total_tokens": 82575360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351288, "loss/out": 5.933079719543457, "created_at": "2025-01-14T13:25:28.732776+00:00"} {"global_step": 316, "acc_step": 0, "speed/wps": 12849.990205759526, "speed/FLOPS": 201826753918394.66, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.603458046913147, "optim/lr": 0.000189, "optim/total_tokens": 82837504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492025, "loss/out": 6.046727180480957, "created_at": "2025-01-14T13:25:33.838998+00:00"} {"global_step": 317, "acc_step": 0, "speed/wps": 12844.769575164317, "speed/FLOPS": 201744756740998.7, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5769224762916565, "optim/lr": 0.00018960000000000003, "optim/total_tokens": 83099648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485138, "loss/out": 5.965921878814697, "created_at": "2025-01-14T13:25:38.949743+00:00"} {"global_step": 318, "acc_step": 0, "speed/wps": 12847.577315594992, "speed/FLOPS": 201788856162702.1, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5439381003379822, "optim/lr": 0.0001902, "optim/total_tokens": 83361792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486988, "loss/out": 5.958599090576172, "created_at": "2025-01-14T13:25:44.057352+00:00"} {"global_step": 319, "acc_step": 0, "speed/wps": 12849.873483463512, "speed/FLOPS": 201824920634342.53, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.46726083755493164, "optim/lr": 0.0001908, "optim/total_tokens": 83623936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481431, "loss/out": 6.014660835266113, "created_at": "2025-01-14T13:25:49.165378+00:00"} {"global_step": 320, "acc_step": 0, "speed/wps": 12843.617156178974, "speed/FLOPS": 201726656417244.25, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5642704963684082, "optim/lr": 0.0001914, "optim/total_tokens": 83886080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356011, "loss/out": 5.958611488342285, "created_at": "2025-01-14T13:25:54.274090+00:00"} {"global_step": 321, "acc_step": 0, "speed/wps": 12858.079558908157, "speed/FLOPS": 201953808325530.0, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5500016212463379, "optim/lr": 0.000192, "optim/total_tokens": 84148224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493013, "loss/out": 5.932612419128418, "created_at": "2025-01-14T13:25:59.377573+00:00"} {"global_step": 322, "acc_step": 0, "speed/wps": 12847.346482262672, "speed/FLOPS": 201785230608019.78, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.641975998878479, "optim/lr": 0.0001926, "optim/total_tokens": 84410368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380182, "loss/out": 5.856934547424316, "created_at": "2025-01-14T13:26:04.484689+00:00"} {"global_step": 323, "acc_step": 0, "speed/wps": 12853.782866343527, "speed/FLOPS": 201886322864528.2, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5401908755302429, "optim/lr": 0.0001932, "optim/total_tokens": 84672512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486752, "loss/out": 5.950081825256348, "created_at": "2025-01-14T13:26:09.590979+00:00"} {"global_step": 324, "acc_step": 0, "speed/wps": 12850.43871336368, "speed/FLOPS": 201833798346631.9, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.565959632396698, "optim/lr": 0.00019380000000000002, "optim/total_tokens": 84934656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290218, "loss/out": 5.923857688903809, "created_at": "2025-01-14T13:26:14.695550+00:00"} {"global_step": 325, "acc_step": 0, "speed/wps": 12849.322432593872, "speed/FLOPS": 201816265623211.56, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5467109084129333, "optim/lr": 0.00019439999999999998, "optim/total_tokens": 85196800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465834, "loss/out": 5.949615478515625, "created_at": "2025-01-14T13:26:19.802454+00:00"} {"global_step": 326, "acc_step": 0, "speed/wps": 12851.533451648216, "speed/FLOPS": 201850992715718.62, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5052239298820496, "optim/lr": 0.00019500000000000002, "optim/total_tokens": 85458944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362914, "loss/out": 6.048948287963867, "created_at": "2025-01-14T13:26:24.909676+00:00"} {"global_step": 327, "acc_step": 0, "speed/wps": 12847.580592464868, "speed/FLOPS": 201788907630446.4, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6060497164726257, "optim/lr": 0.00019559999999999998, "optim/total_tokens": 85721088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491968, "loss/out": 6.090005874633789, "created_at": "2025-01-14T13:26:30.015347+00:00"} {"global_step": 328, "acc_step": 0, "speed/wps": 12851.35956265366, "speed/FLOPS": 201848261550110.6, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5969873666763306, "optim/lr": 0.0001962, "optim/total_tokens": 85983232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480385, "loss/out": 6.016261100769043, "created_at": "2025-01-14T13:26:35.119034+00:00"} {"global_step": 329, "acc_step": 0, "speed/wps": 12847.90707747815, "speed/FLOPS": 201794035526217.44, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6017711162567139, "optim/lr": 0.0001968, "optim/total_tokens": 86245376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466597, "loss/out": 5.9463348388671875, "created_at": "2025-01-14T13:26:40.223868+00:00"} {"global_step": 330, "acc_step": 0, "speed/wps": 12854.044544297447, "speed/FLOPS": 201890432876377.7, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6544436812400818, "optim/lr": 0.0001974, "optim/total_tokens": 86507520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499373, "loss/out": 5.878925800323486, "created_at": "2025-01-14T13:26:45.328221+00:00"} {"global_step": 331, "acc_step": 0, "speed/wps": 12849.487533363577, "speed/FLOPS": 201818858757672.03, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.545063853263855, "optim/lr": 0.00019800000000000002, "optim/total_tokens": 86769664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329704, "loss/out": 5.938628196716309, "created_at": "2025-01-14T13:26:50.432226+00:00"} {"global_step": 332, "acc_step": 0, "speed/wps": 12851.148917958457, "speed/FLOPS": 201844953085716.28, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.517419159412384, "optim/lr": 0.00019859999999999998, "optim/total_tokens": 87031808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477481, "loss/out": 5.879164218902588, "created_at": "2025-01-14T13:26:55.538151+00:00"} {"global_step": 333, "acc_step": 0, "speed/wps": 12850.820428250048, "speed/FLOPS": 201839793703453.56, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5375586748123169, "optim/lr": 0.00019920000000000002, "optim/total_tokens": 87293952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308138, "loss/out": 5.913205146789551, "created_at": "2025-01-14T13:27:00.645970+00:00"} {"global_step": 334, "acc_step": 0, "speed/wps": 12850.103698948731, "speed/FLOPS": 201828536484887.12, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.47427117824554443, "optim/lr": 0.00019980000000000003, "optim/total_tokens": 87556096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497305, "loss/out": 5.925684928894043, "created_at": "2025-01-14T13:27:05.752768+00:00"} {"global_step": 335, "acc_step": 0, "speed/wps": 12850.511822624889, "speed/FLOPS": 201834946628041.88, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5137743949890137, "optim/lr": 0.0002004, "optim/total_tokens": 87818240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420458, "loss/out": 5.897026062011719, "created_at": "2025-01-14T13:27:10.856526+00:00"} {"global_step": 336, "acc_step": 0, "speed/wps": 12850.172111206875, "speed/FLOPS": 201829610993409.56, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.001, "optim/grad_norm": 0.4559541642665863, "optim/lr": 0.000201, "optim/total_tokens": 88080384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497069, "loss/out": 5.808873176574707, "created_at": "2025-01-14T13:27:15.963322+00:00"} {"global_step": 337, "acc_step": 0, "speed/wps": 12857.163433660578, "speed/FLOPS": 201939419319628.78, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.47275039553642273, "optim/lr": 0.0002016, "optim/total_tokens": 88342528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 285748, "loss/out": 5.984225749969482, "created_at": "2025-01-14T13:27:21.064401+00:00"} {"global_step": 338, "acc_step": 0, "speed/wps": 12849.011754510191, "speed/FLOPS": 201811385996990.94, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5509135127067566, "optim/lr": 0.0002022, "optim/total_tokens": 88604672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471398, "loss/out": 5.912846088409424, "created_at": "2025-01-14T13:27:26.172559+00:00"} {"global_step": 339, "acc_step": 0, "speed/wps": 12851.71446427498, "speed/FLOPS": 201853835767759.38, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5624985098838806, "optim/lr": 0.0002028, "optim/total_tokens": 88866816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 387884, "loss/out": 5.955913066864014, "created_at": "2025-01-14T13:27:31.275604+00:00"} {"global_step": 340, "acc_step": 0, "speed/wps": 12856.609004420527, "speed/FLOPS": 201930711246548.16, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5748299360275269, "optim/lr": 0.0002034, "optim/total_tokens": 89128960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489068, "loss/out": 5.8590898513793945, "created_at": "2025-01-14T13:27:36.379795+00:00"} {"global_step": 341, "acc_step": 0, "speed/wps": 12848.35266443918, "speed/FLOPS": 201801034081741.06, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5164408683776855, "optim/lr": 0.00020400000000000003, "optim/total_tokens": 89391104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472161, "loss/out": 5.883159637451172, "created_at": "2025-01-14T13:27:41.484666+00:00"} {"global_step": 342, "acc_step": 0, "speed/wps": 12849.57744122883, "speed/FLOPS": 201820270884242.2, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5589476227760315, "optim/lr": 0.00020459999999999999, "optim/total_tokens": 89653248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479331, "loss/out": 5.920080184936523, "created_at": "2025-01-14T13:27:46.589879+00:00"} {"global_step": 343, "acc_step": 0, "speed/wps": 12851.714769405535, "speed/FLOPS": 201853840560254.0, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6550837755203247, "optim/lr": 0.0002052, "optim/total_tokens": 89915392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493816, "loss/out": 5.834373474121094, "created_at": "2025-01-14T13:27:51.694382+00:00"} {"global_step": 344, "acc_step": 0, "speed/wps": 12848.240536989179, "speed/FLOPS": 201799272966059.5, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5815237760543823, "optim/lr": 0.0002058, "optim/total_tokens": 90177536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324857, "loss/out": 5.768387317657471, "created_at": "2025-01-14T13:27:56.798808+00:00"} {"global_step": 345, "acc_step": 0, "speed/wps": 12852.359434648777, "speed/FLOPS": 201863965913762.0, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.001, "optim/grad_norm": 0.9875274896621704, "optim/lr": 0.0002064, "optim/total_tokens": 90439680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477718, "loss/out": 5.877035140991211, "created_at": "2025-01-14T13:28:01.901613+00:00"} {"global_step": 346, "acc_step": 0, "speed/wps": 12855.354742539204, "speed/FLOPS": 201911011340165.53, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7918460965156555, "optim/lr": 0.00020700000000000002, "optim/total_tokens": 90701824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304559, "loss/out": 5.806723594665527, "created_at": "2025-01-14T13:28:07.006588+00:00"} {"global_step": 347, "acc_step": 0, "speed/wps": 12853.290302330479, "speed/FLOPS": 201878586469849.38, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6899006366729736, "optim/lr": 0.0002076, "optim/total_tokens": 90963968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468211, "loss/out": 5.862010478973389, "created_at": "2025-01-14T13:28:12.109033+00:00"} {"global_step": 348, "acc_step": 0, "speed/wps": 12845.388036506538, "speed/FLOPS": 201754470526233.22, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0961660146713257, "optim/lr": 0.00020820000000000002, "optim/total_tokens": 91226112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369954, "loss/out": 5.977365493774414, "created_at": "2025-01-14T13:28:17.220820+00:00"} {"global_step": 349, "acc_step": 0, "speed/wps": 12853.913554893137, "speed/FLOPS": 201888375507784.7, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8194161653518677, "optim/lr": 0.00020879999999999998, "optim/total_tokens": 91488256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497589, "loss/out": 6.001051425933838, "created_at": "2025-01-14T13:28:22.327248+00:00"} {"global_step": 350, "acc_step": 0, "speed/wps": 12860.938611472346, "speed/FLOPS": 201998713674800.5, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5845593214035034, "optim/lr": 0.0002094, "optim/total_tokens": 91750400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295547, "loss/out": 5.884139060974121, "created_at": "2025-01-14T13:28:27.426400+00:00"} {"global_step": 351, "acc_step": 0, "speed/wps": 12854.45372845586, "speed/FLOPS": 201896859675863.5, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9417420625686646, "optim/lr": 0.00021000000000000004, "optim/total_tokens": 92012544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478244, "loss/out": 5.800256252288818, "created_at": "2025-01-14T13:28:32.530831+00:00"} {"global_step": 352, "acc_step": 0, "speed/wps": 12853.54425700604, "speed/FLOPS": 201882575176999.38, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7620171904563904, "optim/lr": 0.0002106, "optim/total_tokens": 92274688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 342402, "loss/out": 5.908769130706787, "created_at": "2025-01-14T13:28:37.634278+00:00"} {"global_step": 353, "acc_step": 0, "speed/wps": 12848.236173636538, "speed/FLOPS": 201799204433610.4, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5899573564529419, "optim/lr": 0.0002112, "optim/total_tokens": 92536832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496190, "loss/out": 5.974118232727051, "created_at": "2025-01-14T13:28:42.739594+00:00"} {"global_step": 354, "acc_step": 0, "speed/wps": 12855.572042405427, "speed/FLOPS": 201914424333165.88, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8399681448936462, "optim/lr": 0.0002118, "optim/total_tokens": 92798976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486752, "loss/out": 5.931621551513672, "created_at": "2025-01-14T13:28:47.845774+00:00"} {"global_step": 355, "acc_step": 0, "speed/wps": 12859.844545265565, "speed/FLOPS": 201981529861618.5, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6141158938407898, "optim/lr": 0.0002124, "optim/total_tokens": 93061120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489648, "loss/out": 5.923357009887695, "created_at": "2025-01-14T13:28:52.952628+00:00"} {"global_step": 356, "acc_step": 0, "speed/wps": 12852.0933278778, "speed/FLOPS": 201859786341254.78, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7175608277320862, "optim/lr": 0.00021299999999999997, "optim/total_tokens": 93323264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476678, "loss/out": 5.930638313293457, "created_at": "2025-01-14T13:28:58.055451+00:00"} {"global_step": 357, "acc_step": 0, "speed/wps": 12853.885994581555, "speed/FLOPS": 201887942635219.47, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5317423343658447, "optim/lr": 0.00021360000000000001, "optim/total_tokens": 93585408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486468, "loss/out": 5.873422622680664, "created_at": "2025-01-14T13:29:03.159221+00:00"} {"global_step": 358, "acc_step": 0, "speed/wps": 12847.786100347697, "speed/FLOPS": 201792135414143.66, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6191198229789734, "optim/lr": 0.00021420000000000003, "optim/total_tokens": 93847552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488266, "loss/out": 5.88679313659668, "created_at": "2025-01-14T13:29:08.263994+00:00"} {"global_step": 359, "acc_step": 0, "speed/wps": 12856.817147914096, "speed/FLOPS": 201933980426133.78, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5699721574783325, "optim/lr": 0.0002148, "optim/total_tokens": 94109696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485941, "loss/out": 5.789677143096924, "created_at": "2025-01-14T13:29:13.365771+00:00"} {"global_step": 360, "acc_step": 0, "speed/wps": 12859.638653791035, "speed/FLOPS": 201978296053086.47, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.47689351439476013, "optim/lr": 0.0002154, "optim/total_tokens": 94371840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504647, "loss/out": 5.999324798583984, "created_at": "2025-01-14T13:29:18.468246+00:00"} {"global_step": 361, "acc_step": 0, "speed/wps": 12853.337506037453, "speed/FLOPS": 201879327868932.1, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.45175838470458984, "optim/lr": 0.000216, "optim/total_tokens": 94633984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327689, "loss/out": 5.903626441955566, "created_at": "2025-01-14T13:29:23.570620+00:00"} {"global_step": 362, "acc_step": 0, "speed/wps": 12852.044314167695, "speed/FLOPS": 201859016513585.7, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5068917870521545, "optim/lr": 0.0002166, "optim/total_tokens": 94896128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505975, "loss/out": 6.031391143798828, "created_at": "2025-01-14T13:29:28.673923+00:00"} {"global_step": 363, "acc_step": 0, "speed/wps": 12855.18322301255, "speed/FLOPS": 201908317390306.88, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4754778742790222, "optim/lr": 0.00021720000000000002, "optim/total_tokens": 95158272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501228, "loss/out": 5.887032985687256, "created_at": "2025-01-14T13:29:33.775518+00:00"} {"global_step": 364, "acc_step": 0, "speed/wps": 12850.399587085643, "speed/FLOPS": 201833183814670.22, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4410298466682434, "optim/lr": 0.0002178, "optim/total_tokens": 95420416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469784, "loss/out": 5.87729024887085, "created_at": "2025-01-14T13:29:38.881545+00:00"} {"global_step": 365, "acc_step": 0, "speed/wps": 12844.59801208591, "speed/FLOPS": 201742062107099.94, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6780738830566406, "optim/lr": 0.00021840000000000002, "optim/total_tokens": 95682560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470587, "loss/out": 5.911011695861816, "created_at": "2025-01-14T13:29:43.991583+00:00"} {"global_step": 366, "acc_step": 0, "speed/wps": 12844.900711211352, "speed/FLOPS": 201746816412817.2, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7361210584640503, "optim/lr": 0.00021899999999999998, "optim/total_tokens": 95944704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502556, "loss/out": 6.005379676818848, "created_at": "2025-01-14T13:29:49.098732+00:00"} {"global_step": 367, "acc_step": 0, "speed/wps": 12848.713886859969, "speed/FLOPS": 201806707576231.16, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6397921442985535, "optim/lr": 0.0002196, "optim/total_tokens": 96206848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313679, "loss/out": 5.9209418296813965, "created_at": "2025-01-14T13:29:54.205558+00:00"} {"global_step": 368, "acc_step": 0, "speed/wps": 12851.348477647985, "speed/FLOPS": 201848087444864.1, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6900050640106201, "optim/lr": 0.0002202, "optim/total_tokens": 96468992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479906, "loss/out": 5.843560695648193, "created_at": "2025-01-14T13:29:59.311190+00:00"} {"global_step": 369, "acc_step": 0, "speed/wps": 12846.017783285226, "speed/FLOPS": 201764361564755.28, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4821196496486664, "optim/lr": 0.0002208, "optim/total_tokens": 96731136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 384441, "loss/out": 6.029180526733398, "created_at": "2025-01-14T13:30:04.416464+00:00"} {"global_step": 370, "acc_step": 0, "speed/wps": 12851.483559708608, "speed/FLOPS": 201850209094248.34, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5012391209602356, "optim/lr": 0.00022140000000000001, "optim/total_tokens": 96993280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485658, "loss/out": 5.826747894287109, "created_at": "2025-01-14T13:30:09.520524+00:00"} {"global_step": 371, "acc_step": 0, "speed/wps": 12849.25476610801, "speed/FLOPS": 201815202828067.62, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.47276484966278076, "optim/lr": 0.000222, "optim/total_tokens": 97255424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483045, "loss/out": 5.825441360473633, "created_at": "2025-01-14T13:30:14.626737+00:00"} {"global_step": 372, "acc_step": 0, "speed/wps": 12849.390885806026, "speed/FLOPS": 201817340774973.03, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.47882726788520813, "optim/lr": 0.00022260000000000002, "optim/total_tokens": 97517568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470304, "loss/out": 5.818747043609619, "created_at": "2025-01-14T13:30:19.731065+00:00"} {"global_step": 373, "acc_step": 0, "speed/wps": 12850.787356785613, "speed/FLOPS": 201839274270659.66, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5379567742347717, "optim/lr": 0.00022319999999999998, "optim/total_tokens": 97779712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495151, "loss/out": 5.820034503936768, "created_at": "2025-01-14T13:30:24.845247+00:00"} {"global_step": 374, "acc_step": 0, "speed/wps": 12857.700621867238, "speed/FLOPS": 201947856598588.3, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.46108585596084595, "optim/lr": 0.0002238, "optim/total_tokens": 98041856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308833, "loss/out": 5.897951126098633, "created_at": "2025-01-14T13:30:29.946397+00:00"} {"global_step": 375, "acc_step": 0, "speed/wps": 12852.346124940581, "speed/FLOPS": 201863756866505.34, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5096551775932312, "optim/lr": 0.00022440000000000003, "optim/total_tokens": 98304000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484320, "loss/out": 5.758057117462158, "created_at": "2025-01-14T13:30:35.050270+00:00"} {"global_step": 376, "acc_step": 0, "speed/wps": 12846.045234823468, "speed/FLOPS": 201764792728885.72, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5705053210258484, "optim/lr": 0.000225, "optim/total_tokens": 98566144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481715, "loss/out": 5.8030571937561035, "created_at": "2025-01-14T13:30:40.155770+00:00"} {"global_step": 377, "acc_step": 0, "speed/wps": 12857.305988638314, "speed/FLOPS": 201941658341445.2, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5361970067024231, "optim/lr": 0.0002256, "optim/total_tokens": 98828288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499090, "loss/out": 5.91694450378418, "created_at": "2025-01-14T13:30:45.262566+00:00"} {"global_step": 378, "acc_step": 0, "speed/wps": 12855.862089966973, "speed/FLOPS": 201918979928687.06, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6096207499504089, "optim/lr": 0.0002262, "optim/total_tokens": 99090432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483281, "loss/out": 5.884271621704102, "created_at": "2025-01-14T13:30:50.368102+00:00"} {"global_step": 379, "acc_step": 0, "speed/wps": 12853.774052281784, "speed/FLOPS": 201886184427576.72, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0713454484939575, "optim/lr": 0.0002268, "optim/total_tokens": 99352576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475341, "loss/out": 5.865962028503418, "created_at": "2025-01-14T13:30:55.474208+00:00"} {"global_step": 380, "acc_step": 0, "speed/wps": 12857.77090760921, "speed/FLOPS": 201948960532748.72, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8933634161949158, "optim/lr": 0.00022740000000000003, "optim/total_tokens": 99614720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307080, "loss/out": 5.770430564880371, "created_at": "2025-01-14T13:31:00.577793+00:00"} {"global_step": 381, "acc_step": 0, "speed/wps": 12850.794829063005, "speed/FLOPS": 201839391633044.06, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7321620583534241, "optim/lr": 0.000228, "optim/total_tokens": 99876864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471114, "loss/out": 5.9482340812683105, "created_at": "2025-01-14T13:31:05.690798+00:00"} {"global_step": 382, "acc_step": 0, "speed/wps": 12847.02378548683, "speed/FLOPS": 201780162211723.34, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8693387508392334, "optim/lr": 0.00022860000000000003, "optim/total_tokens": 100139008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473254, "loss/out": 5.85022497177124, "created_at": "2025-01-14T13:31:10.796018+00:00"} {"global_step": 383, "acc_step": 0, "speed/wps": 12852.370845339427, "speed/FLOPS": 201864145134338.6, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7181880474090576, "optim/lr": 0.0002292, "optim/total_tokens": 100401152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487982, "loss/out": 5.796220302581787, "created_at": "2025-01-14T13:31:15.900854+00:00"} {"global_step": 384, "acc_step": 0, "speed/wps": 12851.541783829995, "speed/FLOPS": 201851123584083.06, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6870990991592407, "optim/lr": 0.0002298, "optim/total_tokens": 100663296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494118, "loss/out": 5.943824768066406, "created_at": "2025-01-14T13:31:21.006195+00:00"} {"global_step": 385, "acc_step": 0, "speed/wps": 12846.691976914482, "speed/FLOPS": 201774950702141.88, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6832937598228455, "optim/lr": 0.0002304, "optim/total_tokens": 100925440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327543, "loss/out": 5.808470249176025, "created_at": "2025-01-14T13:31:26.111104+00:00"} {"global_step": 386, "acc_step": 0, "speed/wps": 12850.388705549914, "speed/FLOPS": 201833012905198.8, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7815094590187073, "optim/lr": 0.000231, "optim/total_tokens": 101187584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474301, "loss/out": 5.899377346038818, "created_at": "2025-01-14T13:31:31.214586+00:00"} {"global_step": 387, "acc_step": 0, "speed/wps": 12852.623550030246, "speed/FLOPS": 201868114208760.56, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6363404989242554, "optim/lr": 0.00023160000000000002, "optim/total_tokens": 101449728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305736, "loss/out": 5.7328057289123535, "created_at": "2025-01-14T13:31:36.324407+00:00"} {"global_step": 388, "acc_step": 0, "speed/wps": 12852.367134104552, "speed/FLOPS": 201864086844295.12, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.952714741230011, "optim/lr": 0.0002322, "optim/total_tokens": 101711872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470067, "loss/out": 5.895898818969727, "created_at": "2025-01-14T13:31:41.427121+00:00"} {"global_step": 389, "acc_step": 0, "speed/wps": 12836.411747966507, "speed/FLOPS": 201613485579999.0, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9074615836143494, "optim/lr": 0.00023280000000000002, "optim/total_tokens": 101974016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305274, "loss/out": 5.860330104827881, "created_at": "2025-01-14T13:31:46.536294+00:00"} {"global_step": 390, "acc_step": 0, "speed/wps": 12846.590071055793, "speed/FLOPS": 201773350130597.8, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9663647413253784, "optim/lr": 0.00023339999999999998, "optim/total_tokens": 102236160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474258, "loss/out": 5.872204780578613, "created_at": "2025-01-14T13:31:51.643948+00:00"} {"global_step": 391, "acc_step": 0, "speed/wps": 12853.03630838931, "speed/FLOPS": 201874597145979.06, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.937035322189331, "optim/lr": 0.000234, "optim/total_tokens": 102498304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330474, "loss/out": 5.768887519836426, "created_at": "2025-01-14T13:31:56.746509+00:00"} {"global_step": 392, "acc_step": 0, "speed/wps": 12850.13308276209, "speed/FLOPS": 201828997998053.5, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6866153478622437, "optim/lr": 0.0002346, "optim/total_tokens": 102760448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467927, "loss/out": 5.9348368644714355, "created_at": "2025-01-14T13:32:01.852484+00:00"} {"global_step": 393, "acc_step": 0, "speed/wps": 12846.584977095205, "speed/FLOPS": 201773270122946.94, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9031076431274414, "optim/lr": 0.0002352, "optim/total_tokens": 103022592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 405812, "loss/out": 5.896429061889648, "created_at": "2025-01-14T13:32:06.961953+00:00"} {"global_step": 394, "acc_step": 0, "speed/wps": 12849.934797426153, "speed/FLOPS": 201825883654380.0, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5396490693092346, "optim/lr": 0.0002358, "optim/total_tokens": 103284736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498044, "loss/out": 5.952649116516113, "created_at": "2025-01-14T13:32:12.068486+00:00"} {"global_step": 395, "acc_step": 0, "speed/wps": 12842.802408034224, "speed/FLOPS": 201713859678049.66, "speed/curr_iter_time": 1.2874, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6953817009925842, "optim/lr": 0.0002364, "optim/total_tokens": 103546880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501464, "loss/out": 5.906656265258789, "created_at": "2025-01-14T13:32:17.175063+00:00"} {"global_step": 396, "acc_step": 0, "speed/wps": 12843.586062768192, "speed/FLOPS": 201726168052501.1, "speed/curr_iter_time": 1.2874, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6882737278938293, "optim/lr": 0.00023700000000000001, "optim/total_tokens": 103809024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310826, "loss/out": 5.813792705535889, "created_at": "2025-01-14T13:32:22.282786+00:00"} {"global_step": 397, "acc_step": 0, "speed/wps": 12848.59536958457, "speed/FLOPS": 201804846099561.2, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.42080163955688477, "optim/lr": 0.00023760000000000003, "optim/total_tokens": 104071168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489891, "loss/out": 5.883444786071777, "created_at": "2025-01-14T13:32:27.387724+00:00"} {"global_step": 398, "acc_step": 0, "speed/wps": 12846.996841071026, "speed/FLOPS": 201779739012647.66, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5410898923873901, "optim/lr": 0.0002382, "optim/total_tokens": 104333312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350932, "loss/out": 5.748776435852051, "created_at": "2025-01-14T13:32:32.493004+00:00"} {"global_step": 399, "acc_step": 0, "speed/wps": 12850.424349277471, "speed/FLOPS": 201833572738919.7, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.49167177081108093, "optim/lr": 0.0002388, "optim/total_tokens": 104595456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498919, "loss/out": 5.781248569488525, "created_at": "2025-01-14T13:32:37.598553+00:00"} {"global_step": 400, "acc_step": 0, "speed/wps": 12845.858148794585, "speed/FLOPS": 201761854285718.97, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5554496645927429, "optim/lr": 0.0002394, "optim/total_tokens": 104857600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471924, "loss/out": 5.701067924499512, "created_at": "2025-01-14T13:32:42.704972+00:00"} {"global_step": 401, "acc_step": 0, "speed/wps": 12848.917915396212, "speed/FLOPS": 201809912124757.44, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5062044858932495, "optim/lr": 0.00024, "optim/total_tokens": 105119744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487272, "loss/out": 5.806643486022949, "created_at": "2025-01-14T13:32:47.811913+00:00"} {"global_step": 402, "acc_step": 0, "speed/wps": 12848.902416836023, "speed/FLOPS": 201809668698572.4, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5654556751251221, "optim/lr": 0.0002406, "optim/total_tokens": 105381888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462391, "loss/out": 5.772132396697998, "created_at": "2025-01-14T13:32:52.916057+00:00"} {"global_step": 403, "acc_step": 0, "speed/wps": 12852.003178242958, "speed/FLOPS": 201858370417361.03, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.437846839427948, "optim/lr": 0.0002412, "optim/total_tokens": 105644032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481951, "loss/out": 5.779170989990234, "created_at": "2025-01-14T13:32:58.020465+00:00"} {"global_step": 404, "acc_step": 0, "speed/wps": 12852.348094400073, "speed/FLOPS": 201863787799572.97, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.44931092858314514, "optim/lr": 0.00024180000000000002, "optim/total_tokens": 105906176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420669, "loss/out": 5.726454734802246, "created_at": "2025-01-14T13:33:03.128965+00:00"} {"global_step": 405, "acc_step": 0, "speed/wps": 12855.417064665236, "speed/FLOPS": 201911990194794.44, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4394870698451996, "optim/lr": 0.00024239999999999998, "optim/total_tokens": 106168320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497808, "loss/out": 5.781374454498291, "created_at": "2025-01-14T13:33:08.233577+00:00"} {"global_step": 406, "acc_step": 0, "speed/wps": 12847.877323831515, "speed/FLOPS": 201793568204311.22, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5533609986305237, "optim/lr": 0.00024300000000000002, "optim/total_tokens": 106430464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472161, "loss/out": 5.753469944000244, "created_at": "2025-01-14T13:33:13.340651+00:00"} {"global_step": 407, "acc_step": 0, "speed/wps": 12847.054189562497, "speed/FLOPS": 201780639749511.28, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5329041481018066, "optim/lr": 0.00024359999999999999, "optim/total_tokens": 106692608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481188, "loss/out": 5.705434799194336, "created_at": "2025-01-14T13:33:18.448776+00:00"} {"global_step": 408, "acc_step": 0, "speed/wps": 12854.652484379842, "speed/FLOPS": 201899981410768.9, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.764904797077179, "optim/lr": 0.0002442, "optim/total_tokens": 106954752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490923, "loss/out": 5.899787425994873, "created_at": "2025-01-14T13:33:23.555056+00:00"} {"global_step": 409, "acc_step": 0, "speed/wps": 12847.268914967786, "speed/FLOPS": 201784012307064.97, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5911387205123901, "optim/lr": 0.00024480000000000004, "optim/total_tokens": 107216896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349591, "loss/out": 5.818717956542969, "created_at": "2025-01-14T13:33:28.662639+00:00"} {"global_step": 410, "acc_step": 0, "speed/wps": 12852.711637853448, "speed/FLOPS": 201869497749069.28, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8399301767349243, "optim/lr": 0.0002454, "optim/total_tokens": 107479040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480621, "loss/out": 5.789139747619629, "created_at": "2025-01-14T13:33:33.774810+00:00"} {"global_step": 411, "acc_step": 0, "speed/wps": 12844.640791058539, "speed/FLOPS": 201742734009649.53, "speed/curr_iter_time": 1.2875, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6853744983673096, "optim/lr": 0.000246, "optim/total_tokens": 107741184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490985, "loss/out": 5.765639781951904, "created_at": "2025-01-14T13:33:38.883761+00:00"} {"global_step": 412, "acc_step": 0, "speed/wps": 12845.867608609886, "speed/FLOPS": 201762002865117.44, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9718315601348877, "optim/lr": 0.0002466, "optim/total_tokens": 108003328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475904, "loss/out": 5.837750434875488, "created_at": "2025-01-14T13:33:43.991267+00:00"} {"global_step": 413, "acc_step": 0, "speed/wps": 12850.59849222773, "speed/FLOPS": 201836307893251.0, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.45396503806114197, "optim/lr": 0.0002472, "optim/total_tokens": 108265472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485366, "loss/out": 5.733856201171875, "created_at": "2025-01-14T13:33:49.095160+00:00"} {"global_step": 414, "acc_step": 0, "speed/wps": 12853.478034553922, "speed/FLOPS": 201881535062389.66, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5895088315010071, "optim/lr": 0.0002478, "optim/total_tokens": 108527616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323276, "loss/out": 5.817571640014648, "created_at": "2025-01-14T13:33:54.197729+00:00"} {"global_step": 415, "acc_step": 0, "speed/wps": 12846.376428375923, "speed/FLOPS": 201769994578734.66, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5630365014076233, "optim/lr": 0.0002484, "optim/total_tokens": 108789760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501747, "loss/out": 5.859096527099609, "created_at": "2025-01-14T13:33:59.302945+00:00"} {"global_step": 416, "acc_step": 0, "speed/wps": 12855.3524692033, "speed/FLOPS": 201910975634301.28, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8705284595489502, "optim/lr": 0.00024900000000000004, "optim/total_tokens": 109051904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497235, "loss/out": 5.866120338439941, "created_at": "2025-01-14T13:34:04.407508+00:00"} {"global_step": 417, "acc_step": 0, "speed/wps": 12847.809189344314, "speed/FLOPS": 201792498058563.12, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7373548746109009, "optim/lr": 0.0002496, "optim/total_tokens": 109314048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501464, "loss/out": 5.8592915534973145, "created_at": "2025-01-14T13:34:09.515253+00:00"} {"global_step": 418, "acc_step": 0, "speed/wps": 12846.53527601217, "speed/FLOPS": 201772489499141.44, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.47716614603996277, "optim/lr": 0.0002502, "optim/total_tokens": 109576192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480385, "loss/out": 5.883798122406006, "created_at": "2025-01-14T13:34:14.622466+00:00"} {"global_step": 419, "acc_step": 0, "speed/wps": 12853.678805305743, "speed/FLOPS": 201884688442934.94, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5842200517654419, "optim/lr": 0.00025079999999999997, "optim/total_tokens": 109838336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494625, "loss/out": 5.636812210083008, "created_at": "2025-01-14T13:34:19.725402+00:00"} {"global_step": 420, "acc_step": 0, "speed/wps": 12853.24135931256, "speed/FLOPS": 201877817752498.22, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5780616998672485, "optim/lr": 0.0002514, "optim/total_tokens": 110100480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324333, "loss/out": 5.658909320831299, "created_at": "2025-01-14T13:34:24.831604+00:00"} {"global_step": 421, "acc_step": 0, "speed/wps": 12854.616993309346, "speed/FLOPS": 201899423974736.94, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5941076874732971, "optim/lr": 0.000252, "optim/total_tokens": 110362624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478528, "loss/out": 5.843301773071289, "created_at": "2025-01-14T13:34:29.933380+00:00"} {"global_step": 422, "acc_step": 0, "speed/wps": 12848.795056431638, "speed/FLOPS": 201807982455892.9, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6208072304725647, "optim/lr": 0.0002526, "optim/total_tokens": 110624768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320059, "loss/out": 5.746062755584717, "created_at": "2025-01-14T13:34:35.037894+00:00"} {"global_step": 423, "acc_step": 0, "speed/wps": 12850.917350775297, "speed/FLOPS": 201841316004898.1, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6759762167930603, "optim/lr": 0.00025320000000000003, "optim/total_tokens": 110886912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487798, "loss/out": 5.735291481018066, "created_at": "2025-01-14T13:34:40.144404+00:00"} {"global_step": 424, "acc_step": 0, "speed/wps": 12851.307444751883, "speed/FLOPS": 201847442966845.7, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.5583060383796692, "optim/lr": 0.0002538, "optim/total_tokens": 111149056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435873, "loss/out": 5.723684787750244, "created_at": "2025-01-14T13:34:45.252864+00:00"} {"global_step": 425, "acc_step": 0, "speed/wps": 12851.668216039106, "speed/FLOPS": 201853109375663.22, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5344261527061462, "optim/lr": 0.0002544, "optim/total_tokens": 111411200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479815, "loss/out": 5.756167411804199, "created_at": "2025-01-14T13:34:50.355964+00:00"} {"global_step": 426, "acc_step": 0, "speed/wps": 12850.81283625462, "speed/FLOPS": 201839674460732.88, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5598607659339905, "optim/lr": 0.000255, "optim/total_tokens": 111673344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492268, "loss/out": 5.794369220733643, "created_at": "2025-01-14T13:34:55.459209+00:00"} {"global_step": 427, "acc_step": 0, "speed/wps": 12848.743651419647, "speed/FLOPS": 201807175069541.7, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5564725399017334, "optim/lr": 0.0002556, "optim/total_tokens": 111935488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486988, "loss/out": 5.782617568969727, "created_at": "2025-01-14T13:35:00.565672+00:00"} {"global_step": 428, "acc_step": 0, "speed/wps": 12854.484935561086, "speed/FLOPS": 201897349826334.16, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6201850175857544, "optim/lr": 0.0002562, "optim/total_tokens": 112197632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498635, "loss/out": 5.759016513824463, "created_at": "2025-01-14T13:35:05.669889+00:00"} {"global_step": 429, "acc_step": 0, "speed/wps": 12854.20631477755, "speed/FLOPS": 201892973704063.62, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8064281344413757, "optim/lr": 0.0002568, "optim/total_tokens": 112459776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488318, "loss/out": 5.832056045532227, "created_at": "2025-01-14T13:35:10.775525+00:00"} {"global_step": 430, "acc_step": 0, "speed/wps": 12851.43415399688, "speed/FLOPS": 201849433109655.1, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5605369806289673, "optim/lr": 0.0002574, "optim/total_tokens": 112721920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467407, "loss/out": 5.7597551345825195, "created_at": "2025-01-14T13:35:15.880067+00:00"} {"global_step": 431, "acc_step": 0, "speed/wps": 12848.71586459068, "speed/FLOPS": 201806738639209.66, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8831920623779297, "optim/lr": 0.000258, "optim/total_tokens": 112984064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481188, "loss/out": 5.712657928466797, "created_at": "2025-01-14T13:35:20.984365+00:00"} {"global_step": 432, "acc_step": 0, "speed/wps": 12852.520956140736, "speed/FLOPS": 201866502830747.0, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6672476530075073, "optim/lr": 0.0002586, "optim/total_tokens": 113246208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483281, "loss/out": 5.6723456382751465, "created_at": "2025-01-14T13:35:26.087833+00:00"} {"global_step": 433, "acc_step": 0, "speed/wps": 12851.753157292, "speed/FLOPS": 201854443494761.44, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6735336184501648, "optim/lr": 0.0002592, "optim/total_tokens": 113508352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477481, "loss/out": 5.812740325927734, "created_at": "2025-01-14T13:35:31.193880+00:00"} {"global_step": 434, "acc_step": 0, "speed/wps": 12847.63858888157, "speed/FLOPS": 201789818543866.12, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5325084924697876, "optim/lr": 0.00025979999999999997, "optim/total_tokens": 113770496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471114, "loss/out": 5.806347846984863, "created_at": "2025-01-14T13:35:36.299499+00:00"} {"global_step": 435, "acc_step": 0, "speed/wps": 12848.08499233552, "speed/FLOPS": 201796829923533.06, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4789952337741852, "optim/lr": 0.0002604, "optim/total_tokens": 114032640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313467, "loss/out": 5.710825443267822, "created_at": "2025-01-14T13:35:41.406430+00:00"} {"global_step": 436, "acc_step": 0, "speed/wps": 12851.82941267449, "speed/FLOPS": 201855641190329.3, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5099604725837708, "optim/lr": 0.000261, "optim/total_tokens": 114294784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495212, "loss/out": 5.754470348358154, "created_at": "2025-01-14T13:35:46.510205+00:00"} {"global_step": 437, "acc_step": 0, "speed/wps": 12849.293514703675, "speed/FLOPS": 201815811428008.7, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.49337631464004517, "optim/lr": 0.0002616, "optim/total_tokens": 114556928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433481, "loss/out": 5.737088203430176, "created_at": "2025-01-14T13:35:51.617404+00:00"} {"global_step": 438, "acc_step": 0, "speed/wps": 12852.08396451911, "speed/FLOPS": 201859639276839.97, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.49362871050834656, "optim/lr": 0.00026220000000000003, "optim/total_tokens": 114819072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492025, "loss/out": 5.703065872192383, "created_at": "2025-01-14T13:35:56.720694+00:00"} {"global_step": 439, "acc_step": 0, "speed/wps": 12848.357696461555, "speed/FLOPS": 201801113116567.16, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.46780332922935486, "optim/lr": 0.0002628, "optim/total_tokens": 115081216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498919, "loss/out": 5.7575459480285645, "created_at": "2025-01-14T13:36:01.824916+00:00"} {"global_step": 440, "acc_step": 0, "speed/wps": 12858.509253875383, "speed/FLOPS": 201960557275453.06, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.001, "optim/grad_norm": 0.5361219644546509, "optim/lr": 0.0002634, "optim/total_tokens": 115343360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485138, "loss/out": 5.856082916259766, "created_at": "2025-01-14T13:36:06.925519+00:00"} {"global_step": 441, "acc_step": 0, "speed/wps": 12852.227770687934, "speed/FLOPS": 201861897950331.72, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8202723860740662, "optim/lr": 0.00026399999999999997, "optim/total_tokens": 115605504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 5.723684310913086, "created_at": "2025-01-14T13:36:12.028463+00:00"} {"global_step": 442, "acc_step": 0, "speed/wps": 12844.011610636644, "speed/FLOPS": 201732851866539.2, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8284714221954346, "optim/lr": 0.0002646, "optim/total_tokens": 115867648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494625, "loss/out": 5.6934661865234375, "created_at": "2025-01-14T13:36:17.134765+00:00"} {"global_step": 443, "acc_step": 0, "speed/wps": 12844.550075404057, "speed/FLOPS": 201741309195639.53, "speed/curr_iter_time": 1.2874, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6953871846199036, "optim/lr": 0.00026520000000000005, "optim/total_tokens": 116129792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350520, "loss/out": 5.669959545135498, "created_at": "2025-01-14T13:36:22.240711+00:00"} {"global_step": 444, "acc_step": 0, "speed/wps": 12846.400656772152, "speed/FLOPS": 201770375118990.75, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7592396140098572, "optim/lr": 0.0002658, "optim/total_tokens": 116391936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493816, "loss/out": 5.644094944000244, "created_at": "2025-01-14T13:36:27.346536+00:00"} {"global_step": 445, "acc_step": 0, "speed/wps": 12850.852852155906, "speed/FLOPS": 201840302965450.16, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4814157783985138, "optim/lr": 0.0002664, "optim/total_tokens": 116654080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466881, "loss/out": 5.693185329437256, "created_at": "2025-01-14T13:36:32.451642+00:00"} {"global_step": 446, "acc_step": 0, "speed/wps": 12848.732066521432, "speed/FLOPS": 201806993112795.9, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.555625319480896, "optim/lr": 0.000267, "optim/total_tokens": 116916224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489365, "loss/out": 5.757912635803223, "created_at": "2025-01-14T13:36:37.558906+00:00"} {"global_step": 447, "acc_step": 0, "speed/wps": 12858.494364219969, "speed/FLOPS": 201960323412954.25, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7970621585845947, "optim/lr": 0.0002676, "optim/total_tokens": 117178368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502556, "loss/out": 5.67512321472168, "created_at": "2025-01-14T13:36:42.660360+00:00"} {"global_step": 448, "acc_step": 0, "speed/wps": 12852.62305470985, "speed/FLOPS": 201868106429073.06, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5191406011581421, "optim/lr": 0.00026819999999999996, "optim/total_tokens": 117440512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480385, "loss/out": 5.704683303833008, "created_at": "2025-01-14T13:36:47.764920+00:00"} {"global_step": 449, "acc_step": 0, "speed/wps": 12846.108840952766, "speed/FLOPS": 201765791750550.3, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5331554412841797, "optim/lr": 0.00026880000000000003, "optim/total_tokens": 117702656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488845, "loss/out": 5.657786846160889, "created_at": "2025-01-14T13:36:52.870183+00:00"} {"global_step": 450, "acc_step": 0, "speed/wps": 12844.760400664998, "speed/FLOPS": 201744612642879.8, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.40636003017425537, "optim/lr": 0.00026940000000000004, "optim/total_tokens": 117964800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321540, "loss/out": 5.668701171875, "created_at": "2025-01-14T13:36:57.977705+00:00"} {"global_step": 451, "acc_step": 0, "speed/wps": 12853.326574918388, "speed/FLOPS": 201879156180686.25, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4389917254447937, "optim/lr": 0.00027, "optim/total_tokens": 118226944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492268, "loss/out": 5.68864631652832, "created_at": "2025-01-14T13:37:03.081352+00:00"} {"global_step": 452, "acc_step": 0, "speed/wps": 12850.439809268977, "speed/FLOPS": 201833815559330.56, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5222911238670349, "optim/lr": 0.0002706, "optim/total_tokens": 118489088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427880, "loss/out": 5.626659870147705, "created_at": "2025-01-14T13:37:08.184831+00:00"} {"global_step": 453, "acc_step": 0, "speed/wps": 12854.902634591228, "speed/FLOPS": 201903910363581.16, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6221246123313904, "optim/lr": 0.0002712, "optim/total_tokens": 118751232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498044, "loss/out": 5.698084354400635, "created_at": "2025-01-14T13:37:13.293568+00:00"} {"global_step": 454, "acc_step": 0, "speed/wps": 12849.281727034702, "speed/FLOPS": 201815626286469.5, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5861672759056091, "optim/lr": 0.0002718, "optim/total_tokens": 119013376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466629, "loss/out": 5.732425689697266, "created_at": "2025-01-14T13:37:18.397719+00:00"} {"global_step": 455, "acc_step": 0, "speed/wps": 12849.798215684144, "speed/FLOPS": 201823738450438.78, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5802785158157349, "optim/lr": 0.0002724, "optim/total_tokens": 119275520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504121, "loss/out": 5.630902290344238, "created_at": "2025-01-14T13:37:23.501488+00:00"} {"global_step": 456, "acc_step": 0, "speed/wps": 12843.878554009201, "speed/FLOPS": 201730762029366.22, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5722829699516296, "optim/lr": 0.000273, "optim/total_tokens": 119537664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379659, "loss/out": 5.666262149810791, "created_at": "2025-01-14T13:37:28.610628+00:00"} {"global_step": 457, "acc_step": 0, "speed/wps": 12849.340686712465, "speed/FLOPS": 201816552329227.88, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6308867335319519, "optim/lr": 0.00027360000000000004, "optim/total_tokens": 119799808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495387, "loss/out": 5.695687294006348, "created_at": "2025-01-14T13:37:33.717794+00:00"} {"global_step": 458, "acc_step": 0, "speed/wps": 12850.006945914089, "speed/FLOPS": 201827016845525.44, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.5437429547309875, "optim/lr": 0.0002742, "optim/total_tokens": 120061952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471924, "loss/out": 5.661978721618652, "created_at": "2025-01-14T13:37:38.824521+00:00"} {"global_step": 459, "acc_step": 0, "speed/wps": 12843.773267817898, "speed/FLOPS": 201729108365053.44, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5906123518943787, "optim/lr": 0.0002748, "optim/total_tokens": 120324096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470830, "loss/out": 5.457839012145996, "created_at": "2025-01-14T13:37:43.930832+00:00"} {"global_step": 460, "acc_step": 0, "speed/wps": 12850.204581120011, "speed/FLOPS": 201830120978013.7, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5520456433296204, "optim/lr": 0.0002754, "optim/total_tokens": 120586240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498564, "loss/out": 5.501875877380371, "created_at": "2025-01-14T13:37:49.034478+00:00"} {"global_step": 461, "acc_step": 0, "speed/wps": 12850.652624948241, "speed/FLOPS": 201837158122012.78, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.41475632786750793, "optim/lr": 0.000276, "optim/total_tokens": 120848384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314974, "loss/out": 5.616940498352051, "created_at": "2025-01-14T13:37:54.137913+00:00"} {"global_step": 462, "acc_step": 0, "speed/wps": 12850.566541299882, "speed/FLOPS": 201835806060023.97, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5109217762947083, "optim/lr": 0.0002766, "optim/total_tokens": 121110528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489128, "loss/out": 5.556532859802246, "created_at": "2025-01-14T13:37:59.243267+00:00"} {"global_step": 463, "acc_step": 0, "speed/wps": 12846.719006828846, "speed/FLOPS": 201775375244089.94, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.46804922819137573, "optim/lr": 0.0002772, "optim/total_tokens": 121372672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 344512, "loss/out": 5.623963356018066, "created_at": "2025-01-14T13:38:04.350854+00:00"} {"global_step": 464, "acc_step": 0, "speed/wps": 12855.341032081356, "speed/FLOPS": 201910795998584.88, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0395560264587402, "optim/lr": 0.00027780000000000003, "optim/total_tokens": 121634816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496778, "loss/out": 5.60691499710083, "created_at": "2025-01-14T13:38:09.458058+00:00"} {"global_step": 465, "acc_step": 0, "speed/wps": 12859.84366631751, "speed/FLOPS": 201981516056531.5, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.4630742073059082, "optim/lr": 0.0002784, "optim/total_tokens": 121896960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474336, "loss/out": 5.64300537109375, "created_at": "2025-01-14T13:38:14.559021+00:00"} {"global_step": 466, "acc_step": 0, "speed/wps": 12846.52539306613, "speed/FLOPS": 201772334273893.28, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5614262819290161, "optim/lr": 0.000279, "optim/total_tokens": 122159104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500418, "loss/out": 5.610279083251953, "created_at": "2025-01-14T13:38:19.668729+00:00"} {"global_step": 467, "acc_step": 0, "speed/wps": 12853.507515974694, "speed/FLOPS": 201881998108614.6, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6415082216262817, "optim/lr": 0.0002796, "optim/total_tokens": 122421248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483565, "loss/out": 5.7428741455078125, "created_at": "2025-01-14T13:38:24.771258+00:00"} {"global_step": 468, "acc_step": 0, "speed/wps": 12861.844202253524, "speed/FLOPS": 202012937222431.06, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9197076559066772, "optim/lr": 0.0002802, "optim/total_tokens": 122683392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499090, "loss/out": 5.718048095703125, "created_at": "2025-01-14T13:38:29.872325+00:00"} {"global_step": 469, "acc_step": 0, "speed/wps": 12852.899360216239, "speed/FLOPS": 201872446186734.47, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9067375063896179, "optim/lr": 0.0002808, "optim/total_tokens": 122945536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360094, "loss/out": 5.721125602722168, "created_at": "2025-01-14T13:38:34.976990+00:00"} {"global_step": 470, "acc_step": 0, "speed/wps": 12844.719714626495, "speed/FLOPS": 201743973612736.66, "speed/curr_iter_time": 1.288, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9387009143829346, "optim/lr": 0.0002814, "optim/total_tokens": 123207680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485366, "loss/out": 5.7104878425598145, "created_at": "2025-01-14T13:38:40.086183+00:00"} {"global_step": 471, "acc_step": 0, "speed/wps": 12849.660724092797, "speed/FLOPS": 201821578956063.44, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0486432313919067, "optim/lr": 0.000282, "optim/total_tokens": 123469824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479054, "loss/out": 5.694897651672363, "created_at": "2025-01-14T13:38:45.190101+00:00"} {"global_step": 472, "acc_step": 0, "speed/wps": 12850.20523582328, "speed/FLOPS": 201830131261028.2, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6536104679107666, "optim/lr": 0.00028260000000000004, "optim/total_tokens": 123731968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502792, "loss/out": 5.719184875488281, "created_at": "2025-01-14T13:38:50.293725+00:00"} {"global_step": 473, "acc_step": 0, "speed/wps": 12851.222725312587, "speed/FLOPS": 201846112331634.84, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5552478432655334, "optim/lr": 0.0002832, "optim/total_tokens": 123994112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490175, "loss/out": 5.761713981628418, "created_at": "2025-01-14T13:38:55.399089+00:00"} {"global_step": 474, "acc_step": 0, "speed/wps": 12856.188495730219, "speed/FLOPS": 201924106579727.38, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5766162276268005, "optim/lr": 0.0002838, "optim/total_tokens": 124256256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479338, "loss/out": 5.726627349853516, "created_at": "2025-01-14T13:39:00.505134+00:00"} {"global_step": 475, "acc_step": 0, "speed/wps": 12849.54796844823, "speed/FLOPS": 201819807973723.03, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5214052796363831, "optim/lr": 0.0002844, "optim/total_tokens": 124518400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497235, "loss/out": 5.715301513671875, "created_at": "2025-01-14T13:39:05.612181+00:00"} {"global_step": 476, "acc_step": 0, "speed/wps": 12847.435459459777, "speed/FLOPS": 201786628117167.8, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5225076675415039, "optim/lr": 0.000285, "optim/total_tokens": 124780544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300645, "loss/out": 5.736942768096924, "created_at": "2025-01-14T13:39:10.721639+00:00"} {"global_step": 477, "acc_step": 0, "speed/wps": 12848.492661698672, "speed/FLOPS": 201803232931077.75, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4123290181159973, "optim/lr": 0.0002856, "optim/total_tokens": 125042688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497305, "loss/out": 5.511480808258057, "created_at": "2025-01-14T13:39:15.831533+00:00"} {"global_step": 478, "acc_step": 0, "speed/wps": 12850.416030278007, "speed/FLOPS": 201833442077601.62, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.48165640234947205, "optim/lr": 0.0002862, "optim/total_tokens": 125304832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464740, "loss/out": 5.560787677764893, "created_at": "2025-01-14T13:39:20.940176+00:00"} {"global_step": 479, "acc_step": 0, "speed/wps": 12845.17484648436, "speed/FLOPS": 201751122083981.12, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.49480903148651123, "optim/lr": 0.00028680000000000003, "optim/total_tokens": 125566976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476387, "loss/out": 5.614217758178711, "created_at": "2025-01-14T13:39:26.045925+00:00"} {"global_step": 480, "acc_step": 0, "speed/wps": 12852.112754132526, "speed/FLOPS": 201860091457279.72, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5242443680763245, "optim/lr": 0.0002874, "optim/total_tokens": 125829120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501747, "loss/out": 5.608316421508789, "created_at": "2025-01-14T13:39:31.150370+00:00"} {"global_step": 481, "acc_step": 0, "speed/wps": 12850.840355227241, "speed/FLOPS": 201840106684012.9, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6024574637413025, "optim/lr": 0.000288, "optim/total_tokens": 126091264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486468, "loss/out": 5.601006507873535, "created_at": "2025-01-14T13:39:36.253618+00:00"} {"global_step": 482, "acc_step": 0, "speed/wps": 12856.49815390665, "speed/FLOPS": 201928970186903.16, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.47495031356811523, "optim/lr": 0.00028859999999999997, "optim/total_tokens": 126353408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372640, "loss/out": 5.656833648681641, "created_at": "2025-01-14T13:39:41.358264+00:00"} {"global_step": 483, "acc_step": 0, "speed/wps": 12855.30100936022, "speed/FLOPS": 201910167386752.47, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5201757550239563, "optim/lr": 0.0002892, "optim/total_tokens": 126615552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483565, "loss/out": 5.736934661865234, "created_at": "2025-01-14T13:39:46.464836+00:00"} {"global_step": 484, "acc_step": 0, "speed/wps": 12852.085863479446, "speed/FLOPS": 201859669102621.47, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5049770474433899, "optim/lr": 0.0002898, "optim/total_tokens": 126877696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468454, "loss/out": 5.562664985656738, "created_at": "2025-01-14T13:39:51.571263+00:00"} {"global_step": 485, "acc_step": 0, "speed/wps": 12854.207474723411, "speed/FLOPS": 201892991922607.3, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4737946093082428, "optim/lr": 0.0002904, "optim/total_tokens": 127139840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492251, "loss/out": 5.488234519958496, "created_at": "2025-01-14T13:39:56.677438+00:00"} {"global_step": 486, "acc_step": 0, "speed/wps": 12857.097857506045, "speed/FLOPS": 201938389356010.75, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.46180248260498047, "optim/lr": 0.000291, "optim/total_tokens": 127401984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487220, "loss/out": 5.683871269226074, "created_at": "2025-01-14T13:40:01.779666+00:00"} {"global_step": 487, "acc_step": 0, "speed/wps": 12852.474061250272, "speed/FLOPS": 201865766282051.94, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.37165722250938416, "optim/lr": 0.0002916, "optim/total_tokens": 127664128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463984, "loss/out": 5.609628200531006, "created_at": "2025-01-14T13:40:06.882495+00:00"} {"global_step": 488, "acc_step": 0, "speed/wps": 12846.608068810776, "speed/FLOPS": 201773632810071.34, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.42902034521102905, "optim/lr": 0.0002922, "optim/total_tokens": 127926272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491159, "loss/out": 5.610418796539307, "created_at": "2025-01-14T13:40:11.990262+00:00"} {"global_step": 489, "acc_step": 0, "speed/wps": 12848.455513190904, "speed/FLOPS": 201802649462715.88, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3962828814983368, "optim/lr": 0.0002928, "optim/total_tokens": 128188416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289931, "loss/out": 5.444974899291992, "created_at": "2025-01-14T13:40:17.097629+00:00"} {"global_step": 490, "acc_step": 0, "speed/wps": 12856.700463640862, "speed/FLOPS": 201932147739281.66, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4419463276863098, "optim/lr": 0.0002934, "optim/total_tokens": 128450560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470587, "loss/out": 5.604820251464844, "created_at": "2025-01-14T13:40:22.199665+00:00"} {"global_step": 491, "acc_step": 0, "speed/wps": 12852.330423276178, "speed/FLOPS": 201863510250289.53, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.417305052280426, "optim/lr": 0.000294, "optim/total_tokens": 128712704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351463, "loss/out": 5.660537242889404, "created_at": "2025-01-14T13:40:27.304197+00:00"} {"global_step": 492, "acc_step": 0, "speed/wps": 12849.964692008189, "speed/FLOPS": 201826353189870.34, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5150722861289978, "optim/lr": 0.0002946, "optim/total_tokens": 128974848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499965, "loss/out": 5.573485851287842, "created_at": "2025-01-14T13:40:32.408542+00:00"} {"global_step": 493, "acc_step": 0, "speed/wps": 12851.924352923113, "speed/FLOPS": 201857132357395.53, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.46990057826042175, "optim/lr": 0.0002952, "optim/total_tokens": 129236992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 403566, "loss/out": 5.515666961669922, "created_at": "2025-01-14T13:40:37.511417+00:00"} {"global_step": 494, "acc_step": 0, "speed/wps": 12853.804593973038, "speed/FLOPS": 201886664126806.53, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.627427875995636, "optim/lr": 0.0002958, "optim/total_tokens": 129499136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467927, "loss/out": 5.612489700317383, "created_at": "2025-01-14T13:40:42.613608+00:00"} {"global_step": 495, "acc_step": 0, "speed/wps": 12854.008406188917, "speed/FLOPS": 201889865277724.25, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5986887812614441, "optim/lr": 0.0002964, "optim/total_tokens": 129761280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313679, "loss/out": 5.506321907043457, "created_at": "2025-01-14T13:40:47.719945+00:00"} {"global_step": 496, "acc_step": 0, "speed/wps": 12854.552203942734, "speed/FLOPS": 201898406368704.38, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6345177888870239, "optim/lr": 0.000297, "optim/total_tokens": 130023424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486185, "loss/out": 5.527097702026367, "created_at": "2025-01-14T13:40:52.822137+00:00"} {"global_step": 497, "acc_step": 0, "speed/wps": 12847.581305542, "speed/FLOPS": 201788918830302.62, "speed/curr_iter_time": 1.2874, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7809946537017822, "optim/lr": 0.00029759999999999997, "optim/total_tokens": 130285568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293250, "loss/out": 5.6138410568237305, "created_at": "2025-01-14T13:40:57.928770+00:00"} {"global_step": 498, "acc_step": 0, "speed/wps": 12858.405113946914, "speed/FLOPS": 201958921614773.2, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4914883077144623, "optim/lr": 0.00029820000000000004, "optim/total_tokens": 130547712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478574, "loss/out": 5.5513386726379395, "created_at": "2025-01-14T13:41:03.031092+00:00"} {"global_step": 499, "acc_step": 0, "speed/wps": 12849.581235322423, "speed/FLOPS": 201820330475695.7, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4533250033855438, "optim/lr": 0.0002988, "optim/total_tokens": 130809856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348536, "loss/out": 5.611033916473389, "created_at": "2025-01-14T13:41:08.136877+00:00"} {"global_step": 500, "acc_step": 0, "speed/wps": 12837.533531469602, "speed/FLOPS": 201631104731407.16, "speed/curr_iter_time": 1.2877, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4906136095523834, "optim/lr": 0.0002994, "optim/total_tokens": 131072000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489648, "loss/out": 5.554570198059082, "created_at": "2025-01-14T13:41:13.245465+00:00"} {"global_step": 501, "acc_step": 0, "speed/wps": 12850.890899292872, "speed/FLOPS": 201840900548018.47, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.48943135142326355, "optim/lr": 0.00030000000000000003, "optim/total_tokens": 131334144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484328, "loss/out": 5.526907920837402, "created_at": "2025-01-14T13:41:18.351043+00:00"} {"global_step": 502, "acc_step": 0, "speed/wps": 12845.095893875854, "speed/FLOPS": 201749882024772.28, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.46124520897865295, "optim/lr": 0.0003006, "optim/total_tokens": 131596288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481663, "loss/out": 5.607898235321045, "created_at": "2025-01-14T13:41:23.461314+00:00"} {"global_step": 503, "acc_step": 0, "speed/wps": 12853.458796690884, "speed/FLOPS": 201881232905315.06, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3999261260032654, "optim/lr": 0.0003012, "optim/total_tokens": 131858432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494625, "loss/out": 5.5856122970581055, "created_at": "2025-01-14T13:41:28.566046+00:00"} {"global_step": 504, "acc_step": 0, "speed/wps": 12849.313228126748, "speed/FLOPS": 201816121054407.12, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.36615869402885437, "optim/lr": 0.0003018, "optim/total_tokens": 132120576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318185, "loss/out": 5.545513153076172, "created_at": "2025-01-14T13:41:33.671097+00:00"} {"global_step": 505, "acc_step": 0, "speed/wps": 12851.647307739657, "speed/FLOPS": 201852780982089.34, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.45381230115890503, "optim/lr": 0.00030240000000000003, "optim/total_tokens": 132382720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489648, "loss/out": 5.533316612243652, "created_at": "2025-01-14T13:41:38.776581+00:00"} {"global_step": 506, "acc_step": 0, "speed/wps": 12855.992776174791, "speed/FLOPS": 201921032535162.53, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5494678616523743, "optim/lr": 0.00030300000000000005, "optim/total_tokens": 132644864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310540, "loss/out": 5.539834499359131, "created_at": "2025-01-14T13:41:43.879750+00:00"} {"global_step": 507, "acc_step": 0, "speed/wps": 12852.3208084024, "speed/FLOPS": 201863359235484.66, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4040493369102478, "optim/lr": 0.0003036, "optim/total_tokens": 132907008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468974, "loss/out": 5.600900650024414, "created_at": "2025-01-14T13:41:48.982743+00:00"} {"global_step": 508, "acc_step": 0, "speed/wps": 12854.277245617217, "speed/FLOPS": 201894087770369.78, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.5192872881889343, "optim/lr": 0.0003042, "optim/total_tokens": 133169152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324762, "loss/out": 5.590019702911377, "created_at": "2025-01-14T13:41:54.085781+00:00"} {"global_step": 509, "acc_step": 0, "speed/wps": 12851.334170203336, "speed/FLOPS": 201847862726785.38, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.44616085290908813, "optim/lr": 0.0003048, "optim/total_tokens": 133431296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505733, "loss/out": 5.57069206237793, "created_at": "2025-01-14T13:41:59.190377+00:00"} {"global_step": 510, "acc_step": 0, "speed/wps": 12856.942368463824, "speed/FLOPS": 201935947186938.3, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4853731691837311, "optim/lr": 0.0003054, "optim/total_tokens": 133693440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320345, "loss/out": 5.599785804748535, "created_at": "2025-01-14T13:42:04.292202+00:00"} {"global_step": 511, "acc_step": 0, "speed/wps": 12852.665483383176, "speed/FLOPS": 201868772829690.78, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5175105929374695, "optim/lr": 0.000306, "optim/total_tokens": 133955584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487798, "loss/out": 5.507894515991211, "created_at": "2025-01-14T13:42:09.396526+00:00"} {"global_step": 512, "acc_step": 0, "speed/wps": 12849.833149848246, "speed/FLOPS": 201824287139490.56, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7140316367149353, "optim/lr": 0.0003066, "optim/total_tokens": 134217728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350932, "loss/out": 5.568878650665283, "created_at": "2025-01-14T13:42:14.503350+00:00"} {"global_step": 513, "acc_step": 0, "speed/wps": 12853.07657055697, "speed/FLOPS": 201875229518648.2, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8351291418075562, "optim/lr": 0.00030720000000000004, "optim/total_tokens": 134479872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501747, "loss/out": 5.5628509521484375, "created_at": "2025-01-14T13:42:19.605790+00:00"} {"global_step": 514, "acc_step": 0, "speed/wps": 12851.26491708255, "speed/FLOPS": 201846775011359.66, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5123368501663208, "optim/lr": 0.0003078, "optim/total_tokens": 134742016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461605, "loss/out": 5.517441749572754, "created_at": "2025-01-14T13:42:24.708974+00:00"} {"global_step": 515, "acc_step": 0, "speed/wps": 12848.471435294201, "speed/FLOPS": 201802899541227.4, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6658300161361694, "optim/lr": 0.0003084, "optim/total_tokens": 135004160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492251, "loss/out": 5.550783634185791, "created_at": "2025-01-14T13:42:29.819642+00:00"} {"global_step": 516, "acc_step": 0, "speed/wps": 12856.612269456591, "speed/FLOPS": 201930762528426.16, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4822845757007599, "optim/lr": 0.000309, "optim/total_tokens": 135266304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484091, "loss/out": 5.533690929412842, "created_at": "2025-01-14T13:42:34.921207+00:00"} {"global_step": 517, "acc_step": 0, "speed/wps": 12848.202667409627, "speed/FLOPS": 201798678172274.4, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.512549877166748, "optim/lr": 0.0003096, "optim/total_tokens": 135528448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476952, "loss/out": 5.517203330993652, "created_at": "2025-01-14T13:42:40.028599+00:00"} {"global_step": 518, "acc_step": 0, "speed/wps": 12854.466795972452, "speed/FLOPS": 201897064919168.25, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6199058294296265, "optim/lr": 0.0003102, "optim/total_tokens": 135790592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491968, "loss/out": 5.487165451049805, "created_at": "2025-01-14T13:42:45.130501+00:00"} {"global_step": 519, "acc_step": 0, "speed/wps": 12847.680046378175, "speed/FLOPS": 201790469690818.88, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.49254703521728516, "optim/lr": 0.0003108, "optim/total_tokens": 136052736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309482, "loss/out": 5.531081199645996, "created_at": "2025-01-14T13:42:50.246136+00:00"} {"global_step": 520, "acc_step": 0, "speed/wps": 12850.654772243346, "speed/FLOPS": 201837191848233.1, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6772072315216064, "optim/lr": 0.00031140000000000003, "optim/total_tokens": 136314880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486185, "loss/out": 5.6438446044921875, "created_at": "2025-01-14T13:42:55.352646+00:00"} {"global_step": 521, "acc_step": 0, "speed/wps": 12857.106250934597, "speed/FLOPS": 201938521186339.94, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5418576598167419, "optim/lr": 0.000312, "optim/total_tokens": 136577024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314204, "loss/out": 5.592411518096924, "created_at": "2025-01-14T13:43:00.453499+00:00"} {"global_step": 522, "acc_step": 0, "speed/wps": 12849.488929438825, "speed/FLOPS": 201818880684952.2, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.46564698219299316, "optim/lr": 0.0003126, "optim/total_tokens": 136839168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482235, "loss/out": 5.521066665649414, "created_at": "2025-01-14T13:43:05.560322+00:00"} {"global_step": 523, "acc_step": 0, "speed/wps": 12854.306014532342, "speed/FLOPS": 201894539625714.47, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5863716006278992, "optim/lr": 0.0003132, "optim/total_tokens": 137101312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304559, "loss/out": 5.477057933807373, "created_at": "2025-01-14T13:43:10.662493+00:00"} {"global_step": 524, "acc_step": 0, "speed/wps": 12833.604538784146, "speed/FLOPS": 201569394502280.34, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4344337284564972, "optim/lr": 0.0003138, "optim/total_tokens": 137363456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487272, "loss/out": 5.431441307067871, "created_at": "2025-01-14T13:43:15.772650+00:00"} {"global_step": 525, "acc_step": 0, "speed/wps": 12852.017200810487, "speed/FLOPS": 201858590661053.1, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.588493824005127, "optim/lr": 0.0003144, "optim/total_tokens": 137625600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377268, "loss/out": 5.584624290466309, "created_at": "2025-01-14T13:43:20.876826+00:00"} {"global_step": 526, "acc_step": 0, "speed/wps": 12845.687419579897, "speed/FLOPS": 201759172748795.78, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.49618425965309143, "optim/lr": 0.000315, "optim/total_tokens": 137887744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493297, "loss/out": 5.525132179260254, "created_at": "2025-01-14T13:43:25.982152+00:00"} {"global_step": 527, "acc_step": 0, "speed/wps": 12851.832696408186, "speed/FLOPS": 201855692765879.34, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.47165095806121826, "optim/lr": 0.00031560000000000003, "optim/total_tokens": 138149888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478008, "loss/out": 5.443859100341797, "created_at": "2025-01-14T13:43:31.089310+00:00"} {"global_step": 528, "acc_step": 0, "speed/wps": 12852.219641835429, "speed/FLOPS": 201861770275532.9, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.40193095803260803, "optim/lr": 0.0003162, "optim/total_tokens": 138412032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465267, "loss/out": 5.4600830078125, "created_at": "2025-01-14T13:43:36.193971+00:00"} {"global_step": 529, "acc_step": 0, "speed/wps": 12849.487214260706, "speed/FLOPS": 201818853745722.94, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4153381288051605, "optim/lr": 0.0003168, "optim/total_tokens": 138674176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495671, "loss/out": 5.664031982421875, "created_at": "2025-01-14T13:43:41.298025+00:00"} {"global_step": 530, "acc_step": 0, "speed/wps": 12851.847366432688, "speed/FLOPS": 201855923178772.88, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.41897520422935486, "optim/lr": 0.0003174, "optim/total_tokens": 138936320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316603, "loss/out": 5.520536422729492, "created_at": "2025-01-14T13:43:46.404894+00:00"} {"global_step": 531, "acc_step": 0, "speed/wps": 12850.497240130882, "speed/FLOPS": 201834717589935.06, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4209398329257965, "optim/lr": 0.000318, "optim/total_tokens": 139198464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481951, "loss/out": 5.433571815490723, "created_at": "2025-01-14T13:43:51.510475+00:00"} {"global_step": 532, "acc_step": 0, "speed/wps": 12852.682803301679, "speed/FLOPS": 201869044862810.94, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.4324301779270172, "optim/lr": 0.0003186, "optim/total_tokens": 139460608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355021, "loss/out": 5.504531383514404, "created_at": "2025-01-14T13:43:56.615384+00:00"} {"global_step": 533, "acc_step": 0, "speed/wps": 12857.022425200683, "speed/FLOPS": 201937204588000.2, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5316652059555054, "optim/lr": 0.0003192, "optim/total_tokens": 139722752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478244, "loss/out": 5.539496421813965, "created_at": "2025-01-14T13:44:01.717182+00:00"} {"global_step": 534, "acc_step": 0, "speed/wps": 12848.896926891788, "speed/FLOPS": 201809582471454.25, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5367610454559326, "optim/lr": 0.0003198, "optim/total_tokens": 139984896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310540, "loss/out": 5.518862724304199, "created_at": "2025-01-14T13:44:06.824115+00:00"} {"global_step": 535, "acc_step": 0, "speed/wps": 12855.301617613535, "speed/FLOPS": 201910176940206.6, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.45990145206451416, "optim/lr": 0.00032040000000000004, "optim/total_tokens": 140247040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487982, "loss/out": 5.423036575317383, "created_at": "2025-01-14T13:44:11.926909+00:00"} {"global_step": 536, "acc_step": 0, "speed/wps": 12837.622417292794, "speed/FLOPS": 201632500805402.3, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.5837955474853516, "optim/lr": 0.000321, "optim/total_tokens": 140509184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324181, "loss/out": 5.589168071746826, "created_at": "2025-01-14T13:44:17.037149+00:00"} {"global_step": 537, "acc_step": 0, "speed/wps": 12857.110697835276, "speed/FLOPS": 201938591031025.78, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5920233726501465, "optim/lr": 0.0003216, "optim/total_tokens": 140771328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480385, "loss/out": 5.643436431884766, "created_at": "2025-01-14T13:44:22.140472+00:00"} {"global_step": 538, "acc_step": 0, "speed/wps": 12850.419613665172, "speed/FLOPS": 201833498359621.12, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5151547193527222, "optim/lr": 0.00032219999999999997, "optim/total_tokens": 141033472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460271, "loss/out": 5.452399253845215, "created_at": "2025-01-14T13:44:27.246850+00:00"} {"global_step": 539, "acc_step": 0, "speed/wps": 12857.671184640361, "speed/FLOPS": 201947394246489.22, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4555211067199707, "optim/lr": 0.0003228, "optim/total_tokens": 141295616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485082, "loss/out": 5.512012004852295, "created_at": "2025-01-14T13:44:32.351715+00:00"} {"global_step": 540, "acc_step": 0, "speed/wps": 12853.816542557033, "speed/FLOPS": 201886851795737.25, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5002056956291199, "optim/lr": 0.00032340000000000005, "optim/total_tokens": 141557760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477198, "loss/out": 5.603017807006836, "created_at": "2025-01-14T13:44:37.456000+00:00"} {"global_step": 541, "acc_step": 0, "speed/wps": 12852.193253064293, "speed/FLOPS": 201861355803620.28, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3978349268436432, "optim/lr": 0.000324, "optim/total_tokens": 141819904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478008, "loss/out": 5.523412704467773, "created_at": "2025-01-14T13:44:42.564053+00:00"} {"global_step": 542, "acc_step": 0, "speed/wps": 12853.51170917016, "speed/FLOPS": 201882063968511.44, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.35724034905433655, "optim/lr": 0.00032460000000000003, "optim/total_tokens": 142082048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497808, "loss/out": 5.396472930908203, "created_at": "2025-01-14T13:44:47.672661+00:00"} {"global_step": 543, "acc_step": 0, "speed/wps": 12849.983799740514, "speed/FLOPS": 201826653303063.28, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3908410966396332, "optim/lr": 0.0003252, "optim/total_tokens": 142344192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 286472, "loss/out": 5.449796676635742, "created_at": "2025-01-14T13:44:52.776753+00:00"} {"global_step": 544, "acc_step": 0, "speed/wps": 12857.324368812446, "speed/FLOPS": 201941947027336.9, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3547780215740204, "optim/lr": 0.0003258, "optim/total_tokens": 142606336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470830, "loss/out": 5.441495895385742, "created_at": "2025-01-14T13:44:57.877316+00:00"} {"global_step": 545, "acc_step": 0, "speed/wps": 12858.725170677493, "speed/FLOPS": 201963948545530.97, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3586304783821106, "optim/lr": 0.00032639999999999996, "optim/total_tokens": 142868480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311884, "loss/out": 5.511004447937012, "created_at": "2025-01-14T13:45:02.980923+00:00"} {"global_step": 546, "acc_step": 0, "speed/wps": 12854.91167798844, "speed/FLOPS": 201904052402561.3, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.37475183606147766, "optim/lr": 0.000327, "optim/total_tokens": 143130624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467117, "loss/out": 5.3446197509765625, "created_at": "2025-01-14T13:45:08.083856+00:00"} {"global_step": 547, "acc_step": 0, "speed/wps": 12851.76467252907, "speed/FLOPS": 201854624357383.22, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3477819859981537, "optim/lr": 0.00032760000000000005, "optim/total_tokens": 143392768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313070, "loss/out": 5.460488319396973, "created_at": "2025-01-14T13:45:13.186806+00:00"} {"global_step": 548, "acc_step": 0, "speed/wps": 12854.127124312748, "speed/FLOPS": 201891729908994.8, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.33806267380714417, "optim/lr": 0.0003282, "optim/total_tokens": 143654912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489311, "loss/out": 5.4934401512146, "created_at": "2025-01-14T13:45:18.288916+00:00"} {"global_step": 549, "acc_step": 0, "speed/wps": 12854.85796964801, "speed/FLOPS": 201903208839271.84, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.33232802152633667, "optim/lr": 0.0003288, "optim/total_tokens": 143917056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316688, "loss/out": 5.582815170288086, "created_at": "2025-01-14T13:45:23.391743+00:00"} {"global_step": 550, "acc_step": 0, "speed/wps": 12857.386132126638, "speed/FLOPS": 201942917105058.0, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.42116764187812805, "optim/lr": 0.0003294, "optim/total_tokens": 144179200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495739, "loss/out": 5.380782127380371, "created_at": "2025-01-14T13:45:28.492418+00:00"} {"global_step": 551, "acc_step": 0, "speed/wps": 12851.405221906805, "speed/FLOPS": 201848978691423.62, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3990623354911804, "optim/lr": 0.00033, "optim/total_tokens": 144441344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296892, "loss/out": 5.470179557800293, "created_at": "2025-01-14T13:45:33.598000+00:00"} {"global_step": 552, "acc_step": 0, "speed/wps": 12857.54042201997, "speed/FLOPS": 201945340439849.3, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4290459454059601, "optim/lr": 0.0003306, "optim/total_tokens": 144703488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487515, "loss/out": 5.44254207611084, "created_at": "2025-01-14T13:45:38.701788+00:00"} {"global_step": 553, "acc_step": 0, "speed/wps": 12854.971245902761, "speed/FLOPS": 201904987998510.6, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6794743537902832, "optim/lr": 0.00033120000000000003, "optim/total_tokens": 144965632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355837, "loss/out": 5.426507949829102, "created_at": "2025-01-14T13:45:43.813305+00:00"} {"global_step": 554, "acc_step": 0, "speed/wps": 12855.704996176108, "speed/FLOPS": 201916512554832.03, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7703274488449097, "optim/lr": 0.00033180000000000004, "optim/total_tokens": 145227776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497808, "loss/out": 5.489583969116211, "created_at": "2025-01-14T13:45:48.914649+00:00"} {"global_step": 555, "acc_step": 0, "speed/wps": 12854.325945276441, "speed/FLOPS": 201894852665438.25, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9290705919265747, "optim/lr": 0.0003324, "optim/total_tokens": 145489920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483565, "loss/out": 5.499364852905273, "created_at": "2025-01-14T13:45:54.017221+00:00"} {"global_step": 556, "acc_step": 0, "speed/wps": 12859.910664124942, "speed/FLOPS": 201982568349161.84, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6790490746498108, "optim/lr": 0.000333, "optim/total_tokens": 145752064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486694, "loss/out": 5.522308349609375, "created_at": "2025-01-14T13:45:59.118850+00:00"} {"global_step": 557, "acc_step": 0, "speed/wps": 12850.464193796883, "speed/FLOPS": 201834198551849.16, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7325952053070068, "optim/lr": 0.0003336, "optim/total_tokens": 146014208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483281, "loss/out": 5.564845085144043, "created_at": "2025-01-14T13:46:04.227399+00:00"} {"global_step": 558, "acc_step": 0, "speed/wps": 12853.637642482203, "speed/FLOPS": 201884041924227.6, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.6195397973060608, "optim/lr": 0.0003342, "optim/total_tokens": 146276352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468454, "loss/out": 5.541513919830322, "created_at": "2025-01-14T13:46:09.331867+00:00"} {"global_step": 559, "acc_step": 0, "speed/wps": 12856.707940447379, "speed/FLOPS": 201932265172802.2, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5084837079048157, "optim/lr": 0.0003348, "optim/total_tokens": 146538496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505733, "loss/out": 5.416717529296875, "created_at": "2025-01-14T13:46:14.433521+00:00"} {"global_step": 560, "acc_step": 0, "speed/wps": 12851.260570451404, "speed/FLOPS": 201846706741544.62, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4889162480831146, "optim/lr": 0.0003354, "optim/total_tokens": 146800640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478764, "loss/out": 5.4523749351501465, "created_at": "2025-01-14T13:46:19.536853+00:00"} {"global_step": 561, "acc_step": 0, "speed/wps": 12865.295051917994, "speed/FLOPS": 202067137558375.9, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.49738144874572754, "optim/lr": 0.00033600000000000004, "optim/total_tokens": 147062784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470830, "loss/out": 5.440777778625488, "created_at": "2025-01-14T13:46:24.634754+00:00"} {"global_step": 562, "acc_step": 0, "speed/wps": 12854.801434419693, "speed/FLOPS": 201902320875823.56, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5397030711174011, "optim/lr": 0.0003366, "optim/total_tokens": 147324928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496190, "loss/out": 5.541655540466309, "created_at": "2025-01-14T13:46:29.739628+00:00"} {"global_step": 563, "acc_step": 0, "speed/wps": 12858.992052662112, "speed/FLOPS": 201968140293833.7, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.4520018994808197, "optim/lr": 0.0003372, "optim/total_tokens": 147587072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469784, "loss/out": 5.433353424072266, "created_at": "2025-01-14T13:46:34.840615+00:00"} {"global_step": 564, "acc_step": 0, "speed/wps": 12854.11000008419, "speed/FLOPS": 201891460949454.16, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.32934409379959106, "optim/lr": 0.0003378, "optim/total_tokens": 147849216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501464, "loss/out": 5.456298828125, "created_at": "2025-01-14T13:46:39.945334+00:00"} {"global_step": 565, "acc_step": 0, "speed/wps": 12855.976046256568, "speed/FLOPS": 201920769768807.5, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.394039511680603, "optim/lr": 0.0003384, "optim/total_tokens": 148111360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481188, "loss/out": 5.4047112464904785, "created_at": "2025-01-14T13:46:45.047358+00:00"} {"global_step": 566, "acc_step": 0, "speed/wps": 12848.722879322553, "speed/FLOPS": 201806848815212.97, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3937076926231384, "optim/lr": 0.000339, "optim/total_tokens": 148373504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478764, "loss/out": 5.427584648132324, "created_at": "2025-01-14T13:46:50.152262+00:00"} {"global_step": 567, "acc_step": 0, "speed/wps": 12850.85247666098, "speed/FLOPS": 201840297067786.4, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3900976777076721, "optim/lr": 0.0003396, "optim/total_tokens": 148635648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485082, "loss/out": 5.351556777954102, "created_at": "2025-01-14T13:46:55.255755+00:00"} {"global_step": 568, "acc_step": 0, "speed/wps": 12852.850415480398, "speed/FLOPS": 201871677442400.97, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.38691291213035583, "optim/lr": 0.00034020000000000003, "optim/total_tokens": 148897792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327689, "loss/out": 5.381104469299316, "created_at": "2025-01-14T13:47:00.366802+00:00"} {"global_step": 569, "acc_step": 0, "speed/wps": 12855.154962125209, "speed/FLOPS": 201907873514236.7, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.38685813546180725, "optim/lr": 0.00034080000000000004, "optim/total_tokens": 149159936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484895, "loss/out": 5.427928447723389, "created_at": "2025-01-14T13:47:05.471627+00:00"} {"global_step": 570, "acc_step": 0, "speed/wps": 12851.37913924755, "speed/FLOPS": 201848569027418.7, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4349054992198944, "optim/lr": 0.0003414, "optim/total_tokens": 149422080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 412655, "loss/out": 5.456292152404785, "created_at": "2025-01-14T13:47:10.580880+00:00"} {"global_step": 571, "acc_step": 0, "speed/wps": 12850.992971965654, "speed/FLOPS": 201842503739607.06, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4587196111679077, "optim/lr": 0.000342, "optim/total_tokens": 149684224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486411, "loss/out": 5.407688617706299, "created_at": "2025-01-14T13:47:15.685473+00:00"} {"global_step": 572, "acc_step": 0, "speed/wps": 12854.774858924686, "speed/FLOPS": 201901903471155.94, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5037210583686829, "optim/lr": 0.0003426, "optim/total_tokens": 149946368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499373, "loss/out": 5.36794376373291, "created_at": "2025-01-14T13:47:20.787674+00:00"} {"global_step": 573, "acc_step": 0, "speed/wps": 12854.976529687488, "speed/FLOPS": 201905070987610.4, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.42810261249542236, "optim/lr": 0.0003432, "optim/total_tokens": 150208512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480385, "loss/out": 5.3203229904174805, "created_at": "2025-01-14T13:47:25.889458+00:00"} {"global_step": 574, "acc_step": 0, "speed/wps": 12857.686297926039, "speed/FLOPS": 201947631621408.62, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4133589267730713, "optim/lr": 0.0003438, "optim/total_tokens": 150470656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427085, "loss/out": 5.423264503479004, "created_at": "2025-01-14T13:47:30.992512+00:00"} {"global_step": 575, "acc_step": 0, "speed/wps": 12853.055054329285, "speed/FLOPS": 201874891576726.0, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4474192261695862, "optim/lr": 0.0003444, "optim/total_tokens": 150732800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498328, "loss/out": 5.387203693389893, "created_at": "2025-01-14T13:47:36.100716+00:00"} {"global_step": 576, "acc_step": 0, "speed/wps": 12854.959576973124, "speed/FLOPS": 201904804721935.84, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.575516939163208, "optim/lr": 0.00034500000000000004, "optim/total_tokens": 150994944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480954, "loss/out": 5.600126266479492, "created_at": "2025-01-14T13:47:41.208503+00:00"} {"global_step": 577, "acc_step": 0, "speed/wps": 12847.426579041987, "speed/FLOPS": 201786488638003.44, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5419042706489563, "optim/lr": 0.0003456, "optim/total_tokens": 151257088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484611, "loss/out": 5.3954572677612305, "created_at": "2025-01-14T13:47:46.315832+00:00"} {"global_step": 578, "acc_step": 0, "speed/wps": 12853.861301225052, "speed/FLOPS": 201887554792123.34, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4611935317516327, "optim/lr": 0.0003462, "optim/total_tokens": 151519232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486937, "loss/out": 5.551688194274902, "created_at": "2025-01-14T13:47:51.422602+00:00"} {"global_step": 579, "acc_step": 0, "speed/wps": 12851.724925566945, "speed/FLOPS": 201854000076725.0, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4382973313331604, "optim/lr": 0.00034679999999999997, "optim/total_tokens": 151781376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466881, "loss/out": 5.436631202697754, "created_at": "2025-01-14T13:47:56.525675+00:00"} {"global_step": 580, "acc_step": 0, "speed/wps": 12850.386717917494, "speed/FLOPS": 201832981686700.34, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.439729243516922, "optim/lr": 0.0003474, "optim/total_tokens": 152043520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490403, "loss/out": 5.489710807800293, "created_at": "2025-01-14T13:48:01.629153+00:00"} {"global_step": 581, "acc_step": 0, "speed/wps": 12855.706511033539, "speed/FLOPS": 201916536347749.62, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5249972343444824, "optim/lr": 0.000348, "optim/total_tokens": 152305664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305449, "loss/out": 5.449851036071777, "created_at": "2025-01-14T13:48:06.732790+00:00"} {"global_step": 582, "acc_step": 0, "speed/wps": 12849.638761751929, "speed/FLOPS": 201821234007320.66, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4395098090171814, "optim/lr": 0.0003486, "optim/total_tokens": 152567808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476914, "loss/out": 5.500593662261963, "created_at": "2025-01-14T13:48:11.837054+00:00"} {"global_step": 583, "acc_step": 0, "speed/wps": 12847.728750013046, "speed/FLOPS": 201791234648330.34, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5084682703018188, "optim/lr": 0.00034920000000000003, "optim/total_tokens": 152829952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362773, "loss/out": 5.355714321136475, "created_at": "2025-01-14T13:48:16.942801+00:00"} {"global_step": 584, "acc_step": 0, "speed/wps": 12857.731891726748, "speed/FLOPS": 201948347734701.2, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4536938965320587, "optim/lr": 0.0003498, "optim/total_tokens": 153092096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498328, "loss/out": 5.314857482910156, "created_at": "2025-01-14T13:48:22.051441+00:00"} {"global_step": 585, "acc_step": 0, "speed/wps": 12856.02185810454, "speed/FLOPS": 201921489306831.97, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3781108558177948, "optim/lr": 0.0003504, "optim/total_tokens": 153354240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493072, "loss/out": 5.36908483505249, "created_at": "2025-01-14T13:48:27.160680+00:00"} {"global_step": 586, "acc_step": 0, "speed/wps": 12856.582426149082, "speed/FLOPS": 201930293798272.72, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3503402769565582, "optim/lr": 0.000351, "optim/total_tokens": 153616384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494625, "loss/out": 5.4727582931518555, "created_at": "2025-01-14T13:48:32.263392+00:00"} {"global_step": 587, "acc_step": 0, "speed/wps": 12853.982160342863, "speed/FLOPS": 201889453050646.4, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.36899691820144653, "optim/lr": 0.0003516, "optim/total_tokens": 153878528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424393, "loss/out": 5.326079368591309, "created_at": "2025-01-14T13:48:37.365435+00:00"} {"global_step": 588, "acc_step": 0, "speed/wps": 12852.799956944453, "speed/FLOPS": 201870884921756.66, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.42031213641166687, "optim/lr": 0.0003522, "optim/total_tokens": 154140672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484895, "loss/out": 5.4838666915893555, "created_at": "2025-01-14T13:48:42.469823+00:00"} {"global_step": 589, "acc_step": 0, "speed/wps": 12852.633886037582, "speed/FLOPS": 201868276549958.9, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.43843862414360046, "optim/lr": 0.0003528, "optim/total_tokens": 154402816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 394289, "loss/out": 5.347646236419678, "created_at": "2025-01-14T13:48:47.575854+00:00"} {"global_step": 590, "acc_step": 0, "speed/wps": 12845.979774378295, "speed/FLOPS": 201763764582642.2, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.44693616032600403, "optim/lr": 0.0003534, "optim/total_tokens": 154664960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498854, "loss/out": 5.31386137008667, "created_at": "2025-01-14T13:48:52.683576+00:00"} {"global_step": 591, "acc_step": 0, "speed/wps": 12857.839988879328, "speed/FLOPS": 201950045549023.53, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.48966270685195923, "optim/lr": 0.000354, "optim/total_tokens": 154927104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477198, "loss/out": 5.334724426269531, "created_at": "2025-01-14T13:48:57.784213+00:00"} {"global_step": 592, "acc_step": 0, "speed/wps": 12846.959860701067, "speed/FLOPS": 201779158185121.06, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.565738320350647, "optim/lr": 0.0003546, "optim/total_tokens": 155189248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478770, "loss/out": 5.3166961669921875, "created_at": "2025-01-14T13:49:02.889127+00:00"} {"global_step": 593, "acc_step": 0, "speed/wps": 12845.337139442221, "speed/FLOPS": 201753671117898.22, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5267266035079956, "optim/lr": 0.0003552, "optim/total_tokens": 155451392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496190, "loss/out": 5.364927768707275, "created_at": "2025-01-14T13:49:07.997713+00:00"} {"global_step": 594, "acc_step": 0, "speed/wps": 12852.01283254504, "speed/FLOPS": 201858522051441.62, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.389138400554657, "optim/lr": 0.0003558, "optim/total_tokens": 155713536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321402, "loss/out": 5.411325454711914, "created_at": "2025-01-14T13:49:13.104017+00:00"} {"global_step": 595, "acc_step": 0, "speed/wps": 12851.787444874795, "speed/FLOPS": 201854982028365.3, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5236979722976685, "optim/lr": 0.0003564, "optim/total_tokens": 155975680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486468, "loss/out": 5.4182353019714355, "created_at": "2025-01-14T13:49:18.236752+00:00"} {"global_step": 596, "acc_step": 0, "speed/wps": 12846.0466536017, "speed/FLOPS": 201764815012747.47, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.4193706214427948, "optim/lr": 0.000357, "optim/total_tokens": 156237824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490639, "loss/out": 5.453825950622559, "created_at": "2025-01-14T13:49:23.342095+00:00"} {"global_step": 597, "acc_step": 0, "speed/wps": 12856.870374630065, "speed/FLOPS": 201934816424849.25, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.5148798823356628, "optim/lr": 0.0003576, "optim/total_tokens": 156499968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478528, "loss/out": 5.351117134094238, "created_at": "2025-01-14T13:49:28.445982+00:00"} {"global_step": 598, "acc_step": 0, "speed/wps": 12853.032223495431, "speed/FLOPS": 201874532987107.47, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.5042227506637573, "optim/lr": 0.00035820000000000003, "optim/total_tokens": 156762112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478197, "loss/out": 5.441028118133545, "created_at": "2025-01-14T13:49:33.548710+00:00"} {"global_step": 599, "acc_step": 0, "speed/wps": 12851.605873925137, "speed/FLOPS": 201852130207096.28, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8763123154640198, "optim/lr": 0.0003588, "optim/total_tokens": 157024256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297893, "loss/out": 5.486917972564697, "created_at": "2025-01-14T13:49:38.652096+00:00"} {"global_step": 600, "acc_step": 0, "speed/wps": 12851.873321895684, "speed/FLOPS": 201856330844985.75, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6656762361526489, "optim/lr": 0.0003594, "optim/total_tokens": 157286400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498044, "loss/out": 5.467738628387451, "created_at": "2025-01-14T13:49:43.759791+00:00"} {"global_step": 601, "acc_step": 0, "speed/wps": 12853.82927302408, "speed/FLOPS": 201887051745215.72, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5068473815917969, "optim/lr": 0.00035999999999999997, "optim/total_tokens": 157548544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479289, "loss/out": 5.425412654876709, "created_at": "2025-01-14T13:49:48.862982+00:00"} {"global_step": 602, "acc_step": 0, "speed/wps": 12861.144702066167, "speed/FLOPS": 202001950610774.62, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.48959285020828247, "optim/lr": 0.00036060000000000004, "optim/total_tokens": 157810688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501532, "loss/out": 5.457557201385498, "created_at": "2025-01-14T13:49:53.962226+00:00"} {"global_step": 603, "acc_step": 0, "speed/wps": 12847.141802327711, "speed/FLOPS": 201782015828381.53, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5979989171028137, "optim/lr": 0.0003612, "optim/total_tokens": 158072832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481474, "loss/out": 5.358076095581055, "created_at": "2025-01-14T13:49:59.069113+00:00"} {"global_step": 604, "acc_step": 0, "speed/wps": 12857.193437059179, "speed/FLOPS": 201939890564232.78, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4435109496116638, "optim/lr": 0.0003618, "optim/total_tokens": 158334976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497525, "loss/out": 5.356127738952637, "created_at": "2025-01-14T13:50:04.171584+00:00"} {"global_step": 605, "acc_step": 0, "speed/wps": 12856.662276392408, "speed/FLOPS": 201931547956069.9, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.4105420708656311, "optim/lr": 0.0003624, "optim/total_tokens": 158597120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475867, "loss/out": 5.260849952697754, "created_at": "2025-01-14T13:50:09.272679+00:00"} {"global_step": 606, "acc_step": 0, "speed/wps": 12842.856404775319, "speed/FLOPS": 201714707771067.88, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3482518494129181, "optim/lr": 0.000363, "optim/total_tokens": 158859264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471924, "loss/out": 5.3885626792907715, "created_at": "2025-01-14T13:50:14.383773+00:00"} {"global_step": 607, "acc_step": 0, "speed/wps": 12849.448792983054, "speed/FLOPS": 201818250286761.03, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3188956379890442, "optim/lr": 0.0003636, "optim/total_tokens": 159121408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367037, "loss/out": 5.299567699432373, "created_at": "2025-01-14T13:50:19.487859+00:00"} {"global_step": 608, "acc_step": 0, "speed/wps": 12851.83069894024, "speed/FLOPS": 201855661392900.44, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.3353675603866577, "optim/lr": 0.0003642, "optim/total_tokens": 159383552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492025, "loss/out": 5.326774597167969, "created_at": "2025-01-14T13:50:24.593574+00:00"} {"global_step": 609, "acc_step": 0, "speed/wps": 12853.820254629014, "speed/FLOPS": 201886910098928.66, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4082477390766144, "optim/lr": 0.00036480000000000003, "optim/total_tokens": 159645696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463439, "loss/out": 5.294299125671387, "created_at": "2025-01-14T13:50:29.696979+00:00"} {"global_step": 610, "acc_step": 0, "speed/wps": 12856.934811516423, "speed/FLOPS": 201935828494695.06, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4702470004558563, "optim/lr": 0.00036540000000000005, "optim/total_tokens": 159907840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498044, "loss/out": 5.36639404296875, "created_at": "2025-01-14T13:50:34.798496+00:00"} {"global_step": 611, "acc_step": 0, "speed/wps": 12854.726334331546, "speed/FLOPS": 201901141325740.9, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4325975179672241, "optim/lr": 0.000366, "optim/total_tokens": 160169984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470830, "loss/out": 5.26502799987793, "created_at": "2025-01-14T13:50:39.902956+00:00"} {"global_step": 612, "acc_step": 0, "speed/wps": 12850.75911287592, "speed/FLOPS": 201838830661246.72, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3984104096889496, "optim/lr": 0.0003666, "optim/total_tokens": 160432128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486937, "loss/out": 5.334287643432617, "created_at": "2025-01-14T13:50:45.007440+00:00"} {"global_step": 613, "acc_step": 0, "speed/wps": 12855.037892358914, "speed/FLOPS": 201906034772686.3, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.3714909553527832, "optim/lr": 0.0003672, "optim/total_tokens": 160694272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471114, "loss/out": 5.270116329193115, "created_at": "2025-01-14T13:50:50.111353+00:00"} {"global_step": 614, "acc_step": 0, "speed/wps": 12855.267788013696, "speed/FLOPS": 201909645599854.4, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.38099274039268494, "optim/lr": 0.0003678, "optim/total_tokens": 160956416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470067, "loss/out": 5.34397029876709, "created_at": "2025-01-14T13:50:55.215823+00:00"} {"global_step": 615, "acc_step": 0, "speed/wps": 12845.761612105645, "speed/FLOPS": 201760338044363.88, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4364349842071533, "optim/lr": 0.0003684, "optim/total_tokens": 161218560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 384608, "loss/out": 5.362753868103027, "created_at": "2025-01-14T13:51:00.324217+00:00"} {"global_step": 616, "acc_step": 0, "speed/wps": 12853.75491449248, "speed/FLOPS": 201885883842297.62, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.001, "optim/grad_norm": 0.5474957227706909, "optim/lr": 0.000369, "optim/total_tokens": 161480704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496716, "loss/out": 5.442610740661621, "created_at": "2025-01-14T13:51:05.426782+00:00"} {"global_step": 617, "acc_step": 0, "speed/wps": 12850.545561489495, "speed/FLOPS": 201835476543272.56, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5639389157295227, "optim/lr": 0.00036960000000000004, "optim/total_tokens": 161742848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450480, "loss/out": 5.371085166931152, "created_at": "2025-01-14T13:51:10.530609+00:00"} {"global_step": 618, "acc_step": 0, "speed/wps": 12847.547699522587, "speed/FLOPS": 201788391001588.25, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4531283676624298, "optim/lr": 0.0003702, "optim/total_tokens": 162004992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476678, "loss/out": 5.435503959655762, "created_at": "2025-01-14T13:51:15.635289+00:00"} {"global_step": 619, "acc_step": 0, "speed/wps": 12851.624331690664, "speed/FLOPS": 201852420111667.53, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3555757701396942, "optim/lr": 0.0003708, "optim/total_tokens": 162267136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487220, "loss/out": 5.351337432861328, "created_at": "2025-01-14T13:51:20.740709+00:00"} {"global_step": 620, "acc_step": 0, "speed/wps": 12848.908432318765, "speed/FLOPS": 201809763179995.03, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3713454604148865, "optim/lr": 0.0003714, "optim/total_tokens": 162529280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484895, "loss/out": 5.346839904785156, "created_at": "2025-01-14T13:51:25.846466+00:00"} {"global_step": 621, "acc_step": 0, "speed/wps": 12848.046837483009, "speed/FLOPS": 201796230649144.1, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.409704327583313, "optim/lr": 0.000372, "optim/total_tokens": 162791424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488266, "loss/out": 5.256409645080566, "created_at": "2025-01-14T13:51:30.956595+00:00"} {"global_step": 622, "acc_step": 0, "speed/wps": 12848.764371970132, "speed/FLOPS": 201807500514260.06, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.4398232698440552, "optim/lr": 0.0003726, "optim/total_tokens": 163053568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314047, "loss/out": 5.3345746994018555, "created_at": "2025-01-14T13:51:36.063927+00:00"} {"global_step": 623, "acc_step": 0, "speed/wps": 12849.320419482168, "speed/FLOPS": 201816234004525.97, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.461512953042984, "optim/lr": 0.0003732, "optim/total_tokens": 163315712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492268, "loss/out": 5.2736616134643555, "created_at": "2025-01-14T13:51:41.171467+00:00"} {"global_step": 624, "acc_step": 0, "speed/wps": 12845.73515135874, "speed/FLOPS": 201759922441972.84, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4411807358264923, "optim/lr": 0.00037380000000000003, "optim/total_tokens": 163577856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482473, "loss/out": 5.3105363845825195, "created_at": "2025-01-14T13:51:46.280257+00:00"} {"global_step": 625, "acc_step": 0, "speed/wps": 12852.819277347577, "speed/FLOPS": 201871188375236.75, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5332130789756775, "optim/lr": 0.0003744, "optim/total_tokens": 163840000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488561, "loss/out": 5.3997483253479, "created_at": "2025-01-14T13:51:51.383112+00:00"} {"global_step": 626, "acc_step": 0, "speed/wps": 12854.977562961454, "speed/FLOPS": 201905087216597.9, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.5069957971572876, "optim/lr": 0.000375, "optim/total_tokens": 164102144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461368, "loss/out": 5.328115463256836, "created_at": "2025-01-14T13:51:56.487159+00:00"} {"global_step": 627, "acc_step": 0, "speed/wps": 12845.673739114829, "speed/FLOPS": 201758957878292.94, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.4929039180278778, "optim/lr": 0.0003756, "optim/total_tokens": 164364288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483045, "loss/out": 5.340815544128418, "created_at": "2025-01-14T13:52:01.595227+00:00"} {"global_step": 628, "acc_step": 0, "speed/wps": 12847.072423164214, "speed/FLOPS": 201780926133281.84, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5219955444335938, "optim/lr": 0.00037620000000000004, "optim/total_tokens": 164626432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 393766, "loss/out": 5.324161529541016, "created_at": "2025-01-14T13:52:06.703312+00:00"} {"global_step": 629, "acc_step": 0, "speed/wps": 12851.149162041414, "speed/FLOPS": 201844956919374.53, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4203993082046509, "optim/lr": 0.0003768, "optim/total_tokens": 164888576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500938, "loss/out": 5.314966201782227, "created_at": "2025-01-14T13:52:11.808899+00:00"} {"global_step": 630, "acc_step": 0, "speed/wps": 12847.946706997292, "speed/FLOPS": 201794657962272.22, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5023699402809143, "optim/lr": 0.0003774, "optim/total_tokens": 165150720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486988, "loss/out": 5.201671123504639, "created_at": "2025-01-14T13:52:16.918774+00:00"} {"global_step": 631, "acc_step": 0, "speed/wps": 12852.323740280111, "speed/FLOPS": 201863405284652.9, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.48960304260253906, "optim/lr": 0.000378, "optim/total_tokens": 165412864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469548, "loss/out": 5.372811317443848, "created_at": "2025-01-14T13:52:22.026582+00:00"} {"global_step": 632, "acc_step": 0, "speed/wps": 12859.855179606498, "speed/FLOPS": 201981696888555.97, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.48353466391563416, "optim/lr": 0.00037860000000000004, "optim/total_tokens": 165675008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498854, "loss/out": 5.296011447906494, "created_at": "2025-01-14T13:52:27.127438+00:00"} {"global_step": 633, "acc_step": 0, "speed/wps": 12829.908383107655, "speed/FLOPS": 201511341298330.9, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.47540369629859924, "optim/lr": 0.00037920000000000006, "optim/total_tokens": 165937152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362773, "loss/out": 5.345120906829834, "created_at": "2025-01-14T13:52:32.243427+00:00"} {"global_step": 634, "acc_step": 0, "speed/wps": 12853.952089750826, "speed/FLOPS": 201888980750677.16, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5130524039268494, "optim/lr": 0.00037979999999999996, "optim/total_tokens": 166199296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494342, "loss/out": 5.42067813873291, "created_at": "2025-01-14T13:52:37.348906+00:00"} {"global_step": 635, "acc_step": 0, "speed/wps": 12858.94575203318, "speed/FLOPS": 201967413078832.56, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.461721807718277, "optim/lr": 0.0003804, "optim/total_tokens": 166461440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479574, "loss/out": 5.150485992431641, "created_at": "2025-01-14T13:52:42.449126+00:00"} {"global_step": 636, "acc_step": 0, "speed/wps": 12860.49193254172, "speed/FLOPS": 201991697968393.84, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.42589128017425537, "optim/lr": 0.000381, "optim/total_tokens": 166723584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 507021, "loss/out": 5.306809425354004, "created_at": "2025-01-14T13:52:47.552744+00:00"} {"global_step": 637, "acc_step": 0, "speed/wps": 12855.148929050098, "speed/FLOPS": 201907778756501.75, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.42273545265197754, "optim/lr": 0.0003816, "optim/total_tokens": 166985728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498328, "loss/out": 5.244204044342041, "created_at": "2025-01-14T13:52:52.654551+00:00"} {"global_step": 638, "acc_step": 0, "speed/wps": 12853.101582528556, "speed/FLOPS": 201875622366032.0, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.37556910514831543, "optim/lr": 0.0003822000000000001, "optim/total_tokens": 167247872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477481, "loss/out": 5.2909135818481445, "created_at": "2025-01-14T13:52:57.756898+00:00"} {"global_step": 639, "acc_step": 0, "speed/wps": 12855.573850802513, "speed/FLOPS": 201914452736527.1, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3891848921775818, "optim/lr": 0.0003828, "optim/total_tokens": 167510016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477198, "loss/out": 5.213242053985596, "created_at": "2025-01-14T13:53:02.858387+00:00"} {"global_step": 640, "acc_step": 0, "speed/wps": 12862.628593228044, "speed/FLOPS": 202025257160552.06, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.42608460783958435, "optim/lr": 0.0003834, "optim/total_tokens": 167772160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493816, "loss/out": 5.253995895385742, "created_at": "2025-01-14T13:53:07.966319+00:00"} {"global_step": 641, "acc_step": 0, "speed/wps": 12859.817001816786, "speed/FLOPS": 201981097253906.75, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.42867860198020935, "optim/lr": 0.000384, "optim/total_tokens": 168034304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422030, "loss/out": 5.241005897521973, "created_at": "2025-01-14T13:53:13.069165+00:00"} {"global_step": 642, "acc_step": 0, "speed/wps": 12854.953875197541, "speed/FLOPS": 201904715167715.28, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3586028516292572, "optim/lr": 0.0003846, "optim/total_tokens": 168296448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487982, "loss/out": 5.104392051696777, "created_at": "2025-01-14T13:53:18.174978+00:00"} {"global_step": 643, "acc_step": 0, "speed/wps": 12853.048344755507, "speed/FLOPS": 201874786193649.9, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3781599700450897, "optim/lr": 0.0003852, "optim/total_tokens": 168558592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473774, "loss/out": 5.24458646774292, "created_at": "2025-01-14T13:53:23.283151+00:00"} {"global_step": 644, "acc_step": 0, "speed/wps": 12856.311860196709, "speed/FLOPS": 201926044188194.12, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.42724326252937317, "optim/lr": 0.0003858, "optim/total_tokens": 168820736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480904, "loss/out": 5.2422075271606445, "created_at": "2025-01-14T13:53:28.384651+00:00"} {"global_step": 645, "acc_step": 0, "speed/wps": 12857.731898774817, "speed/FLOPS": 201948347845400.8, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.33195555210113525, "optim/lr": 0.0003864, "optim/total_tokens": 169082880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493816, "loss/out": 5.266575336456299, "created_at": "2025-01-14T13:53:33.488633+00:00"} {"global_step": 646, "acc_step": 0, "speed/wps": 12858.54502490381, "speed/FLOPS": 201961119108609.03, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.43157294392585754, "optim/lr": 0.00038700000000000003, "optim/total_tokens": 169345024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482471, "loss/out": 5.186590194702148, "created_at": "2025-01-14T13:53:38.591631+00:00"} {"global_step": 647, "acc_step": 0, "speed/wps": 12853.997199213629, "speed/FLOPS": 201889689256777.34, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.4858323037624359, "optim/lr": 0.00038760000000000004, "optim/total_tokens": 169607168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466881, "loss/out": 5.274299621582031, "created_at": "2025-01-14T13:53:43.699861+00:00"} {"global_step": 648, "acc_step": 0, "speed/wps": 12854.725906948546, "speed/FLOPS": 201901134613103.6, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.001, "optim/grad_norm": 0.5427124500274658, "optim/lr": 0.00038819999999999995, "optim/total_tokens": 169869312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345567, "loss/out": 5.225938320159912, "created_at": "2025-01-14T13:53:48.804402+00:00"} {"global_step": 649, "acc_step": 0, "speed/wps": 12862.263830273398, "speed/FLOPS": 202019528057112.5, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5427679419517517, "optim/lr": 0.00038879999999999996, "optim/total_tokens": 170131456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490923, "loss/out": 5.194784164428711, "created_at": "2025-01-14T13:53:53.903086+00:00"} {"global_step": 650, "acc_step": 0, "speed/wps": 12856.144965705133, "speed/FLOPS": 201923422880866.88, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.5366753935813904, "optim/lr": 0.0003894, "optim/total_tokens": 170393600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 382288, "loss/out": 5.289037704467773, "created_at": "2025-01-14T13:53:59.004304+00:00"} {"global_step": 651, "acc_step": 0, "speed/wps": 12854.095080698886, "speed/FLOPS": 201891226620006.56, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.45479270815849304, "optim/lr": 0.00039000000000000005, "optim/total_tokens": 170655744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493297, "loss/out": 5.364711761474609, "created_at": "2025-01-14T13:54:04.108425+00:00"} {"global_step": 652, "acc_step": 0, "speed/wps": 12854.452735184796, "speed/FLOPS": 201896844075176.56, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.43312767148017883, "optim/lr": 0.00039060000000000006, "optim/total_tokens": 170917888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417722, "loss/out": 5.179538726806641, "created_at": "2025-01-14T13:54:09.214677+00:00"} {"global_step": 653, "acc_step": 0, "speed/wps": 12854.34522330521, "speed/FLOPS": 201895155453370.88, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.42022255063056946, "optim/lr": 0.00039119999999999997, "optim/total_tokens": 171180032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470830, "loss/out": 5.256683349609375, "created_at": "2025-01-14T13:54:14.319206+00:00"} {"global_step": 654, "acc_step": 0, "speed/wps": 12861.301898271311, "speed/FLOPS": 202004419593186.62, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.3893366754055023, "optim/lr": 0.0003918, "optim/total_tokens": 171442176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293964, "loss/out": 5.21720027923584, "created_at": "2025-01-14T13:54:19.428061+00:00"} {"global_step": 655, "acc_step": 0, "speed/wps": 12856.23347996269, "speed/FLOPS": 201924813118913.12, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.3836725354194641, "optim/lr": 0.0003924, "optim/total_tokens": 171704320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502556, "loss/out": 5.150598526000977, "created_at": "2025-01-14T13:54:24.529392+00:00"} {"global_step": 656, "acc_step": 0, "speed/wps": 12860.904058795544, "speed/FLOPS": 201998170977531.0, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.020939826965332, "optim/lr": 0.000393, "optim/total_tokens": 171966464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 405290, "loss/out": 5.120996475219727, "created_at": "2025-01-14T13:54:29.631662+00:00"} {"global_step": 657, "acc_step": 0, "speed/wps": 12860.227716956975, "speed/FLOPS": 201987548099565.2, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4435175359249115, "optim/lr": 0.0003936, "optim/total_tokens": 172228608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483281, "loss/out": 5.164475917816162, "created_at": "2025-01-14T13:54:34.731339+00:00"} {"global_step": 658, "acc_step": 0, "speed/wps": 12857.746709135527, "speed/FLOPS": 201948580462467.3, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5205458998680115, "optim/lr": 0.0003942, "optim/total_tokens": 172490752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458152, "loss/out": 5.194496154785156, "created_at": "2025-01-14T13:54:39.832187+00:00"} {"global_step": 659, "acc_step": 0, "speed/wps": 12855.275704664247, "speed/FLOPS": 201909769941729.9, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5640769004821777, "optim/lr": 0.0003948, "optim/total_tokens": 172752896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498854, "loss/out": 5.272034168243408, "created_at": "2025-01-14T13:54:44.933924+00:00"} {"global_step": 660, "acc_step": 0, "speed/wps": 12854.333759798064, "speed/FLOPS": 201894975403238.62, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5088822841644287, "optim/lr": 0.0003954, "optim/total_tokens": 173015040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481188, "loss/out": 5.1761016845703125, "created_at": "2025-01-14T13:54:50.038124+00:00"} {"global_step": 661, "acc_step": 0, "speed/wps": 12853.46673693931, "speed/FLOPS": 201881357617827.62, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6797100305557251, "optim/lr": 0.00039600000000000003, "optim/total_tokens": 173277184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490695, "loss/out": 5.332183837890625, "created_at": "2025-01-14T13:54:55.146853+00:00"} {"global_step": 662, "acc_step": 0, "speed/wps": 12857.357405804074, "speed/FLOPS": 201942465918688.03, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.5601035952568054, "optim/lr": 0.00039660000000000004, "optim/total_tokens": 173539328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 506495, "loss/out": 5.289274215698242, "created_at": "2025-01-14T13:55:00.247475+00:00"} {"global_step": 663, "acc_step": 0, "speed/wps": 12857.584203473585, "speed/FLOPS": 201946028087740.34, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5853129029273987, "optim/lr": 0.00039719999999999995, "optim/total_tokens": 173801472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325675, "loss/out": 5.21466588973999, "created_at": "2025-01-14T13:55:05.349889+00:00"} {"global_step": 664, "acc_step": 0, "speed/wps": 12858.492298890971, "speed/FLOPS": 201960290974124.28, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.6033051013946533, "optim/lr": 0.0003978, "optim/total_tokens": 174063616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493533, "loss/out": 5.261418342590332, "created_at": "2025-01-14T13:55:10.482916+00:00"} {"global_step": 665, "acc_step": 0, "speed/wps": 12856.037784873723, "speed/FLOPS": 201921739458627.66, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5230613946914673, "optim/lr": 0.00039840000000000003, "optim/total_tokens": 174325760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480051, "loss/out": 5.177125930786133, "created_at": "2025-01-14T13:55:15.584237+00:00"} {"global_step": 666, "acc_step": 0, "speed/wps": 12858.927397770382, "speed/FLOPS": 201967124799913.94, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6558496356010437, "optim/lr": 0.00039900000000000005, "optim/total_tokens": 174587904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492251, "loss/out": 5.230354309082031, "created_at": "2025-01-14T13:55:20.689475+00:00"} {"global_step": 667, "acc_step": 0, "speed/wps": 12859.51759663, "speed/FLOPS": 201976394684022.44, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6118533611297607, "optim/lr": 0.00039960000000000006, "optim/total_tokens": 174850048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351463, "loss/out": 5.35990571975708, "created_at": "2025-01-14T13:55:25.796224+00:00"} {"global_step": 668, "acc_step": 0, "speed/wps": 12852.774584624784, "speed/FLOPS": 201870486414611.03, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5117793083190918, "optim/lr": 0.00040019999999999997, "optim/total_tokens": 175112192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496190, "loss/out": 5.235782146453857, "created_at": "2025-01-14T13:55:30.903404+00:00"} {"global_step": 669, "acc_step": 0, "speed/wps": 12858.910866994744, "speed/FLOPS": 201966865161366.88, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5537223219871521, "optim/lr": 0.0004008, "optim/total_tokens": 175374336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479054, "loss/out": 5.162482261657715, "created_at": "2025-01-14T13:55:36.006203+00:00"} {"global_step": 670, "acc_step": 0, "speed/wps": 12853.398590111612, "speed/FLOPS": 201880287278255.06, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.49951255321502686, "optim/lr": 0.0004014, "optim/total_tokens": 175636480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495975, "loss/out": 5.308373928070068, "created_at": "2025-01-14T13:55:41.110552+00:00"} {"global_step": 671, "acc_step": 0, "speed/wps": 12860.903045723264, "speed/FLOPS": 201998155065838.72, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4484674632549286, "optim/lr": 0.000402, "optim/total_tokens": 175898624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482235, "loss/out": 5.24555778503418, "created_at": "2025-01-14T13:55:46.220310+00:00"} {"global_step": 672, "acc_step": 0, "speed/wps": 12851.460497403688, "speed/FLOPS": 201849846869058.66, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.42042434215545654, "optim/lr": 0.00040260000000000003, "optim/total_tokens": 176160768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322458, "loss/out": 5.156556129455566, "created_at": "2025-01-14T13:55:51.323606+00:00"} {"global_step": 673, "acc_step": 0, "speed/wps": 12859.917622931194, "speed/FLOPS": 201982677646776.44, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3830193877220154, "optim/lr": 0.0004032, "optim/total_tokens": 176422912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491685, "loss/out": 5.06093692779541, "created_at": "2025-01-14T13:55:56.426295+00:00"} {"global_step": 674, "acc_step": 0, "speed/wps": 12855.483452708246, "speed/FLOPS": 201913032910236.62, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.41030997037887573, "optim/lr": 0.0004038, "optim/total_tokens": 176685056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290025, "loss/out": 5.148035049438477, "created_at": "2025-01-14T13:56:01.532825+00:00"} {"global_step": 675, "acc_step": 0, "speed/wps": 12858.945827226786, "speed/FLOPS": 201967414259851.47, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3942943215370178, "optim/lr": 0.0004044, "optim/total_tokens": 176947200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491741, "loss/out": 5.155962944030762, "created_at": "2025-01-14T13:56:06.633026+00:00"} {"global_step": 676, "acc_step": 0, "speed/wps": 12860.868103054094, "speed/FLOPS": 201997606243203.16, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3413190543651581, "optim/lr": 0.00040500000000000003, "optim/total_tokens": 177209344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351701, "loss/out": 5.131813049316406, "created_at": "2025-01-14T13:56:11.734224+00:00"} {"global_step": 677, "acc_step": 0, "speed/wps": 12862.897388946472, "speed/FLOPS": 202029478966674.0, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.35713431239128113, "optim/lr": 0.0004056, "optim/total_tokens": 177471488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498564, "loss/out": 5.120794773101807, "created_at": "2025-01-14T13:56:16.832767+00:00"} {"global_step": 678, "acc_step": 0, "speed/wps": 12858.8851110266, "speed/FLOPS": 201966460628494.78, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3773255944252014, "optim/lr": 0.0004062, "optim/total_tokens": 177733632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322220, "loss/out": 5.111187934875488, "created_at": "2025-01-14T13:56:21.935441+00:00"} {"global_step": 679, "acc_step": 0, "speed/wps": 12854.009298427472, "speed/FLOPS": 201889879291556.84, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.50323086977005, "optim/lr": 0.0004068, "optim/total_tokens": 177995776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465834, "loss/out": 5.185899257659912, "created_at": "2025-01-14T13:56:27.045796+00:00"} {"global_step": 680, "acc_step": 0, "speed/wps": 12862.240564691047, "speed/FLOPS": 202019162639171.4, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.49308058619499207, "optim/lr": 0.00040740000000000004, "optim/total_tokens": 178257920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304391, "loss/out": 5.142009735107422, "created_at": "2025-01-14T13:56:32.146670+00:00"} {"global_step": 681, "acc_step": 0, "speed/wps": 12859.228192466131, "speed/FLOPS": 201971849193945.9, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5493642091751099, "optim/lr": 0.00040800000000000005, "optim/total_tokens": 178520064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495387, "loss/out": 5.060433387756348, "created_at": "2025-01-14T13:56:37.249262+00:00"} {"global_step": 682, "acc_step": 0, "speed/wps": 12856.70375456217, "speed/FLOPS": 201932199427722.97, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.45226338505744934, "optim/lr": 0.00040859999999999996, "optim/total_tokens": 178782208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303572, "loss/out": 5.125431537628174, "created_at": "2025-01-14T13:56:42.358410+00:00"} {"global_step": 683, "acc_step": 0, "speed/wps": 12859.05414687816, "speed/FLOPS": 201969115568823.1, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.41301605105400085, "optim/lr": 0.00040919999999999997, "optim/total_tokens": 179044352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482761, "loss/out": 5.203868389129639, "created_at": "2025-01-14T13:56:47.459661+00:00"} {"global_step": 684, "acc_step": 0, "speed/wps": 12862.536276130139, "speed/FLOPS": 202023807193672.44, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3980434238910675, "optim/lr": 0.0004098, "optim/total_tokens": 179306496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350408, "loss/out": 5.040219306945801, "created_at": "2025-01-14T13:56:52.564650+00:00"} {"global_step": 685, "acc_step": 0, "speed/wps": 12859.962576860054, "speed/FLOPS": 201983383710002.38, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.457266241312027, "optim/lr": 0.0004104, "optim/total_tokens": 179568640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496480, "loss/out": 5.187182426452637, "created_at": "2025-01-14T13:56:57.668729+00:00"} {"global_step": 686, "acc_step": 0, "speed/wps": 12857.19753869345, "speed/FLOPS": 201939954986035.25, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.44230523705482483, "optim/lr": 0.000411, "optim/total_tokens": 179830784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467691, "loss/out": 5.1361083984375, "created_at": "2025-01-14T13:57:02.770867+00:00"} {"global_step": 687, "acc_step": 0, "speed/wps": 12860.10345240072, "speed/FLOPS": 201985596353951.78, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.36361461877822876, "optim/lr": 0.0004116, "optim/total_tokens": 180092928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492777, "loss/out": 5.138423919677734, "created_at": "2025-01-14T13:57:07.870682+00:00"} {"global_step": 688, "acc_step": 0, "speed/wps": 12860.01496240265, "speed/FLOPS": 201984206496934.12, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.34396401047706604, "optim/lr": 0.0004122, "optim/total_tokens": 180355072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465030, "loss/out": 5.209305763244629, "created_at": "2025-01-14T13:57:12.974334+00:00"} {"global_step": 689, "acc_step": 0, "speed/wps": 12866.468756927405, "speed/FLOPS": 202085572208388.4, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3227488398551941, "optim/lr": 0.0004128, "optim/total_tokens": 180617216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475341, "loss/out": 5.238852500915527, "created_at": "2025-01-14T13:57:18.079556+00:00"} {"global_step": 690, "acc_step": 0, "speed/wps": 12860.904032939869, "speed/FLOPS": 201998170571432.06, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3685060739517212, "optim/lr": 0.0004134, "optim/total_tokens": 180879360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482761, "loss/out": 5.010303020477295, "created_at": "2025-01-14T13:57:23.185875+00:00"} {"global_step": 691, "acc_step": 0, "speed/wps": 12860.044871002077, "speed/FLOPS": 201984676252586.84, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3571675419807434, "optim/lr": 0.00041400000000000003, "optim/total_tokens": 181141504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296475, "loss/out": 5.171358108520508, "created_at": "2025-01-14T13:57:28.288685+00:00"} {"global_step": 692, "acc_step": 0, "speed/wps": 12855.851015981938, "speed/FLOPS": 201918805996534.84, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3619183599948883, "optim/lr": 0.0004146, "optim/total_tokens": 181403648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488266, "loss/out": 5.101841926574707, "created_at": "2025-01-14T13:57:33.390858+00:00"} {"global_step": 693, "acc_step": 0, "speed/wps": 12851.292496666492, "speed/FLOPS": 201847208186623.8, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3439238965511322, "optim/lr": 0.0004152, "optim/total_tokens": 181665792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487982, "loss/out": 5.020626068115234, "created_at": "2025-01-14T13:57:38.495887+00:00"} {"global_step": 694, "acc_step": 0, "speed/wps": 12852.35073521381, "speed/FLOPS": 201863829277181.56, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3763890862464905, "optim/lr": 0.0004158, "optim/total_tokens": 181927936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496258, "loss/out": 5.085421562194824, "created_at": "2025-01-14T13:57:43.598725+00:00"} {"global_step": 695, "acc_step": 0, "speed/wps": 12849.598835534804, "speed/FLOPS": 201820606911216.34, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4827967882156372, "optim/lr": 0.00041640000000000004, "optim/total_tokens": 182190080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476914, "loss/out": 5.101654529571533, "created_at": "2025-01-14T13:57:48.705479+00:00"} {"global_step": 696, "acc_step": 0, "speed/wps": 12851.771420664747, "speed/FLOPS": 201854730346126.94, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.43114009499549866, "optim/lr": 0.00041700000000000005, "optim/total_tokens": 182452224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497808, "loss/out": 5.0876054763793945, "created_at": "2025-01-14T13:57:53.814229+00:00"} {"global_step": 697, "acc_step": 0, "speed/wps": 12852.740568783442, "speed/FLOPS": 201869952149080.1, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.48000577092170715, "optim/lr": 0.00041759999999999996, "optim/total_tokens": 182714368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305203, "loss/out": 5.11417293548584, "created_at": "2025-01-14T13:57:58.919004+00:00"} {"global_step": 698, "acc_step": 0, "speed/wps": 12864.360497385176, "speed/FLOPS": 202052459095225.5, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.43022117018699646, "optim/lr": 0.0004182, "optim/total_tokens": 182976512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476632, "loss/out": 5.188321113586426, "created_at": "2025-01-14T13:58:04.018937+00:00"} {"global_step": 699, "acc_step": 0, "speed/wps": 12855.240416815568, "speed/FLOPS": 201909215697575.8, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.43317607045173645, "optim/lr": 0.0004188, "optim/total_tokens": 183238656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308138, "loss/out": 5.057827949523926, "created_at": "2025-01-14T13:58:09.130801+00:00"} {"global_step": 700, "acc_step": 0, "speed/wps": 12851.153327881806, "speed/FLOPS": 201845022349622.34, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3452761173248291, "optim/lr": 0.00041940000000000006, "optim/total_tokens": 183500800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482235, "loss/out": 5.223508834838867, "created_at": "2025-01-14T13:58:14.235836+00:00"} {"global_step": 701, "acc_step": 0, "speed/wps": 12863.871649202783, "speed/FLOPS": 202044781062773.25, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3608836829662323, "optim/lr": 0.00042000000000000007, "optim/total_tokens": 183762944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429998, "loss/out": 5.02902889251709, "created_at": "2025-01-14T13:58:19.337733+00:00"} {"global_step": 702, "acc_step": 0, "speed/wps": 12863.362938117443, "speed/FLOPS": 202036791056135.7, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.38050663471221924, "optim/lr": 0.0004206, "optim/total_tokens": 184025088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488845, "loss/out": 5.110091686248779, "created_at": "2025-01-14T13:58:24.438422+00:00"} {"global_step": 703, "acc_step": 0, "speed/wps": 12865.245514083615, "speed/FLOPS": 202066359498614.88, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.38294440507888794, "optim/lr": 0.0004212, "optim/total_tokens": 184287232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 5.134426593780518, "created_at": "2025-01-14T13:58:29.536180+00:00"} {"global_step": 704, "acc_step": 0, "speed/wps": 12857.767103979013, "speed/FLOPS": 201948900791509.72, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3761948049068451, "optim/lr": 0.0004218, "optim/total_tokens": 184549376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489311, "loss/out": 5.054905891418457, "created_at": "2025-01-14T13:58:34.642755+00:00"} {"global_step": 705, "acc_step": 0, "speed/wps": 12850.41929216932, "speed/FLOPS": 201833493310086.97, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3477899432182312, "optim/lr": 0.0004224, "optim/total_tokens": 184811520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467670, "loss/out": 4.975529670715332, "created_at": "2025-01-14T13:58:39.750061+00:00"} {"global_step": 706, "acc_step": 0, "speed/wps": 12865.156475836551, "speed/FLOPS": 202064961030593.38, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.39353981614112854, "optim/lr": 0.000423, "optim/total_tokens": 185073664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496716, "loss/out": 5.014349460601807, "created_at": "2025-01-14T13:58:44.847598+00:00"} {"global_step": 707, "acc_step": 0, "speed/wps": 12853.612900775017, "speed/FLOPS": 201883653321717.56, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3074222803115845, "optim/lr": 0.0004236, "optim/total_tokens": 185335808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 347602, "loss/out": 5.14771842956543, "created_at": "2025-01-14T13:58:49.949774+00:00"} {"global_step": 708, "acc_step": 0, "speed/wps": 12862.860331087211, "speed/FLOPS": 202028896922071.47, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3867987394332886, "optim/lr": 0.0004242, "optim/total_tokens": 185597952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462675, "loss/out": 4.980899333953857, "created_at": "2025-01-14T13:58:55.053526+00:00"} {"global_step": 709, "acc_step": 0, "speed/wps": 12857.57780867517, "speed/FLOPS": 201945927648643.88, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.3837169110774994, "optim/lr": 0.0004248, "optim/total_tokens": 185860096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357066, "loss/out": 5.035639762878418, "created_at": "2025-01-14T13:59:00.160482+00:00"} {"global_step": 710, "acc_step": 0, "speed/wps": 12857.785732156042, "speed/FLOPS": 201949193372627.78, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3806677460670471, "optim/lr": 0.00042540000000000004, "optim/total_tokens": 186122240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497808, "loss/out": 4.917393207550049, "created_at": "2025-01-14T13:59:05.288853+00:00"} {"global_step": 711, "acc_step": 0, "speed/wps": 12862.257000584323, "speed/FLOPS": 202019420787460.6, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.47806456685066223, "optim/lr": 0.00042599999999999995, "optim/total_tokens": 186384384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481431, "loss/out": 5.008562088012695, "created_at": "2025-01-14T13:59:10.390118+00:00"} {"global_step": 712, "acc_step": 0, "speed/wps": 12833.748472381934, "speed/FLOPS": 201571655177219.12, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.46508920192718506, "optim/lr": 0.00042659999999999996, "optim/total_tokens": 186646528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333977, "loss/out": 5.083526611328125, "created_at": "2025-01-14T13:59:15.501203+00:00"} {"global_step": 713, "acc_step": 0, "speed/wps": 12859.128569084198, "speed/FLOPS": 201970284471835.75, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.43072304129600525, "optim/lr": 0.00042720000000000003, "optim/total_tokens": 186908672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481715, "loss/out": 4.917918682098389, "created_at": "2025-01-14T13:59:20.601240+00:00"} {"global_step": 714, "acc_step": 0, "speed/wps": 12855.76374237389, "speed/FLOPS": 201917435244593.94, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.45141029357910156, "optim/lr": 0.00042780000000000004, "optim/total_tokens": 187170816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 393011, "loss/out": 5.033632755279541, "created_at": "2025-01-14T13:59:25.702657+00:00"} {"global_step": 715, "acc_step": 0, "speed/wps": 12863.434118188541, "speed/FLOPS": 202037909036963.94, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4460071325302124, "optim/lr": 0.00042840000000000006, "optim/total_tokens": 187432960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493072, "loss/out": 4.96547794342041, "created_at": "2025-01-14T13:59:30.803472+00:00"} {"global_step": 716, "acc_step": 0, "speed/wps": 12853.579685805067, "speed/FLOPS": 201883131634972.34, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5504286885261536, "optim/lr": 0.00042899999999999997, "optim/total_tokens": 187695104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456556, "loss/out": 4.929163932800293, "created_at": "2025-01-14T13:59:35.910402+00:00"} {"global_step": 717, "acc_step": 0, "speed/wps": 12858.250887884145, "speed/FLOPS": 201956499282528.6, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6206048727035522, "optim/lr": 0.0004296, "optim/total_tokens": 187957248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493816, "loss/out": 5.036435127258301, "created_at": "2025-01-14T13:59:41.010826+00:00"} {"global_step": 718, "acc_step": 0, "speed/wps": 12854.446395160343, "speed/FLOPS": 201896744496380.4, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.518502414226532, "optim/lr": 0.0004302, "optim/total_tokens": 188219392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480621, "loss/out": 5.071951389312744, "created_at": "2025-01-14T13:59:46.115254+00:00"} {"global_step": 719, "acc_step": 0, "speed/wps": 12861.225349440369, "speed/FLOPS": 202003217288607.56, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.49369677901268005, "optim/lr": 0.0004308, "optim/total_tokens": 188481536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495448, "loss/out": 5.104565620422363, "created_at": "2025-01-14T13:59:51.214540+00:00"} {"global_step": 720, "acc_step": 0, "speed/wps": 12851.15156062637, "speed/FLOPS": 201844994592447.25, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4230648875236511, "optim/lr": 0.0004314, "optim/total_tokens": 188743680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349305, "loss/out": 5.043728351593018, "created_at": "2025-01-14T13:59:56.317790+00:00"} {"global_step": 721, "acc_step": 0, "speed/wps": 12856.259706817431, "speed/FLOPS": 201925225047706.3, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3794776201248169, "optim/lr": 0.000432, "optim/total_tokens": 189005824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500702, "loss/out": 4.9643402099609375, "created_at": "2025-01-14T14:00:01.419682+00:00"} {"global_step": 722, "acc_step": 0, "speed/wps": 12855.761067276497, "speed/FLOPS": 201917393228513.4, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4000356197357178, "optim/lr": 0.0004326, "optim/total_tokens": 189267968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 406528, "loss/out": 5.028817176818848, "created_at": "2025-01-14T14:00:06.525762+00:00"} {"global_step": 723, "acc_step": 0, "speed/wps": 12855.837168242979, "speed/FLOPS": 201918588498765.7, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3653193712234497, "optim/lr": 0.0004332, "optim/total_tokens": 189530112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493533, "loss/out": 5.046268463134766, "created_at": "2025-01-14T14:00:11.632168+00:00"} {"global_step": 724, "acc_step": 0, "speed/wps": 12861.36971990995, "speed/FLOPS": 202005484825218.16, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3882192373275757, "optim/lr": 0.00043380000000000003, "optim/total_tokens": 189792256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474537, "loss/out": 5.037394046783447, "created_at": "2025-01-14T14:00:16.732496+00:00"} {"global_step": 725, "acc_step": 0, "speed/wps": 12860.018830818097, "speed/FLOPS": 201984267255714.47, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5139102935791016, "optim/lr": 0.00043440000000000004, "optim/total_tokens": 190054400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485608, "loss/out": 5.0363922119140625, "created_at": "2025-01-14T14:00:21.832212+00:00"} {"global_step": 726, "acc_step": 0, "speed/wps": 12857.023696062437, "speed/FLOPS": 201937224548630.38, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5179881453514099, "optim/lr": 0.000435, "optim/total_tokens": 190316544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484091, "loss/out": 4.996272563934326, "created_at": "2025-01-14T14:00:26.935871+00:00"} {"global_step": 727, "acc_step": 0, "speed/wps": 12853.924782854263, "speed/FLOPS": 201888551858343.0, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5168842077255249, "optim/lr": 0.0004356, "optim/total_tokens": 190578688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472161, "loss/out": 4.980965614318848, "created_at": "2025-01-14T14:00:32.040937+00:00"} {"global_step": 728, "acc_step": 0, "speed/wps": 12860.000449989991, "speed/FLOPS": 201983978559551.16, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.529224157333374, "optim/lr": 0.00043620000000000003, "optim/total_tokens": 190840832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495671, "loss/out": 5.077060699462891, "created_at": "2025-01-14T14:00:37.142100+00:00"} {"global_step": 729, "acc_step": 0, "speed/wps": 12855.571753531653, "speed/FLOPS": 201914419796006.28, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.4504464566707611, "optim/lr": 0.00043680000000000005, "optim/total_tokens": 191102976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319534, "loss/out": 4.972342014312744, "created_at": "2025-01-14T14:00:42.248549+00:00"} {"global_step": 730, "acc_step": 0, "speed/wps": 12859.428839746952, "speed/FLOPS": 201975000635209.44, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.38565993309020996, "optim/lr": 0.00043740000000000006, "optim/total_tokens": 191365120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469548, "loss/out": 5.029489517211914, "created_at": "2025-01-14T14:00:47.351056+00:00"} {"global_step": 731, "acc_step": 0, "speed/wps": 12852.443941375717, "speed/FLOPS": 201865293208033.03, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4473980665206909, "optim/lr": 0.00043799999999999997, "optim/total_tokens": 191627264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300113, "loss/out": 5.04716682434082, "created_at": "2025-01-14T14:00:52.461519+00:00"} {"global_step": 732, "acc_step": 0, "speed/wps": 12858.04036964126, "speed/FLOPS": 201953192804241.7, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.33793768286705017, "optim/lr": 0.0004386, "optim/total_tokens": 191889408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476632, "loss/out": 4.960119247436523, "created_at": "2025-01-14T14:00:57.562594+00:00"} {"global_step": 733, "acc_step": 0, "speed/wps": 12858.920778405653, "speed/FLOPS": 201967020833694.9, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.40221142768859863, "optim/lr": 0.0004392, "optim/total_tokens": 192151552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487798, "loss/out": 5.161287784576416, "created_at": "2025-01-14T14:01:02.663161+00:00"} {"global_step": 734, "acc_step": 0, "speed/wps": 12861.724438827483, "speed/FLOPS": 202011056173252.88, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.35459670424461365, "optim/lr": 0.0004398, "optim/total_tokens": 192413696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503075, "loss/out": 4.850226402282715, "created_at": "2025-01-14T14:01:07.764898+00:00"} {"global_step": 735, "acc_step": 0, "speed/wps": 12866.64852252125, "speed/FLOPS": 202088395674063.56, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.42291751503944397, "optim/lr": 0.0004404, "optim/total_tokens": 192675840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465267, "loss/out": 4.940491676330566, "created_at": "2025-01-14T14:01:12.864390+00:00"} {"global_step": 736, "acc_step": 0, "speed/wps": 12868.116134228001, "speed/FLOPS": 202111446532627.9, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.3878694474697113, "optim/lr": 0.000441, "optim/total_tokens": 192937984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492487, "loss/out": 4.9327850341796875, "created_at": "2025-01-14T14:01:17.968171+00:00"} {"global_step": 737, "acc_step": 0, "speed/wps": 12860.801753404103, "speed/FLOPS": 201996564130776.06, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4284377098083496, "optim/lr": 0.0004416, "optim/total_tokens": 193200128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326486, "loss/out": 4.996632099151611, "created_at": "2025-01-14T14:01:23.068802+00:00"} {"global_step": 738, "acc_step": 0, "speed/wps": 12857.535373405111, "speed/FLOPS": 201945261144415.44, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4066256582736969, "optim/lr": 0.0004422, "optim/total_tokens": 193462272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468974, "loss/out": 4.788027763366699, "created_at": "2025-01-14T14:01:28.170771+00:00"} {"global_step": 739, "acc_step": 0, "speed/wps": 12856.903991854015, "speed/FLOPS": 201935344429546.28, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5805922746658325, "optim/lr": 0.00044280000000000003, "optim/total_tokens": 193724416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308138, "loss/out": 5.1248555183410645, "created_at": "2025-01-14T14:01:33.273125+00:00"} {"global_step": 740, "acc_step": 0, "speed/wps": 12861.707496520341, "speed/FLOPS": 202010790071038.06, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.45321494340896606, "optim/lr": 0.0004434, "optim/total_tokens": 193986560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477718, "loss/out": 5.052459239959717, "created_at": "2025-01-14T14:01:38.375133+00:00"} {"global_step": 741, "acc_step": 0, "speed/wps": 12865.418388928956, "speed/FLOPS": 202069074735615.9, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4521288275718689, "optim/lr": 0.000444, "optim/total_tokens": 194248704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355439, "loss/out": 5.089069366455078, "created_at": "2025-01-14T14:01:43.472860+00:00"} {"global_step": 742, "acc_step": 0, "speed/wps": 12860.881026131328, "speed/FLOPS": 201997809217889.28, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4569128751754761, "optim/lr": 0.0004446, "optim/total_tokens": 194510848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493362, "loss/out": 5.024258136749268, "created_at": "2025-01-14T14:01:48.576256+00:00"} {"global_step": 743, "acc_step": 0, "speed/wps": 12842.39698171282, "speed/FLOPS": 201707491900557.88, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.38974934816360474, "optim/lr": 0.00044520000000000003, "optim/total_tokens": 194772992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485082, "loss/out": 4.930826187133789, "created_at": "2025-01-14T14:01:53.683173+00:00"} {"global_step": 744, "acc_step": 0, "speed/wps": 12865.807952398134, "speed/FLOPS": 202075193365215.62, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.5340532660484314, "optim/lr": 0.00044580000000000005, "optim/total_tokens": 195035136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502792, "loss/out": 5.011207580566406, "created_at": "2025-01-14T14:01:58.786065+00:00"} {"global_step": 745, "acc_step": 0, "speed/wps": 12850.353047940125, "speed/FLOPS": 201832452853438.38, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5474467277526855, "optim/lr": 0.00044639999999999995, "optim/total_tokens": 195297280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422528, "loss/out": 4.890892505645752, "created_at": "2025-01-14T14:02:03.894989+00:00"} {"global_step": 746, "acc_step": 0, "speed/wps": 12858.576498592165, "speed/FLOPS": 201961613446134.03, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5452858805656433, "optim/lr": 0.00044699999999999997, "optim/total_tokens": 195559424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488549, "loss/out": 4.9731645584106445, "created_at": "2025-01-14T14:02:09.000089+00:00"} {"global_step": 747, "acc_step": 0, "speed/wps": 12862.067429648167, "speed/FLOPS": 202016443315410.5, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.4741012752056122, "optim/lr": 0.0004476, "optim/total_tokens": 195821568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357462, "loss/out": 5.086505889892578, "created_at": "2025-01-14T14:02:14.099097+00:00"} {"global_step": 748, "acc_step": 0, "speed/wps": 12863.920656749775, "speed/FLOPS": 202045550793642.16, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3973974883556366, "optim/lr": 0.0004482, "optim/total_tokens": 196083712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474821, "loss/out": 4.874587059020996, "created_at": "2025-01-14T14:02:19.201895+00:00"} {"global_step": 749, "acc_step": 0, "speed/wps": 12863.988129747402, "speed/FLOPS": 202046610549787.5, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.36625057458877563, "optim/lr": 0.00044880000000000007, "optim/total_tokens": 196345856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369430, "loss/out": 4.845345973968506, "created_at": "2025-01-14T14:02:24.303053+00:00"} {"global_step": 750, "acc_step": 0, "speed/wps": 12866.599475366664, "speed/FLOPS": 202087625321102.94, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.41267624497413635, "optim/lr": 0.0004494, "optim/total_tokens": 196608000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502556, "loss/out": 4.823025703430176, "created_at": "2025-01-14T14:02:29.400665+00:00"} {"global_step": 751, "acc_step": 0, "speed/wps": 12862.49391157215, "speed/FLOPS": 202023141800074.25, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.33043813705444336, "optim/lr": 0.00045, "optim/total_tokens": 196870144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430811, "loss/out": 4.93367338180542, "created_at": "2025-01-14T14:02:34.502550+00:00"} {"global_step": 752, "acc_step": 0, "speed/wps": 12856.744688032519, "speed/FLOPS": 201932842344123.3, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.3977082371711731, "optim/lr": 0.0004506, "optim/total_tokens": 197132288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489830, "loss/out": 4.938331604003906, "created_at": "2025-01-14T14:02:39.606990+00:00"} {"global_step": 753, "acc_step": 0, "speed/wps": 12859.148359694263, "speed/FLOPS": 201970595310562.0, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3575085401535034, "optim/lr": 0.0004512, "optim/total_tokens": 197394432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475867, "loss/out": 4.864139556884766, "created_at": "2025-01-14T14:02:44.709574+00:00"} {"global_step": 754, "acc_step": 0, "speed/wps": 12855.381886419882, "speed/FLOPS": 201911437672111.2, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3757317364215851, "optim/lr": 0.00045180000000000003, "optim/total_tokens": 197656576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486468, "loss/out": 4.825809478759766, "created_at": "2025-01-14T14:02:49.812107+00:00"} {"global_step": 755, "acc_step": 0, "speed/wps": 12859.672642703152, "speed/FLOPS": 201978829895656.94, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3470119833946228, "optim/lr": 0.0004524, "optim/total_tokens": 197918720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490639, "loss/out": 4.933311462402344, "created_at": "2025-01-14T14:02:54.914929+00:00"} {"global_step": 756, "acc_step": 0, "speed/wps": 12866.277810301734, "speed/FLOPS": 202082573129244.1, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3955690264701843, "optim/lr": 0.000453, "optim/total_tokens": 198180864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469548, "loss/out": 4.857002258300781, "created_at": "2025-01-14T14:03:00.013303+00:00"} {"global_step": 757, "acc_step": 0, "speed/wps": 12855.872792880078, "speed/FLOPS": 201919148032644.72, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3888667821884155, "optim/lr": 0.0004536, "optim/total_tokens": 198443008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505245, "loss/out": 4.810028076171875, "created_at": "2025-01-14T14:03:05.115444+00:00"} {"global_step": 758, "acc_step": 0, "speed/wps": 12854.521416835714, "speed/FLOPS": 201897922814882.72, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.38122040033340454, "optim/lr": 0.00045420000000000004, "optim/total_tokens": 198705152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487456, "loss/out": 4.887269496917725, "created_at": "2025-01-14T14:03:10.218557+00:00"} {"global_step": 759, "acc_step": 0, "speed/wps": 12860.981459399973, "speed/FLOPS": 201999386660383.12, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3376298248767853, "optim/lr": 0.00045480000000000005, "optim/total_tokens": 198967296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 384726, "loss/out": 4.752747535705566, "created_at": "2025-01-14T14:03:15.323336+00:00"} {"global_step": 760, "acc_step": 0, "speed/wps": 12855.033402272928, "speed/FLOPS": 201905964249715.88, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.44935721158981323, "optim/lr": 0.00045539999999999996, "optim/total_tokens": 199229440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492552, "loss/out": 4.840820789337158, "created_at": "2025-01-14T14:03:20.426434+00:00"} {"global_step": 761, "acc_step": 0, "speed/wps": 12857.27767613139, "speed/FLOPS": 201941213654618.47, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5097945928573608, "optim/lr": 0.000456, "optim/total_tokens": 199491584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473241, "loss/out": 4.923666954040527, "created_at": "2025-01-14T14:03:25.527218+00:00"} {"global_step": 762, "acc_step": 0, "speed/wps": 12848.679908931437, "speed/FLOPS": 201806173906173.0, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.5992700457572937, "optim/lr": 0.00045660000000000004, "optim/total_tokens": 199753728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494625, "loss/out": 4.871248245239258, "created_at": "2025-01-14T14:03:30.631415+00:00"} {"global_step": 763, "acc_step": 0, "speed/wps": 12860.095562706005, "speed/FLOPS": 201985472435454.72, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5401471257209778, "optim/lr": 0.00045720000000000005, "optim/total_tokens": 200015872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 4.8644118309021, "created_at": "2025-01-14T14:03:35.734460+00:00"} {"global_step": 764, "acc_step": 0, "speed/wps": 12859.777738505505, "speed/FLOPS": 201980480569649.66, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6369185447692871, "optim/lr": 0.00045780000000000007, "optim/total_tokens": 200278016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497808, "loss/out": 5.077691078186035, "created_at": "2025-01-14T14:03:40.834503+00:00"} {"global_step": 765, "acc_step": 0, "speed/wps": 12857.1531584919, "speed/FLOPS": 201939257933985.1, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5448211431503296, "optim/lr": 0.0004584, "optim/total_tokens": 200540160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488845, "loss/out": 5.129371166229248, "created_at": "2025-01-14T14:03:45.935505+00:00"} {"global_step": 766, "acc_step": 0, "speed/wps": 12860.416133708362, "speed/FLOPS": 201990507443557.47, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.5127292275428772, "optim/lr": 0.000459, "optim/total_tokens": 200802304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492777, "loss/out": 5.013075828552246, "created_at": "2025-01-14T14:03:51.040497+00:00"} {"global_step": 767, "acc_step": 0, "speed/wps": 12864.155521900535, "speed/FLOPS": 202049239673573.38, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.45794859528541565, "optim/lr": 0.0004596, "optim/total_tokens": 201064448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485082, "loss/out": 4.854440689086914, "created_at": "2025-01-14T14:03:56.141554+00:00"} {"global_step": 768, "acc_step": 0, "speed/wps": 12855.87559720127, "speed/FLOPS": 201919192078362.6, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4229035973548889, "optim/lr": 0.0004602, "optim/total_tokens": 201326592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471634, "loss/out": 4.887819290161133, "created_at": "2025-01-14T14:04:01.242816+00:00"} {"global_step": 769, "acc_step": 0, "speed/wps": 12860.212115892267, "speed/FLOPS": 201987303063405.84, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3963823616504669, "optim/lr": 0.0004608, "optim/total_tokens": 201588736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499373, "loss/out": 4.9638895988464355, "created_at": "2025-01-14T14:04:06.344219+00:00"} {"global_step": 770, "acc_step": 0, "speed/wps": 12858.565046334166, "speed/FLOPS": 201961433572685.12, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3693874478340149, "optim/lr": 0.0004614, "optim/total_tokens": 201850880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322598, "loss/out": 4.95272159576416, "created_at": "2025-01-14T14:04:11.449823+00:00"} {"global_step": 771, "acc_step": 0, "speed/wps": 12853.36684594788, "speed/FLOPS": 201879788692542.6, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3568331003189087, "optim/lr": 0.000462, "optim/total_tokens": 202113024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489068, "loss/out": 4.829586982727051, "created_at": "2025-01-14T14:04:16.554472+00:00"} {"global_step": 772, "acc_step": 0, "speed/wps": 12856.461597991078, "speed/FLOPS": 201928396026016.0, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3695448040962219, "optim/lr": 0.0004626, "optim/total_tokens": 202375168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478008, "loss/out": 4.983425140380859, "created_at": "2025-01-14T14:04:21.655623+00:00"} {"global_step": 773, "acc_step": 0, "speed/wps": 12857.179393797396, "speed/FLOPS": 201939669995509.03, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3455802798271179, "optim/lr": 0.00046320000000000004, "optim/total_tokens": 202637312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494928, "loss/out": 4.877732276916504, "created_at": "2025-01-14T14:04:26.765901+00:00"} {"global_step": 774, "acc_step": 0, "speed/wps": 12862.451544942147, "speed/FLOPS": 202022476373932.2, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3355058431625366, "optim/lr": 0.0004638, "optim/total_tokens": 202899456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468765, "loss/out": 4.862149238586426, "created_at": "2025-01-14T14:04:31.865091+00:00"} {"global_step": 775, "acc_step": 0, "speed/wps": 12859.347411219524, "speed/FLOPS": 201973721688291.66, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.40512916445732117, "optim/lr": 0.0004644, "optim/total_tokens": 203161600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497808, "loss/out": 4.943824768066406, "created_at": "2025-01-14T14:04:36.968806+00:00"} {"global_step": 776, "acc_step": 0, "speed/wps": 12858.290696368755, "speed/FLOPS": 201957124529481.97, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.45853570103645325, "optim/lr": 0.000465, "optim/total_tokens": 203423744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326732, "loss/out": 4.986367702484131, "created_at": "2025-01-14T14:04:42.069519+00:00"} {"global_step": 777, "acc_step": 0, "speed/wps": 12853.946731673066, "speed/FLOPS": 201888896594703.16, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5914759635925293, "optim/lr": 0.00046560000000000004, "optim/total_tokens": 203685888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473815, "loss/out": 4.8942718505859375, "created_at": "2025-01-14T14:04:47.173970+00:00"} {"global_step": 778, "acc_step": 0, "speed/wps": 12853.74704197712, "speed/FLOPS": 201885760193625.9, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.707091212272644, "optim/lr": 0.00046620000000000006, "optim/total_tokens": 203948032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300225, "loss/out": 4.941197395324707, "created_at": "2025-01-14T14:04:52.284120+00:00"} {"global_step": 779, "acc_step": 0, "speed/wps": 12858.948028989931, "speed/FLOPS": 201967448841567.22, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7030535936355591, "optim/lr": 0.00046679999999999996, "optim/total_tokens": 204210176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491741, "loss/out": 5.01506233215332, "created_at": "2025-01-14T14:04:57.387766+00:00"} {"global_step": 780, "acc_step": 0, "speed/wps": 12853.940039952075, "speed/FLOPS": 201888791492029.6, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.5118850469589233, "optim/lr": 0.0004674, "optim/total_tokens": 204472320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 4.917745113372803, "created_at": "2025-01-14T14:05:02.492547+00:00"} {"global_step": 781, "acc_step": 0, "speed/wps": 12853.145728168161, "speed/FLOPS": 201876315733964.47, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4208963215351105, "optim/lr": 0.000468, "optim/total_tokens": 204734464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503311, "loss/out": 4.926414489746094, "created_at": "2025-01-14T14:05:07.595068+00:00"} {"global_step": 782, "acc_step": 0, "speed/wps": 12857.823794512795, "speed/FLOPS": 201949791194243.72, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3762664794921875, "optim/lr": 0.0004686, "optim/total_tokens": 204996608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 4.958448886871338, "created_at": "2025-01-14T14:05:12.701540+00:00"} {"global_step": 783, "acc_step": 0, "speed/wps": 12857.411034014865, "speed/FLOPS": 201943308223431.38, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.43711116909980774, "optim/lr": 0.0004692, "optim/total_tokens": 205258752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486988, "loss/out": 4.857237815856934, "created_at": "2025-01-14T14:05:17.802230+00:00"} {"global_step": 784, "acc_step": 0, "speed/wps": 12861.463538828539, "speed/FLOPS": 202006958380255.3, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4381551146507263, "optim/lr": 0.0004698, "optim/total_tokens": 205520896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480335, "loss/out": 4.910282611846924, "created_at": "2025-01-14T14:05:22.901565+00:00"} {"global_step": 785, "acc_step": 0, "speed/wps": 12859.506652646693, "speed/FLOPS": 201976222793726.0, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.544479250907898, "optim/lr": 0.0004704, "optim/total_tokens": 205783040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498328, "loss/out": 4.889705657958984, "created_at": "2025-01-14T14:05:28.005772+00:00"} {"global_step": 786, "acc_step": 0, "speed/wps": 12862.205231541879, "speed/FLOPS": 202018607683510.9, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.4691060185432434, "optim/lr": 0.000471, "optim/total_tokens": 206045184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354783, "loss/out": 4.912530899047852, "created_at": "2025-01-14T14:05:33.105919+00:00"} {"global_step": 787, "acc_step": 0, "speed/wps": 12860.591986019503, "speed/FLOPS": 201993269445750.3, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.42808833718299866, "optim/lr": 0.0004716, "optim/total_tokens": 206307328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484091, "loss/out": 4.892138481140137, "created_at": "2025-01-14T14:05:38.206507+00:00"} {"global_step": 788, "acc_step": 0, "speed/wps": 12855.02797049423, "speed/FLOPS": 201905878936167.38, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.43943342566490173, "optim/lr": 0.00047220000000000004, "optim/total_tokens": 206569472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 403953, "loss/out": 4.895941257476807, "created_at": "2025-01-14T14:05:43.311401+00:00"} {"global_step": 789, "acc_step": 0, "speed/wps": 12860.744150484988, "speed/FLOPS": 201995659397776.66, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.47703489661216736, "optim/lr": 0.0004728, "optim/total_tokens": 206831616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493533, "loss/out": 4.924601078033447, "created_at": "2025-01-14T14:05:48.410956+00:00"} {"global_step": 790, "acc_step": 0, "speed/wps": 12859.009276844343, "speed/FLOPS": 201968410823284.12, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4491906762123108, "optim/lr": 0.0004734, "optim/total_tokens": 207093760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482235, "loss/out": 5.074805736541748, "created_at": "2025-01-14T14:05:53.511681+00:00"} {"global_step": 791, "acc_step": 0, "speed/wps": 12855.723674745444, "speed/FLOPS": 201916805927430.5, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3968197703361511, "optim/lr": 0.00047400000000000003, "optim/total_tokens": 207355904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488785, "loss/out": 5.01220178604126, "created_at": "2025-01-14T14:05:58.613771+00:00"} {"global_step": 792, "acc_step": 0, "speed/wps": 12863.617998786067, "speed/FLOPS": 202040797134426.5, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.41959917545318604, "optim/lr": 0.00047460000000000004, "optim/total_tokens": 207618048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484611, "loss/out": 4.8138532638549805, "created_at": "2025-01-14T14:06:03.718835+00:00"} {"global_step": 793, "acc_step": 0, "speed/wps": 12855.17841578786, "speed/FLOPS": 201908241886237.3, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.33279746770858765, "optim/lr": 0.00047520000000000006, "optim/total_tokens": 207880192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467691, "loss/out": 4.869890213012695, "created_at": "2025-01-14T14:06:08.820949+00:00"} {"global_step": 794, "acc_step": 0, "speed/wps": 12865.905503382592, "speed/FLOPS": 202076725537475.66, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3451291620731354, "optim/lr": 0.00047579999999999996, "optim/total_tokens": 208142336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489068, "loss/out": 4.826775550842285, "created_at": "2025-01-14T14:06:13.918363+00:00"} {"global_step": 795, "acc_step": 0, "speed/wps": 12860.671761503418, "speed/FLOPS": 201994522429348.22, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.38473448157310486, "optim/lr": 0.0004764, "optim/total_tokens": 208404480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475631, "loss/out": 4.841124057769775, "created_at": "2025-01-14T14:06:19.018142+00:00"} {"global_step": 796, "acc_step": 0, "speed/wps": 12862.70459918444, "speed/FLOPS": 202026450938540.38, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.37349194288253784, "optim/lr": 0.000477, "optim/total_tokens": 208666624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480904, "loss/out": 4.956757545471191, "created_at": "2025-01-14T14:06:24.118537+00:00"} {"global_step": 797, "acc_step": 0, "speed/wps": 12861.98783432516, "speed/FLOPS": 202015193161487.25, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.39375945925712585, "optim/lr": 0.0004776, "optim/total_tokens": 208928768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481663, "loss/out": 4.806784629821777, "created_at": "2025-01-14T14:06:29.220174+00:00"} {"global_step": 798, "acc_step": 0, "speed/wps": 12855.037218376141, "speed/FLOPS": 201906024186860.7, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.33433255553245544, "optim/lr": 0.00047819999999999997, "optim/total_tokens": 209190912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478008, "loss/out": 4.853964805603027, "created_at": "2025-01-14T14:06:34.321936+00:00"} {"global_step": 799, "acc_step": 0, "speed/wps": 12857.264003856895, "speed/FLOPS": 201940998912759.84, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.31139907240867615, "optim/lr": 0.0004788, "optim/total_tokens": 209453056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482471, "loss/out": 4.693953990936279, "created_at": "2025-01-14T14:06:39.428025+00:00"} {"global_step": 800, "acc_step": 0, "speed/wps": 12857.110432383643, "speed/FLOPS": 201938586861743.1, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.28966429829597473, "optim/lr": 0.0004794, "optim/total_tokens": 209715200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476914, "loss/out": 4.868239879608154, "created_at": "2025-01-14T14:06:44.529928+00:00"} {"global_step": 801, "acc_step": 0, "speed/wps": 12855.976083836007, "speed/FLOPS": 201920770359044.25, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.297990620136261, "optim/lr": 0.00048, "optim/total_tokens": 209977344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494106, "loss/out": 4.823974609375, "created_at": "2025-01-14T14:06:49.631288+00:00"} {"global_step": 802, "acc_step": 0, "speed/wps": 12853.438953156101, "speed/FLOPS": 201880921235333.4, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.31323689222335815, "optim/lr": 0.0004806, "optim/total_tokens": 210239488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463410, "loss/out": 4.651525497436523, "created_at": "2025-01-14T14:06:54.736158+00:00"} {"global_step": 803, "acc_step": 0, "speed/wps": 12859.167800232923, "speed/FLOPS": 201970900650935.72, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.29529210925102234, "optim/lr": 0.0004812, "optim/total_tokens": 210501632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473774, "loss/out": 4.745042324066162, "created_at": "2025-01-14T14:06:59.838427+00:00"} {"global_step": 804, "acc_step": 0, "speed/wps": 12860.970966579185, "speed/FLOPS": 201999221856214.0, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2823804020881653, "optim/lr": 0.0004818, "optim/total_tokens": 210763776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484611, "loss/out": 4.790071487426758, "created_at": "2025-01-14T14:07:04.937846+00:00"} {"global_step": 805, "acc_step": 0, "speed/wps": 12859.321171765407, "speed/FLOPS": 201973309561607.97, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2754075825214386, "optim/lr": 0.0004824, "optim/total_tokens": 211025920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478764, "loss/out": 4.801684379577637, "created_at": "2025-01-14T14:07:10.039543+00:00"} {"global_step": 806, "acc_step": 0, "speed/wps": 12855.07228201271, "speed/FLOPS": 201906574909455.5, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.26222291588783264, "optim/lr": 0.00048300000000000003, "optim/total_tokens": 211288064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484037, "loss/out": 4.741184234619141, "created_at": "2025-01-14T14:07:15.142974+00:00"} {"global_step": 807, "acc_step": 0, "speed/wps": 12858.380828500633, "speed/FLOPS": 201958540178467.62, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2663223147392273, "optim/lr": 0.00048360000000000005, "optim/total_tokens": 211550208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483565, "loss/out": 4.6689982414245605, "created_at": "2025-01-14T14:07:20.246653+00:00"} {"global_step": 808, "acc_step": 0, "speed/wps": 12861.78781478288, "speed/FLOPS": 202012051579720.1, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.2949617803096771, "optim/lr": 0.00048419999999999995, "optim/total_tokens": 211812352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496999, "loss/out": 4.812342166900635, "created_at": "2025-01-14T14:07:25.345640+00:00"} {"global_step": 809, "acc_step": 0, "speed/wps": 12859.99746525493, "speed/FLOPS": 201983931680185.56, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3585401475429535, "optim/lr": 0.00048479999999999997, "optim/total_tokens": 212074496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492025, "loss/out": 4.714770793914795, "created_at": "2025-01-14T14:07:30.445603+00:00"} {"global_step": 810, "acc_step": 0, "speed/wps": 12858.997344452342, "speed/FLOPS": 201968223408670.94, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3840472996234894, "optim/lr": 0.00048540000000000003, "optim/total_tokens": 212336640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503601, "loss/out": 4.717437744140625, "created_at": "2025-01-14T14:07:35.545641+00:00"} {"global_step": 811, "acc_step": 0, "speed/wps": 12853.296758601376, "speed/FLOPS": 201878687874455.66, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.38733530044555664, "optim/lr": 0.00048600000000000005, "optim/total_tokens": 212598784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482471, "loss/out": 4.767239570617676, "created_at": "2025-01-14T14:07:40.648270+00:00"} {"global_step": 812, "acc_step": 0, "speed/wps": 12862.068257179517, "speed/FLOPS": 202016456312927.5, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3298405706882477, "optim/lr": 0.00048660000000000006, "optim/total_tokens": 212860928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365983, "loss/out": 4.693765163421631, "created_at": "2025-01-14T14:07:45.752332+00:00"} {"global_step": 813, "acc_step": 0, "speed/wps": 12863.859867693658, "speed/FLOPS": 202044596017982.94, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3481306731700897, "optim/lr": 0.00048719999999999997, "optim/total_tokens": 213123072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492777, "loss/out": 4.71644401550293, "created_at": "2025-01-14T14:07:50.850495+00:00"} {"global_step": 814, "acc_step": 0, "speed/wps": 12860.963757475902, "speed/FLOPS": 201999108627340.6, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.37676191329956055, "optim/lr": 0.0004878, "optim/total_tokens": 213385216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445691, "loss/out": 4.764822959899902, "created_at": "2025-01-14T14:07:55.951706+00:00"} {"global_step": 815, "acc_step": 0, "speed/wps": 12858.277045435267, "speed/FLOPS": 201956910122813.2, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.47802311182022095, "optim/lr": 0.0004884, "optim/total_tokens": 213647360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492777, "loss/out": 4.806674957275391, "created_at": "2025-01-14T14:08:01.055632+00:00"} {"global_step": 816, "acc_step": 0, "speed/wps": 12857.631464567106, "speed/FLOPS": 201946770388157.66, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5308772325515747, "optim/lr": 0.0004890000000000001, "optim/total_tokens": 213909504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485941, "loss/out": 4.778356552124023, "created_at": "2025-01-14T14:08:06.157424+00:00"} {"global_step": 817, "acc_step": 0, "speed/wps": 12857.716026541604, "speed/FLOPS": 201948098550167.47, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.38931378722190857, "optim/lr": 0.0004896000000000001, "optim/total_tokens": 214171648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493013, "loss/out": 4.751186847686768, "created_at": "2025-01-14T14:08:11.258077+00:00"} {"global_step": 818, "acc_step": 0, "speed/wps": 12856.675728932922, "speed/FLOPS": 201931759246704.4, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5058292150497437, "optim/lr": 0.0004902, "optim/total_tokens": 214433792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483045, "loss/out": 4.798308372497559, "created_at": "2025-01-14T14:08:16.360371+00:00"} {"global_step": 819, "acc_step": 0, "speed/wps": 12861.861356498748, "speed/FLOPS": 202013206653424.72, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.43752092123031616, "optim/lr": 0.0004908, "optim/total_tokens": 214695936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488785, "loss/out": 4.762120723724365, "created_at": "2025-01-14T14:08:21.459711+00:00"} {"global_step": 820, "acc_step": 0, "speed/wps": 12862.018154109786, "speed/FLOPS": 202015669375368.56, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.40522462129592896, "optim/lr": 0.0004914, "optim/total_tokens": 214958080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485422, "loss/out": 4.762701034545898, "created_at": "2025-01-14T14:08:26.558557+00:00"} {"global_step": 821, "acc_step": 0, "speed/wps": 12860.058852368196, "speed/FLOPS": 201984895849154.16, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.38206014037132263, "optim/lr": 0.000492, "optim/total_tokens": 215220224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480621, "loss/out": 4.686897277832031, "created_at": "2025-01-14T14:08:31.658663+00:00"} {"global_step": 822, "acc_step": 0, "speed/wps": 12865.683603501324, "speed/FLOPS": 202073240294917.94, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.354207843542099, "optim/lr": 0.0004926, "optim/total_tokens": 215482368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493072, "loss/out": 4.652959823608398, "created_at": "2025-01-14T14:08:36.759172+00:00"} {"global_step": 823, "acc_step": 0, "speed/wps": 12856.594640630487, "speed/FLOPS": 201930485643487.66, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.33838972449302673, "optim/lr": 0.0004932, "optim/total_tokens": 215744512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466597, "loss/out": 4.78907585144043, "created_at": "2025-01-14T14:08:41.861834+00:00"} {"global_step": 824, "acc_step": 0, "speed/wps": 12862.545936858973, "speed/FLOPS": 202023958928694.03, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.34246668219566345, "optim/lr": 0.0004938, "optim/total_tokens": 216006656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489648, "loss/out": 4.679923057556152, "created_at": "2025-01-14T14:08:46.960538+00:00"} {"global_step": 825, "acc_step": 0, "speed/wps": 12860.145345267167, "speed/FLOPS": 201986254338986.2, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.39094990491867065, "optim/lr": 0.0004944, "optim/total_tokens": 216268800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492835, "loss/out": 4.777047157287598, "created_at": "2025-01-14T14:08:52.061333+00:00"} {"global_step": 826, "acc_step": 0, "speed/wps": 12856.645175945016, "speed/FLOPS": 201931279370045.12, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.37317830324172974, "optim/lr": 0.000495, "optim/total_tokens": 216530944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359278, "loss/out": 4.796064853668213, "created_at": "2025-01-14T14:08:57.164863+00:00"} {"global_step": 827, "acc_step": 0, "speed/wps": 12867.444047688348, "speed/FLOPS": 202100890489978.97, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.41835781931877136, "optim/lr": 0.0004956, "optim/total_tokens": 216793088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467150, "loss/out": 4.679154872894287, "created_at": "2025-01-14T14:09:02.264208+00:00"} {"global_step": 828, "acc_step": 0, "speed/wps": 12862.532128768891, "speed/FLOPS": 202023742053665.0, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4526378810405731, "optim/lr": 0.0004962, "optim/total_tokens": 217055232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329704, "loss/out": 4.795183181762695, "created_at": "2025-01-14T14:09:07.363353+00:00"} {"global_step": 829, "acc_step": 0, "speed/wps": 12859.777284935404, "speed/FLOPS": 201980473445707.94, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3830559551715851, "optim/lr": 0.0004968, "optim/total_tokens": 217317376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463984, "loss/out": 4.789534568786621, "created_at": "2025-01-14T14:09:12.466017+00:00"} {"global_step": 830, "acc_step": 0, "speed/wps": 12862.08347954364, "speed/FLOPS": 202016695401073.78, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4351239502429962, "optim/lr": 0.0004974000000000001, "optim/total_tokens": 217579520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390560, "loss/out": 4.835249900817871, "created_at": "2025-01-14T14:09:17.564928+00:00"} {"global_step": 831, "acc_step": 0, "speed/wps": 12858.363211271257, "speed/FLOPS": 201958263475671.7, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.001, "optim/grad_norm": 0.3926013708114624, "optim/lr": 0.0004980000000000001, "optim/total_tokens": 217841664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484091, "loss/out": 4.698730945587158, "created_at": "2025-01-14T14:09:22.666444+00:00"} {"global_step": 832, "acc_step": 0, "speed/wps": 12859.69024002009, "speed/FLOPS": 201979106285700.72, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.37342676520347595, "optim/lr": 0.0004986, "optim/total_tokens": 218103808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 382214, "loss/out": 4.80980920791626, "created_at": "2025-01-14T14:09:27.769113+00:00"} {"global_step": 833, "acc_step": 0, "speed/wps": 12859.69362882517, "speed/FLOPS": 201979159511541.2, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3362068235874176, "optim/lr": 0.0004992, "optim/total_tokens": 218365952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495212, "loss/out": 4.740807056427002, "created_at": "2025-01-14T14:09:32.871864+00:00"} {"global_step": 834, "acc_step": 0, "speed/wps": 12859.132535660863, "speed/FLOPS": 201970346772372.94, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.30124160647392273, "optim/lr": 0.0004998, "optim/total_tokens": 218628096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449930, "loss/out": 4.806679725646973, "created_at": "2025-01-14T14:09:37.972064+00:00"} {"global_step": 835, "acc_step": 0, "speed/wps": 12860.220083288767, "speed/FLOPS": 201987428202316.22, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3690384030342102, "optim/lr": 0.0005004, "optim/total_tokens": 218890240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486185, "loss/out": 4.666149139404297, "created_at": "2025-01-14T14:09:43.075029+00:00"} {"global_step": 836, "acc_step": 0, "speed/wps": 12862.728397812834, "speed/FLOPS": 202026824728702.06, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4072844386100769, "optim/lr": 0.000501, "optim/total_tokens": 219152384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 383840, "loss/out": 4.674252510070801, "created_at": "2025-01-14T14:09:48.176845+00:00"} {"global_step": 837, "acc_step": 0, "speed/wps": 12857.45574019397, "speed/FLOPS": 201944010395406.94, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.43767738342285156, "optim/lr": 0.0005015999999999999, "optim/total_tokens": 219414528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498854, "loss/out": 4.6538286209106445, "created_at": "2025-01-14T14:09:53.277749+00:00"} {"global_step": 838, "acc_step": 0, "speed/wps": 12858.756191749662, "speed/FLOPS": 201964435774096.8, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.30957403779029846, "optim/lr": 0.0005022, "optim/total_tokens": 219676672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301810, "loss/out": 4.788723468780518, "created_at": "2025-01-14T14:09:58.379705+00:00"} {"global_step": 839, "acc_step": 0, "speed/wps": 12862.726277046288, "speed/FLOPS": 202026791419149.2, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3419252932071686, "optim/lr": 0.0005028, "optim/total_tokens": 219938816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474301, "loss/out": 4.658124923706055, "created_at": "2025-01-14T14:10:03.485127+00:00"} {"global_step": 840, "acc_step": 0, "speed/wps": 12855.6694053494, "speed/FLOPS": 201915953551991.66, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.31605440378189087, "optim/lr": 0.0005034, "optim/total_tokens": 220200960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291904, "loss/out": 4.640711784362793, "created_at": "2025-01-14T14:10:08.591268+00:00"} {"global_step": 841, "acc_step": 0, "speed/wps": 12864.321690788474, "speed/FLOPS": 202051849584298.47, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3023720979690552, "optim/lr": 0.000504, "optim/total_tokens": 220463104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476678, "loss/out": 4.76224422454834, "created_at": "2025-01-14T14:10:13.691933+00:00"} {"global_step": 842, "acc_step": 0, "speed/wps": 12863.900592770326, "speed/FLOPS": 202045235661274.12, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3153117299079895, "optim/lr": 0.0005046, "optim/total_tokens": 220725248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 347964, "loss/out": 4.758365631103516, "created_at": "2025-01-14T14:10:18.792366+00:00"} {"global_step": 843, "acc_step": 0, "speed/wps": 12855.765463925303, "speed/FLOPS": 201917462283924.56, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.31290990114212036, "optim/lr": 0.0005052, "optim/total_tokens": 220987392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484895, "loss/out": 4.6648101806640625, "created_at": "2025-01-14T14:10:23.893753+00:00"} {"global_step": 844, "acc_step": 0, "speed/wps": 12854.759407286845, "speed/FLOPS": 201901660781950.88, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2822316586971283, "optim/lr": 0.0005058, "optim/total_tokens": 221249536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377427, "loss/out": 4.646515846252441, "created_at": "2025-01-14T14:10:29.000033+00:00"} {"global_step": 845, "acc_step": 0, "speed/wps": 12859.223262370731, "speed/FLOPS": 201971771760022.97, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.30629202723503113, "optim/lr": 0.0005064000000000001, "optim/total_tokens": 221511680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502792, "loss/out": 4.703128814697266, "created_at": "2025-01-14T14:10:34.106645+00:00"} {"global_step": 846, "acc_step": 0, "speed/wps": 12860.33550353324, "speed/FLOPS": 201989241035859.72, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.38713690638542175, "optim/lr": 0.0005070000000000001, "optim/total_tokens": 221773824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482471, "loss/out": 4.720309257507324, "created_at": "2025-01-14T14:10:39.207569+00:00"} {"global_step": 847, "acc_step": 0, "speed/wps": 12865.747117276305, "speed/FLOPS": 202074237866031.1, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3491450548171997, "optim/lr": 0.0005076, "optim/total_tokens": 222035968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489830, "loss/out": 4.681114196777344, "created_at": "2025-01-14T14:10:44.308214+00:00"} {"global_step": 848, "acc_step": 0, "speed/wps": 12862.256370513544, "speed/FLOPS": 202019410891333.22, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.32349008321762085, "optim/lr": 0.0005082, "optim/total_tokens": 222298112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501012, "loss/out": 4.72821044921875, "created_at": "2025-01-14T14:10:49.412868+00:00"} {"global_step": 849, "acc_step": 0, "speed/wps": 12860.481087965485, "speed/FLOPS": 201991527639422.1, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.39040154218673706, "optim/lr": 0.0005088, "optim/total_tokens": 222560256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483565, "loss/out": 4.715796947479248, "created_at": "2025-01-14T14:10:54.516094+00:00"} {"global_step": 850, "acc_step": 0, "speed/wps": 12858.610239873016, "speed/FLOPS": 201962143399315.2, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.34308600425720215, "optim/lr": 0.0005094, "optim/total_tokens": 222822400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501464, "loss/out": 4.671285629272461, "created_at": "2025-01-14T14:10:59.619798+00:00"} {"global_step": 851, "acc_step": 0, "speed/wps": 12852.70675266139, "speed/FLOPS": 201869421020415.06, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3119444251060486, "optim/lr": 0.00051, "optim/total_tokens": 223084544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478008, "loss/out": 4.5403151512146, "created_at": "2025-01-14T14:11:04.722541+00:00"} {"global_step": 852, "acc_step": 0, "speed/wps": 12857.355437158274, "speed/FLOPS": 201942434998400.53, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.30642083287239075, "optim/lr": 0.0005105999999999999, "optim/total_tokens": 223346688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463984, "loss/out": 4.49078369140625, "created_at": "2025-01-14T14:11:09.823398+00:00"} {"global_step": 853, "acc_step": 0, "speed/wps": 12853.482117383373, "speed/FLOPS": 201881599188836.6, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.34450963139533997, "optim/lr": 0.0005112, "optim/total_tokens": 223608832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484320, "loss/out": 4.669759750366211, "created_at": "2025-01-14T14:11:14.926330+00:00"} {"global_step": 854, "acc_step": 0, "speed/wps": 12861.027504375326, "speed/FLOPS": 202000109859993.47, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3984493315219879, "optim/lr": 0.0005118, "optim/total_tokens": 223870976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502792, "loss/out": 4.932067394256592, "created_at": "2025-01-14T14:11:20.031852+00:00"} {"global_step": 855, "acc_step": 0, "speed/wps": 12858.412576289003, "speed/FLOPS": 201959038821110.03, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.3470887243747711, "optim/lr": 0.0005124, "optim/total_tokens": 224133120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479006, "loss/out": 4.63648796081543, "created_at": "2025-01-14T14:11:25.136129+00:00"} {"global_step": 856, "acc_step": 0, "speed/wps": 12859.433739450264, "speed/FLOPS": 201975077591782.84, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.33197155594825745, "optim/lr": 0.000513, "optim/total_tokens": 224395264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495671, "loss/out": 4.680200099945068, "created_at": "2025-01-14T14:11:30.238049+00:00"} {"global_step": 857, "acc_step": 0, "speed/wps": 12855.874737585096, "speed/FLOPS": 201919178576909.34, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4047089219093323, "optim/lr": 0.0005136, "optim/total_tokens": 224657408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309406, "loss/out": 4.67003870010376, "created_at": "2025-01-14T14:11:35.339834+00:00"} {"global_step": 858, "acc_step": 0, "speed/wps": 12859.497170382896, "speed/FLOPS": 201976073861743.1, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.38378313183784485, "optim/lr": 0.0005142, "optim/total_tokens": 224919552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468211, "loss/out": 4.696712970733643, "created_at": "2025-01-14T14:11:40.442742+00:00"} {"global_step": 859, "acc_step": 0, "speed/wps": 12860.693253880452, "speed/FLOPS": 201994859996663.16, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.34857669472694397, "optim/lr": 0.0005148, "optim/total_tokens": 225181696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315917, "loss/out": 4.675132751464844, "created_at": "2025-01-14T14:11:45.542356+00:00"} {"global_step": 860, "acc_step": 0, "speed/wps": 12857.382528408538, "speed/FLOPS": 201942860503713.22, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3331966996192932, "optim/lr": 0.0005154000000000001, "optim/total_tokens": 225443840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471114, "loss/out": 4.65495491027832, "created_at": "2025-01-14T14:11:50.643472+00:00"} {"global_step": 861, "acc_step": 0, "speed/wps": 12861.762804255757, "speed/FLOPS": 202011658755023.56, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3532050848007202, "optim/lr": 0.000516, "optim/total_tokens": 225705984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323038, "loss/out": 4.551705360412598, "created_at": "2025-01-14T14:11:55.742873+00:00"} {"global_step": 862, "acc_step": 0, "speed/wps": 12846.242129061546, "speed/FLOPS": 201767885223455.78, "speed/curr_iter_time": 1.2874, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.34103646874427795, "optim/lr": 0.0005166, "optim/total_tokens": 225968128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494402, "loss/out": 4.757623672485352, "created_at": "2025-01-14T14:12:00.855796+00:00"} {"global_step": 863, "acc_step": 0, "speed/wps": 12856.887945486142, "speed/FLOPS": 201935092399288.8, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.35620611906051636, "optim/lr": 0.0005172, "optim/total_tokens": 226230272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419644, "loss/out": 4.683281421661377, "created_at": "2025-01-14T14:12:05.960165+00:00"} {"global_step": 864, "acc_step": 0, "speed/wps": 12860.897735910838, "speed/FLOPS": 201998071667938.12, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.001, "optim/grad_norm": 0.35926660895347595, "optim/lr": 0.0005178, "optim/total_tokens": 226492416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472964, "loss/out": 4.680234432220459, "created_at": "2025-01-14T14:12:11.063796+00:00"} {"global_step": 865, "acc_step": 0, "speed/wps": 12862.090741623173, "speed/FLOPS": 202016809462012.0, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3430221676826477, "optim/lr": 0.0005184, "optim/total_tokens": 226754560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309769, "loss/out": 4.641012191772461, "created_at": "2025-01-14T14:12:16.163258+00:00"} {"global_step": 866, "acc_step": 0, "speed/wps": 12861.168654764308, "speed/FLOPS": 202002326820813.38, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4090535640716553, "optim/lr": 0.0005189999999999999, "optim/total_tokens": 227016704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497825, "loss/out": 4.564682960510254, "created_at": "2025-01-14T14:12:21.264919+00:00"} {"global_step": 867, "acc_step": 0, "speed/wps": 12851.954308402437, "speed/FLOPS": 201857602849361.44, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.34999632835388184, "optim/lr": 0.0005195999999999999, "optim/total_tokens": 227278848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 339188, "loss/out": 4.5930891036987305, "created_at": "2025-01-14T14:12:26.368098+00:00"} {"global_step": 868, "acc_step": 0, "speed/wps": 12858.009333227094, "speed/FLOPS": 201952705334708.7, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.39943641424179077, "optim/lr": 0.0005202, "optim/total_tokens": 227540992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470587, "loss/out": 4.650018692016602, "created_at": "2025-01-14T14:12:31.476021+00:00"} {"global_step": 869, "acc_step": 0, "speed/wps": 12855.834136118228, "speed/FLOPS": 201918540875079.9, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.38322097063064575, "optim/lr": 0.0005208, "optim/total_tokens": 227803136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432146, "loss/out": 4.662790775299072, "created_at": "2025-01-14T14:12:36.577622+00:00"} {"global_step": 870, "acc_step": 0, "speed/wps": 12851.94989088762, "speed/FLOPS": 201857533466220.94, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.34469732642173767, "optim/lr": 0.0005214, "optim/total_tokens": 228065280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488266, "loss/out": 4.633161544799805, "created_at": "2025-01-14T14:12:41.680554+00:00"} {"global_step": 871, "acc_step": 0, "speed/wps": 12858.509232728618, "speed/FLOPS": 201960556943314.06, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.34133872389793396, "optim/lr": 0.000522, "optim/total_tokens": 228327424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491222, "loss/out": 4.698971748352051, "created_at": "2025-01-14T14:12:46.783957+00:00"} {"global_step": 872, "acc_step": 0, "speed/wps": 12855.16067816968, "speed/FLOPS": 201907963292569.75, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.44542908668518066, "optim/lr": 0.0005226, "optim/total_tokens": 228589568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486175, "loss/out": 4.592930793762207, "created_at": "2025-01-14T14:12:51.889559+00:00"} {"global_step": 873, "acc_step": 0, "speed/wps": 12853.097516374493, "speed/FLOPS": 201875558501494.97, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.35631000995635986, "optim/lr": 0.0005232, "optim/total_tokens": 228851712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491442, "loss/out": 4.532290458679199, "created_at": "2025-01-14T14:12:56.992650+00:00"} {"global_step": 874, "acc_step": 0, "speed/wps": 12863.175801142186, "speed/FLOPS": 202033851812786.12, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3735620379447937, "optim/lr": 0.0005238, "optim/total_tokens": 229113856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494928, "loss/out": 4.659246444702148, "created_at": "2025-01-14T14:13:02.091844+00:00"} {"global_step": 875, "acc_step": 0, "speed/wps": 12860.988614460524, "speed/FLOPS": 201999499040441.16, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.35519570112228394, "optim/lr": 0.0005244000000000001, "optim/total_tokens": 229376000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500182, "loss/out": 4.681088447570801, "created_at": "2025-01-14T14:13:07.192475+00:00"} {"global_step": 876, "acc_step": 0, "speed/wps": 12856.675221555293, "speed/FLOPS": 201931751277641.53, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3391437232494354, "optim/lr": 0.000525, "optim/total_tokens": 229638144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487798, "loss/out": 4.690669059753418, "created_at": "2025-01-14T14:13:12.293385+00:00"} {"global_step": 877, "acc_step": 0, "speed/wps": 12860.341569681303, "speed/FLOPS": 201989336313050.78, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.31714701652526855, "optim/lr": 0.0005256, "optim/total_tokens": 229900288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501012, "loss/out": 4.6508612632751465, "created_at": "2025-01-14T14:13:17.392956+00:00"} {"global_step": 878, "acc_step": 0, "speed/wps": 12862.252702938598, "speed/FLOPS": 202019353287028.9, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.33762305974960327, "optim/lr": 0.0005262, "optim/total_tokens": 230162432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479574, "loss/out": 4.656124591827393, "created_at": "2025-01-14T14:13:22.492749+00:00"} {"global_step": 879, "acc_step": 0, "speed/wps": 12857.750475163844, "speed/FLOPS": 201948639613117.12, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.30842357873916626, "optim/lr": 0.0005268, "optim/total_tokens": 230424576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353859, "loss/out": 4.622524738311768, "created_at": "2025-01-14T14:13:27.596710+00:00"} {"global_step": 880, "acc_step": 0, "speed/wps": 12858.792565686492, "speed/FLOPS": 201965007076757.75, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31981995701789856, "optim/lr": 0.0005274, "optim/total_tokens": 230686720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489128, "loss/out": 4.558173179626465, "created_at": "2025-01-14T14:13:32.697559+00:00"} {"global_step": 881, "acc_step": 0, "speed/wps": 12853.830822658454, "speed/FLOPS": 201887076084353.0, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.320696622133255, "optim/lr": 0.0005279999999999999, "optim/total_tokens": 230948864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480143, "loss/out": 4.621810436248779, "created_at": "2025-01-14T14:13:37.800882+00:00"} {"global_step": 882, "acc_step": 0, "speed/wps": 12859.187743706503, "speed/FLOPS": 201971213890593.53, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.34749841690063477, "optim/lr": 0.0005286, "optim/total_tokens": 231211008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493816, "loss/out": 4.6021809577941895, "created_at": "2025-01-14T14:13:42.904661+00:00"} {"global_step": 883, "acc_step": 0, "speed/wps": 12863.642386322232, "speed/FLOPS": 202041180174193.97, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3804238736629486, "optim/lr": 0.0005292, "optim/total_tokens": 231473152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471924, "loss/out": 4.656158447265625, "created_at": "2025-01-14T14:13:48.005076+00:00"} {"global_step": 884, "acc_step": 0, "speed/wps": 12857.8207097643, "speed/FLOPS": 201949742744029.75, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.33897456526756287, "optim/lr": 0.0005298000000000001, "optim/total_tokens": 231735296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491685, "loss/out": 4.533088684082031, "created_at": "2025-01-14T14:13:53.110837+00:00"} {"global_step": 885, "acc_step": 0, "speed/wps": 12862.764897678386, "speed/FLOPS": 202027398009246.6, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3188791275024414, "optim/lr": 0.0005304000000000001, "optim/total_tokens": 231997440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481188, "loss/out": 4.624672889709473, "created_at": "2025-01-14T14:13:58.213076+00:00"} {"global_step": 886, "acc_step": 0, "speed/wps": 12855.181168140783, "speed/FLOPS": 201908285115722.12, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.340232789516449, "optim/lr": 0.000531, "optim/total_tokens": 232259584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479006, "loss/out": 4.647604942321777, "created_at": "2025-01-14T14:14:03.316281+00:00"} {"global_step": 887, "acc_step": 0, "speed/wps": 12860.32306101421, "speed/FLOPS": 201989045609000.53, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3289027214050293, "optim/lr": 0.0005316, "optim/total_tokens": 232521728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498044, "loss/out": 4.674092769622803, "created_at": "2025-01-14T14:14:08.419036+00:00"} {"global_step": 888, "acc_step": 0, "speed/wps": 12852.44247423556, "speed/FLOPS": 201865270164580.78, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2920702397823334, "optim/lr": 0.0005322, "optim/total_tokens": 232783872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486752, "loss/out": 4.665356159210205, "created_at": "2025-01-14T14:14:13.526627+00:00"} {"global_step": 889, "acc_step": 0, "speed/wps": 12860.712797809289, "speed/FLOPS": 201995166960921.72, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.36134302616119385, "optim/lr": 0.0005328, "optim/total_tokens": 233046016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496716, "loss/out": 4.550950527191162, "created_at": "2025-01-14T14:14:18.626330+00:00"} {"global_step": 890, "acc_step": 0, "speed/wps": 12862.927120503067, "speed/FLOPS": 202029945941625.75, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.26656439900398254, "optim/lr": 0.0005334000000000001, "optim/total_tokens": 233308160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472161, "loss/out": 4.488427639007568, "created_at": "2025-01-14T14:14:23.729352+00:00"} {"global_step": 891, "acc_step": 0, "speed/wps": 12860.812852373256, "speed/FLOPS": 201996738455338.2, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3500635325908661, "optim/lr": 0.000534, "optim/total_tokens": 233570304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476431, "loss/out": 4.585043907165527, "created_at": "2025-01-14T14:14:28.831772+00:00"} {"global_step": 892, "acc_step": 0, "speed/wps": 12859.578347003575, "speed/FLOPS": 201977348852119.22, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.40680599212646484, "optim/lr": 0.0005346, "optim/total_tokens": 233832448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495151, "loss/out": 4.735556602478027, "created_at": "2025-01-14T14:14:33.931847+00:00"} {"global_step": 893, "acc_step": 0, "speed/wps": 12854.450030106931, "speed/FLOPS": 201896801588211.53, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.38416507840156555, "optim/lr": 0.0005352, "optim/total_tokens": 234094592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478008, "loss/out": 4.569552898406982, "created_at": "2025-01-14T14:14:39.035598+00:00"} {"global_step": 894, "acc_step": 0, "speed/wps": 12854.467864386934, "speed/FLOPS": 201897081700085.84, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.40186095237731934, "optim/lr": 0.0005358, "optim/total_tokens": 234356736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498044, "loss/out": 4.57025146484375, "created_at": "2025-01-14T14:14:44.139972+00:00"} {"global_step": 895, "acc_step": 0, "speed/wps": 12862.669823062171, "speed/FLOPS": 202025904731752.75, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3654231131076813, "optim/lr": 0.0005363999999999999, "optim/total_tokens": 234618880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467670, "loss/out": 4.705233573913574, "created_at": "2025-01-14T14:14:49.241376+00:00"} {"global_step": 896, "acc_step": 0, "speed/wps": 12857.876794623107, "speed/FLOPS": 201950623633805.97, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.39484846591949463, "optim/lr": 0.000537, "optim/total_tokens": 234881024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494118, "loss/out": 4.568873882293701, "created_at": "2025-01-14T14:14:54.342584+00:00"} {"global_step": 897, "acc_step": 0, "speed/wps": 12858.582311666505, "speed/FLOPS": 201961704748454.53, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.48693835735321045, "optim/lr": 0.0005376000000000001, "optim/total_tokens": 235143168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489648, "loss/out": 4.58382511138916, "created_at": "2025-01-14T14:14:59.442818+00:00"} {"global_step": 898, "acc_step": 0, "speed/wps": 12850.409649647916, "speed/FLOPS": 201833341861038.12, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.001, "optim/grad_norm": 0.43384072184562683, "optim/lr": 0.0005382000000000001, "optim/total_tokens": 235405312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472193, "loss/out": 4.605859756469727, "created_at": "2025-01-14T14:15:04.551328+00:00"} {"global_step": 899, "acc_step": 0, "speed/wps": 12864.905879360695, "speed/FLOPS": 202061025068584.4, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3542860746383667, "optim/lr": 0.0005388000000000001, "optim/total_tokens": 235667456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495954, "loss/out": 4.735786437988281, "created_at": "2025-01-14T14:15:09.649962+00:00"} {"global_step": 900, "acc_step": 0, "speed/wps": 12858.18476729483, "speed/FLOPS": 201955460767814.16, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.4099637269973755, "optim/lr": 0.0005394, "optim/total_tokens": 235929600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327543, "loss/out": 4.601048469543457, "created_at": "2025-01-14T14:15:14.750771+00:00"} {"global_step": 901, "acc_step": 0, "speed/wps": 12864.991174441917, "speed/FLOPS": 202062364745042.72, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.38124606013298035, "optim/lr": 0.00054, "optim/total_tokens": 236191744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504930, "loss/out": 4.679248332977295, "created_at": "2025-01-14T14:15:19.849683+00:00"} {"global_step": 902, "acc_step": 0, "speed/wps": 12855.560308973832, "speed/FLOPS": 201914240043499.25, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.37470683455467224, "optim/lr": 0.0005406, "optim/total_tokens": 236453888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327304, "loss/out": 4.642304420471191, "created_at": "2025-01-14T14:15:24.961474+00:00"} {"global_step": 903, "acc_step": 0, "speed/wps": 12861.722694522397, "speed/FLOPS": 202011028776544.6, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.34024176001548767, "optim/lr": 0.0005412, "optim/total_tokens": 236716032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473774, "loss/out": 4.6096625328063965, "created_at": "2025-01-14T14:15:30.067527+00:00"} {"global_step": 904, "acc_step": 0, "speed/wps": 12860.931621004249, "speed/FLOPS": 201998603879893.16, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.3117559552192688, "optim/lr": 0.0005418, "optim/total_tokens": 236978176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 282237, "loss/out": 4.6112751960754395, "created_at": "2025-01-14T14:15:35.170581+00:00"} {"global_step": 905, "acc_step": 0, "speed/wps": 12853.468504831559, "speed/FLOPS": 201881385385004.72, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3176608681678772, "optim/lr": 0.0005424, "optim/total_tokens": 237240320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494342, "loss/out": 4.612694263458252, "created_at": "2025-01-14T14:15:40.275381+00:00"} {"global_step": 906, "acc_step": 0, "speed/wps": 12861.19333624434, "speed/FLOPS": 202002714477373.2, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2916924059391022, "optim/lr": 0.000543, "optim/total_tokens": 237502464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467407, "loss/out": 4.574943542480469, "created_at": "2025-01-14T14:15:45.376894+00:00"} {"global_step": 907, "acc_step": 0, "speed/wps": 12867.6325465416, "speed/FLOPS": 202103851123493.84, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.28534936904907227, "optim/lr": 0.0005436, "optim/total_tokens": 237764608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497525, "loss/out": 4.548276901245117, "created_at": "2025-01-14T14:15:50.478901+00:00"} {"global_step": 908, "acc_step": 0, "speed/wps": 12854.792017800692, "speed/FLOPS": 201902172974882.38, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.30785998702049255, "optim/lr": 0.0005442, "optim/total_tokens": 238026752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319002, "loss/out": 4.58580207824707, "created_at": "2025-01-14T14:15:55.592789+00:00"} {"global_step": 909, "acc_step": 0, "speed/wps": 12866.663633392545, "speed/FLOPS": 202088633011061.78, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3376721441745758, "optim/lr": 0.0005448, "optim/total_tokens": 238288896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 4.666883945465088, "created_at": "2025-01-14T14:16:00.692990+00:00"} {"global_step": 910, "acc_step": 0, "speed/wps": 12860.88992515887, "speed/FLOPS": 201997948989345.38, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3176765739917755, "optim/lr": 0.0005454, "optim/total_tokens": 238551040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357700, "loss/out": 4.51607608795166, "created_at": "2025-01-14T14:16:05.792592+00:00"} {"global_step": 911, "acc_step": 0, "speed/wps": 12854.164918466116, "speed/FLOPS": 201892323518108.1, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3534946143627167, "optim/lr": 0.000546, "optim/total_tokens": 238813184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490175, "loss/out": 4.506312370300293, "created_at": "2025-01-14T14:16:10.897805+00:00"} {"global_step": 912, "acc_step": 0, "speed/wps": 12855.5284226127, "speed/FLOPS": 201913739224381.22, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.35498538613319397, "optim/lr": 0.0005466000000000001, "optim/total_tokens": 239075328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475383, "loss/out": 4.479129791259766, "created_at": "2025-01-14T14:16:16.003430+00:00"} {"global_step": 913, "acc_step": 0, "speed/wps": 12854.140393097117, "speed/FLOPS": 201891938313486.56, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.40857017040252686, "optim/lr": 0.0005472000000000001, "optim/total_tokens": 239337472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494106, "loss/out": 4.635398864746094, "created_at": "2025-01-14T14:16:21.105841+00:00"} {"global_step": 914, "acc_step": 0, "speed/wps": 12862.235912054877, "speed/FLOPS": 202019089563126.97, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.42460277676582336, "optim/lr": 0.0005478000000000001, "optim/total_tokens": 239599616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477763, "loss/out": 4.648759841918945, "created_at": "2025-01-14T14:16:26.205938+00:00"} {"global_step": 915, "acc_step": 0, "speed/wps": 12860.544625596713, "speed/FLOPS": 201992525585230.38, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5005937814712524, "optim/lr": 0.0005484, "optim/total_tokens": 239861760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501822, "loss/out": 4.617083549499512, "created_at": "2025-01-14T14:16:31.305456+00:00"} {"global_step": 916, "acc_step": 0, "speed/wps": 12862.38717735218, "speed/FLOPS": 202021465392481.44, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.49993354082107544, "optim/lr": 0.000549, "optim/total_tokens": 240123904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486988, "loss/out": 4.663130760192871, "created_at": "2025-01-14T14:16:36.406910+00:00"} {"global_step": 917, "acc_step": 0, "speed/wps": 12859.308658361135, "speed/FLOPS": 201973113021398.66, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4761417508125305, "optim/lr": 0.0005496, "optim/total_tokens": 240386048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500182, "loss/out": 4.519882678985596, "created_at": "2025-01-14T14:16:41.506801+00:00"} {"global_step": 918, "acc_step": 0, "speed/wps": 12850.901284144562, "speed/FLOPS": 201841063656384.53, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.4030960500240326, "optim/lr": 0.0005502, "optim/total_tokens": 240648192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362152, "loss/out": 4.524092674255371, "created_at": "2025-01-14T14:16:46.611647+00:00"} {"global_step": 919, "acc_step": 0, "speed/wps": 12855.48635784663, "speed/FLOPS": 201913078539427.03, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.40868815779685974, "optim/lr": 0.0005508, "optim/total_tokens": 240910336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483808, "loss/out": 4.641756534576416, "created_at": "2025-01-14T14:16:51.713553+00:00"} {"global_step": 920, "acc_step": 0, "speed/wps": 12857.29053796255, "speed/FLOPS": 201941415667350.84, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.40069782733917236, "optim/lr": 0.0005514, "optim/total_tokens": 241172480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415649, "loss/out": 4.571640968322754, "created_at": "2025-01-14T14:16:56.816908+00:00"} {"global_step": 921, "acc_step": 0, "speed/wps": 12863.330888413657, "speed/FLOPS": 202036287671496.8, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.44473662972450256, "optim/lr": 0.000552, "optim/total_tokens": 241434624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488561, "loss/out": 4.6266584396362305, "created_at": "2025-01-14T14:17:01.917186+00:00"} {"global_step": 922, "acc_step": 0, "speed/wps": 12863.908842228278, "speed/FLOPS": 202045365230347.22, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.34874409437179565, "optim/lr": 0.0005526, "optim/total_tokens": 241696768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435338, "loss/out": 4.643572807312012, "created_at": "2025-01-14T14:17:07.015918+00:00"} {"global_step": 923, "acc_step": 0, "speed/wps": 12853.499845674058, "speed/FLOPS": 201881877636003.0, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.38292574882507324, "optim/lr": 0.0005532, "optim/total_tokens": 241958912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489068, "loss/out": 4.480652332305908, "created_at": "2025-01-14T14:17:12.119496+00:00"} {"global_step": 924, "acc_step": 0, "speed/wps": 12857.186239237666, "speed/FLOPS": 201939777512555.06, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.36255282163619995, "optim/lr": 0.0005538, "optim/total_tokens": 242221056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482235, "loss/out": 4.463596343994141, "created_at": "2025-01-14T14:17:17.224428+00:00"} {"global_step": 925, "acc_step": 0, "speed/wps": 12861.861610392263, "speed/FLOPS": 202013210641171.25, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.38310495018959045, "optim/lr": 0.0005544, "optim/total_tokens": 242483200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486175, "loss/out": 4.5399017333984375, "created_at": "2025-01-14T14:17:22.323408+00:00"} {"global_step": 926, "acc_step": 0, "speed/wps": 12858.461690276097, "speed/FLOPS": 201959810223767.03, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3740118443965912, "optim/lr": 0.000555, "optim/total_tokens": 242745344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481188, "loss/out": 4.505893230438232, "created_at": "2025-01-14T14:17:27.440816+00:00"} {"global_step": 927, "acc_step": 0, "speed/wps": 12859.640030922472, "speed/FLOPS": 201978317682828.06, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3069991171360016, "optim/lr": 0.0005556000000000001, "optim/total_tokens": 243007488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298243, "loss/out": 4.55321741104126, "created_at": "2025-01-14T14:17:32.541268+00:00"} {"global_step": 928, "acc_step": 0, "speed/wps": 12856.913392775012, "speed/FLOPS": 201935492083928.75, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.251179963350296, "optim/lr": 0.0005562000000000001, "optim/total_tokens": 243269632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479574, "loss/out": 4.4324951171875, "created_at": "2025-01-14T14:17:37.645333+00:00"} {"global_step": 929, "acc_step": 0, "speed/wps": 12853.365049910464, "speed/FLOPS": 201879760483307.0, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.25166788697242737, "optim/lr": 0.0005568, "optim/total_tokens": 243531776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323942, "loss/out": 4.556826591491699, "created_at": "2025-01-14T14:17:42.748204+00:00"} {"global_step": 930, "acc_step": 0, "speed/wps": 12856.38382428699, "speed/FLOPS": 201927174483121.0, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.25691571831703186, "optim/lr": 0.0005574, "optim/total_tokens": 243793920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481715, "loss/out": 4.460890769958496, "created_at": "2025-01-14T14:17:47.849726+00:00"} {"global_step": 931, "acc_step": 0, "speed/wps": 12863.81159430239, "speed/FLOPS": 202043837818038.47, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.24327780306339264, "optim/lr": 0.000558, "optim/total_tokens": 244056064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479574, "loss/out": 4.524758815765381, "created_at": "2025-01-14T14:17:52.948615+00:00"} {"global_step": 932, "acc_step": 0, "speed/wps": 12857.544523875631, "speed/FLOPS": 201945404865129.0, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2678694725036621, "optim/lr": 0.0005586, "optim/total_tokens": 244318208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485892, "loss/out": 4.487585067749023, "created_at": "2025-01-14T14:17:58.049641+00:00"} {"global_step": 933, "acc_step": 0, "speed/wps": 12857.89428365117, "speed/FLOPS": 201950898323024.16, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2882000505924225, "optim/lr": 0.0005592, "optim/total_tokens": 244580352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474011, "loss/out": 4.622339248657227, "created_at": "2025-01-14T14:18:03.153134+00:00"} {"global_step": 934, "acc_step": 0, "speed/wps": 12859.25713864111, "speed/FLOPS": 201972303833400.12, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.28834861516952515, "optim/lr": 0.0005598, "optim/total_tokens": 244842496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498564, "loss/out": 4.473458766937256, "created_at": "2025-01-14T14:18:08.253225+00:00"} {"global_step": 935, "acc_step": 0, "speed/wps": 12858.933135968535, "speed/FLOPS": 201967214926201.03, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.29064294695854187, "optim/lr": 0.0005604, "optim/total_tokens": 245104640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475057, "loss/out": 4.432231426239014, "created_at": "2025-01-14T14:18:13.353357+00:00"} {"global_step": 936, "acc_step": 0, "speed/wps": 12859.757793201825, "speed/FLOPS": 201980167301247.62, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3178619146347046, "optim/lr": 0.000561, "optim/total_tokens": 245366784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496480, "loss/out": 4.524355411529541, "created_at": "2025-01-14T14:18:18.456439+00:00"} {"global_step": 937, "acc_step": 0, "speed/wps": 12852.82135493995, "speed/FLOPS": 201871221006679.88, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3078952431678772, "optim/lr": 0.0005616, "optim/total_tokens": 245628928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309195, "loss/out": 4.477552890777588, "created_at": "2025-01-14T14:18:23.561833+00:00"} {"global_step": 938, "acc_step": 0, "speed/wps": 12855.19259323613, "speed/FLOPS": 201908464562544.25, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.001, "optim/grad_norm": 0.3111628293991089, "optim/lr": 0.0005622, "optim/total_tokens": 245891072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466881, "loss/out": 4.491835594177246, "created_at": "2025-01-14T14:18:28.664619+00:00"} {"global_step": 939, "acc_step": 0, "speed/wps": 12857.564060584906, "speed/FLOPS": 201945711715994.4, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.30431950092315674, "optim/lr": 0.0005628, "optim/total_tokens": 246153216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294835, "loss/out": 4.4290361404418945, "created_at": "2025-01-14T14:18:33.766407+00:00"} {"global_step": 940, "acc_step": 0, "speed/wps": 12867.137908820412, "speed/FLOPS": 202096082158379.16, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.31672176718711853, "optim/lr": 0.0005634, "optim/total_tokens": 246415360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484328, "loss/out": 4.5548996925354, "created_at": "2025-01-14T14:18:38.867332+00:00"} {"global_step": 941, "acc_step": 0, "speed/wps": 12846.966022126755, "speed/FLOPS": 201779254958778.2, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.38085007667541504, "optim/lr": 0.000564, "optim/total_tokens": 246677504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365889, "loss/out": 4.554397106170654, "created_at": "2025-01-14T14:18:43.973262+00:00"} {"global_step": 942, "acc_step": 0, "speed/wps": 12860.626877829574, "speed/FLOPS": 201993817469573.88, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3062122166156769, "optim/lr": 0.0005646000000000001, "optim/total_tokens": 246939648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478764, "loss/out": 4.611220359802246, "created_at": "2025-01-14T14:18:49.073541+00:00"} {"global_step": 943, "acc_step": 0, "speed/wps": 12857.129331157776, "speed/FLOPS": 201938883692960.5, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3278749883174896, "optim/lr": 0.0005652000000000001, "optim/total_tokens": 247201792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322066, "loss/out": 4.545398712158203, "created_at": "2025-01-14T14:18:54.174488+00:00"} {"global_step": 944, "acc_step": 0, "speed/wps": 12861.894868178746, "speed/FLOPS": 202013733000408.94, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31428176164627075, "optim/lr": 0.0005658, "optim/total_tokens": 247463936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486185, "loss/out": 4.743680953979492, "created_at": "2025-01-14T14:18:59.279190+00:00"} {"global_step": 945, "acc_step": 0, "speed/wps": 12857.431986914224, "speed/FLOPS": 201943637317508.1, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2953815162181854, "optim/lr": 0.0005664, "optim/total_tokens": 247726080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308950, "loss/out": 4.571117401123047, "created_at": "2025-01-14T14:19:04.382922+00:00"} {"global_step": 946, "acc_step": 0, "speed/wps": 12860.355147438328, "speed/FLOPS": 201989549570382.66, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.27473580837249756, "optim/lr": 0.000567, "optim/total_tokens": 247988224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496716, "loss/out": 4.569021224975586, "created_at": "2025-01-14T14:19:09.482549+00:00"} {"global_step": 947, "acc_step": 0, "speed/wps": 12853.459057296162, "speed/FLOPS": 201881236998479.06, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.001, "optim/grad_norm": 0.26359957456588745, "optim/lr": 0.0005676, "optim/total_tokens": 248250368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315785, "loss/out": 4.5774054527282715, "created_at": "2025-01-14T14:19:14.590702+00:00"} {"global_step": 948, "acc_step": 0, "speed/wps": 12854.122660691393, "speed/FLOPS": 201891659801687.78, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.29956790804862976, "optim/lr": 0.0005682, "optim/total_tokens": 248512512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473774, "loss/out": 4.525623321533203, "created_at": "2025-01-14T14:19:19.695320+00:00"} {"global_step": 949, "acc_step": 0, "speed/wps": 12863.307616619511, "speed/FLOPS": 202035922155990.97, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.34024256467819214, "optim/lr": 0.0005688, "optim/total_tokens": 248774656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356123, "loss/out": 4.511989593505859, "created_at": "2025-01-14T14:19:24.793856+00:00"} {"global_step": 950, "acc_step": 0, "speed/wps": 12859.329589246545, "speed/FLOPS": 201973441769716.06, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.35343852639198303, "optim/lr": 0.0005694, "optim/total_tokens": 249036800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478528, "loss/out": 4.4434967041015625, "created_at": "2025-01-14T14:19:29.894024+00:00"} {"global_step": 951, "acc_step": 0, "speed/wps": 12861.258272495472, "speed/FLOPS": 202003734390429.0, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3346382677555084, "optim/lr": 0.00057, "optim/total_tokens": 249298944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348536, "loss/out": 4.4451375007629395, "created_at": "2025-01-14T14:19:34.999612+00:00"} {"global_step": 952, "acc_step": 0, "speed/wps": 12863.151316618965, "speed/FLOPS": 202033467249703.7, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3185487687587738, "optim/lr": 0.0005706, "optim/total_tokens": 249561088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484895, "loss/out": 4.40767240524292, "created_at": "2025-01-14T14:19:40.100340+00:00"} {"global_step": 953, "acc_step": 0, "speed/wps": 12859.072522651411, "speed/FLOPS": 201969404185592.97, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.29358431696891785, "optim/lr": 0.0005712, "optim/total_tokens": 249823232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461605, "loss/out": 4.4567365646362305, "created_at": "2025-01-14T14:19:45.204786+00:00"} {"global_step": 954, "acc_step": 0, "speed/wps": 12857.895524142878, "speed/FLOPS": 201950917806651.03, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2720082402229309, "optim/lr": 0.0005718, "optim/total_tokens": 250085376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492777, "loss/out": 4.42241096496582, "created_at": "2025-01-14T14:19:50.305552+00:00"} {"global_step": 955, "acc_step": 0, "speed/wps": 12850.615116629377, "speed/FLOPS": 201836569002323.2, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2743711769580841, "optim/lr": 0.0005724, "optim/total_tokens": 250347520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463722, "loss/out": 4.466122627258301, "created_at": "2025-01-14T14:19:55.411827+00:00"} {"global_step": 956, "acc_step": 0, "speed/wps": 12859.230671614674, "speed/FLOPS": 201971888132380.47, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3032642900943756, "optim/lr": 0.000573, "optim/total_tokens": 250609664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497825, "loss/out": 4.442583084106445, "created_at": "2025-01-14T14:20:00.512467+00:00"} {"global_step": 957, "acc_step": 0, "speed/wps": 12858.386542715034, "speed/FLOPS": 201958629928056.97, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2935159206390381, "optim/lr": 0.0005736000000000001, "optim/total_tokens": 250871808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301989, "loss/out": 4.4792304039001465, "created_at": "2025-01-14T14:20:05.613845+00:00"} {"global_step": 958, "acc_step": 0, "speed/wps": 12854.573706397603, "speed/FLOPS": 201898744094305.56, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.36008918285369873, "optim/lr": 0.0005742, "optim/total_tokens": 251133952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479574, "loss/out": 4.592630386352539, "created_at": "2025-01-14T14:20:10.719937+00:00"} {"global_step": 959, "acc_step": 0, "speed/wps": 12853.197126151153, "speed/FLOPS": 201877123009915.6, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.32137012481689453, "optim/lr": 0.0005748, "optim/total_tokens": 251396096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350820, "loss/out": 4.45505428314209, "created_at": "2025-01-14T14:20:15.824642+00:00"} {"global_step": 960, "acc_step": 0, "speed/wps": 12856.028711707206, "speed/FLOPS": 201921596952079.7, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.35009825229644775, "optim/lr": 0.0005754, "optim/total_tokens": 251658240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 4.424757480621338, "created_at": "2025-01-14T14:20:20.926799+00:00"} {"global_step": 961, "acc_step": 0, "speed/wps": 12856.937426026596, "speed/FLOPS": 201935869559170.03, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3685438334941864, "optim/lr": 0.000576, "optim/total_tokens": 251920384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463485, "loss/out": 4.418242454528809, "created_at": "2025-01-14T14:20:26.030560+00:00"} {"global_step": 962, "acc_step": 0, "speed/wps": 12844.134692288379, "speed/FLOPS": 201734785033011.6, "speed/curr_iter_time": 1.2882, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.43183860182762146, "optim/lr": 0.0005766, "optim/total_tokens": 252182528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 4.446852684020996, "created_at": "2025-01-14T14:20:31.139425+00:00"} {"global_step": 963, "acc_step": 0, "speed/wps": 12855.158355585672, "speed/FLOPS": 201907926813196.38, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.37587612867355347, "optim/lr": 0.0005771999999999999, "optim/total_tokens": 252444672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502653, "loss/out": 4.3997087478637695, "created_at": "2025-01-14T14:20:36.247248+00:00"} {"global_step": 964, "acc_step": 0, "speed/wps": 12850.050500086609, "speed/FLOPS": 201827700923654.6, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3824197053909302, "optim/lr": 0.0005778, "optim/total_tokens": 252706816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326583, "loss/out": 4.555803298950195, "created_at": "2025-01-14T14:20:41.354807+00:00"} {"global_step": 965, "acc_step": 0, "speed/wps": 12852.325305983168, "speed/FLOPS": 201863429876170.9, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3435152769088745, "optim/lr": 0.0005784, "optim/total_tokens": 252968960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 4.587104797363281, "created_at": "2025-01-14T14:20:46.460764+00:00"} {"global_step": 966, "acc_step": 0, "speed/wps": 12854.728079087716, "speed/FLOPS": 201901168729534.1, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.33556854724884033, "optim/lr": 0.000579, "optim/total_tokens": 253231104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356674, "loss/out": 4.422351837158203, "created_at": "2025-01-14T14:20:51.564400+00:00"} {"global_step": 967, "acc_step": 0, "speed/wps": 12853.229198156156, "speed/FLOPS": 201877626744825.7, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.32438814640045166, "optim/lr": 0.0005796, "optim/total_tokens": 253493248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503075, "loss/out": 4.487958908081055, "created_at": "2025-01-14T14:20:56.667139+00:00"} {"global_step": 968, "acc_step": 0, "speed/wps": 12861.914034916455, "speed/FLOPS": 202014034040362.53, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31352829933166504, "optim/lr": 0.0005802, "optim/total_tokens": 253755392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 378236, "loss/out": 4.28378438949585, "created_at": "2025-01-14T14:21:01.768572+00:00"} {"global_step": 969, "acc_step": 0, "speed/wps": 12864.762572529178, "speed/FLOPS": 202058774237871.22, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2794014811515808, "optim/lr": 0.0005808, "optim/total_tokens": 254017536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470067, "loss/out": 4.382750988006592, "created_at": "2025-01-14T14:21:06.867577+00:00"} {"global_step": 970, "acc_step": 0, "speed/wps": 12861.43445343047, "speed/FLOPS": 202006501554111.1, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.29376503825187683, "optim/lr": 0.0005814, "optim/total_tokens": 254279680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330235, "loss/out": 4.475175857543945, "created_at": "2025-01-14T14:21:11.966757+00:00"} {"global_step": 971, "acc_step": 0, "speed/wps": 12856.478756672588, "speed/FLOPS": 201928665526687.6, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.3103432357311249, "optim/lr": 0.000582, "optim/total_tokens": 254541824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496190, "loss/out": 4.501010894775391, "created_at": "2025-01-14T14:21:17.070674+00:00"} {"global_step": 972, "acc_step": 0, "speed/wps": 12858.111622390714, "speed/FLOPS": 201954311926583.4, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3381205201148987, "optim/lr": 0.0005826000000000001, "optim/total_tokens": 254803968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464504, "loss/out": 4.450632572174072, "created_at": "2025-01-14T14:21:22.174552+00:00"} {"global_step": 973, "acc_step": 0, "speed/wps": 12849.281558103297, "speed/FLOPS": 201815623633169.66, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.296653151512146, "optim/lr": 0.0005832, "optim/total_tokens": 255066112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476151, "loss/out": 4.342710018157959, "created_at": "2025-01-14T14:21:27.280199+00:00"} {"global_step": 974, "acc_step": 0, "speed/wps": 12852.856256244673, "speed/FLOPS": 201871769179629.97, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2722635269165039, "optim/lr": 0.0005838, "optim/total_tokens": 255328256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312861, "loss/out": 4.4195756912231445, "created_at": "2025-01-14T14:21:32.386147+00:00"} {"global_step": 975, "acc_step": 0, "speed/wps": 12840.896864895412, "speed/FLOPS": 201683930504563.97, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.2723389267921448, "optim/lr": 0.0005844, "optim/total_tokens": 255590400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468211, "loss/out": 4.419827461242676, "created_at": "2025-01-14T14:21:37.496363+00:00"} {"global_step": 976, "acc_step": 0, "speed/wps": 12857.102240974502, "speed/FLOPS": 201938458204406.38, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.25979217886924744, "optim/lr": 0.000585, "optim/total_tokens": 255852544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417484, "loss/out": 4.4358720779418945, "created_at": "2025-01-14T14:21:42.604950+00:00"} {"global_step": 977, "acc_step": 0, "speed/wps": 12853.906346655744, "speed/FLOPS": 201888262292511.28, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2617455720901489, "optim/lr": 0.0005856, "optim/total_tokens": 256114688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484563, "loss/out": 4.517578601837158, "created_at": "2025-01-14T14:21:47.709023+00:00"} {"global_step": 978, "acc_step": 0, "speed/wps": 12853.093828196561, "speed/FLOPS": 201875500573592.44, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2985531985759735, "optim/lr": 0.0005861999999999999, "optim/total_tokens": 256376832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488845, "loss/out": 4.47098445892334, "created_at": "2025-01-14T14:21:52.813012+00:00"} {"global_step": 979, "acc_step": 0, "speed/wps": 12859.657120537147, "speed/FLOPS": 201978586098710.16, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.33992713689804077, "optim/lr": 0.0005868, "optim/total_tokens": 256638976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463174, "loss/out": 4.419950485229492, "created_at": "2025-01-14T14:21:57.914871+00:00"} {"global_step": 980, "acc_step": 0, "speed/wps": 12856.786535528765, "speed/FLOPS": 201933499616556.34, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.40911591053009033, "optim/lr": 0.0005874, "optim/total_tokens": 256901120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503601, "loss/out": 4.372509002685547, "created_at": "2025-01-14T14:22:03.018825+00:00"} {"global_step": 981, "acc_step": 0, "speed/wps": 12859.929761495085, "speed/FLOPS": 201982868299602.5, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.44634369015693665, "optim/lr": 0.000588, "optim/total_tokens": 257163264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324181, "loss/out": 4.434146881103516, "created_at": "2025-01-14T14:22:08.119789+00:00"} {"global_step": 982, "acc_step": 0, "speed/wps": 12852.815521250075, "speed/FLOPS": 201871129380564.06, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.42570722103118896, "optim/lr": 0.0005886000000000001, "optim/total_tokens": 257425408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496190, "loss/out": 4.509035110473633, "created_at": "2025-01-14T14:22:13.223267+00:00"} {"global_step": 983, "acc_step": 0, "speed/wps": 12856.239593915412, "speed/FLOPS": 201924909146942.1, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.42741644382476807, "optim/lr": 0.0005892, "optim/total_tokens": 257687552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474537, "loss/out": 4.5767974853515625, "created_at": "2025-01-14T14:22:18.324830+00:00"} {"global_step": 984, "acc_step": 0, "speed/wps": 12862.926448044042, "speed/FLOPS": 202029935379732.7, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3713100552558899, "optim/lr": 0.0005898, "optim/total_tokens": 257949696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468211, "loss/out": 4.530827045440674, "created_at": "2025-01-14T14:22:23.424050+00:00"} {"global_step": 985, "acc_step": 0, "speed/wps": 12857.319327414874, "speed/FLOPS": 201941867845260.44, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3667430877685547, "optim/lr": 0.0005904, "optim/total_tokens": 258211840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503311, "loss/out": 4.547821521759033, "created_at": "2025-01-14T14:22:28.529546+00:00"} {"global_step": 986, "acc_step": 0, "speed/wps": 12860.64528630307, "speed/FLOPS": 201994106599946.0, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3455289602279663, "optim/lr": 0.000591, "optim/total_tokens": 258473984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477198, "loss/out": 4.512895584106445, "created_at": "2025-01-14T14:22:33.629253+00:00"} {"global_step": 987, "acc_step": 0, "speed/wps": 12854.996669166349, "speed/FLOPS": 201905387305800.47, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.300749808549881, "optim/lr": 0.0005916, "optim/total_tokens": 258736128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476632, "loss/out": 4.565988540649414, "created_at": "2025-01-14T14:22:38.738859+00:00"} {"global_step": 988, "acc_step": 0, "speed/wps": 12857.081014302965, "speed/FLOPS": 201938124810361.38, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3243314027786255, "optim/lr": 0.0005922, "optim/total_tokens": 258998272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498564, "loss/out": 4.428442001342773, "created_at": "2025-01-14T14:22:43.841053+00:00"} {"global_step": 989, "acc_step": 0, "speed/wps": 12848.380264367954, "speed/FLOPS": 201801467576548.72, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3537156879901886, "optim/lr": 0.0005928, "optim/total_tokens": 259260416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305449, "loss/out": 4.504040718078613, "created_at": "2025-01-14T14:22:48.948461+00:00"} {"global_step": 990, "acc_step": 0, "speed/wps": 12850.848113880937, "speed/FLOPS": 201840228544330.72, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3098101317882538, "optim/lr": 0.0005934, "optim/total_tokens": 259522560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480385, "loss/out": 4.417638778686523, "created_at": "2025-01-14T14:22:54.053298+00:00"} {"global_step": 991, "acc_step": 0, "speed/wps": 12854.568425292216, "speed/FLOPS": 201898661147288.47, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2903057336807251, "optim/lr": 0.000594, "optim/total_tokens": 259784704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357177, "loss/out": 4.487049102783203, "created_at": "2025-01-14T14:22:59.159573+00:00"} {"global_step": 992, "acc_step": 0, "speed/wps": 12852.416894407357, "speed/FLOPS": 201864868398228.5, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.3125966489315033, "optim/lr": 0.0005945999999999999, "optim/total_tokens": 260046848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500182, "loss/out": 4.332414627075195, "created_at": "2025-01-14T14:23:04.262598+00:00"} {"global_step": 993, "acc_step": 0, "speed/wps": 12863.381859980122, "speed/FLOPS": 202037088249990.4, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2965582311153412, "optim/lr": 0.0005951999999999999, "optim/total_tokens": 260308992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 4.487627983093262, "created_at": "2025-01-14T14:23:09.364757+00:00"} {"global_step": 994, "acc_step": 0, "speed/wps": 12850.799691680346, "speed/FLOPS": 201839468007131.34, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.001, "optim/grad_norm": 0.34883853793144226, "optim/lr": 0.0005958000000000001, "optim/total_tokens": 260571136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486988, "loss/out": 4.442656517028809, "created_at": "2025-01-14T14:23:14.468033+00:00"} {"global_step": 995, "acc_step": 0, "speed/wps": 12853.939737063814, "speed/FLOPS": 201888786734753.25, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3687189817428589, "optim/lr": 0.0005964000000000001, "optim/total_tokens": 260833280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463700, "loss/out": 4.475280284881592, "created_at": "2025-01-14T14:23:19.571614+00:00"} {"global_step": 996, "acc_step": 0, "speed/wps": 12864.487530010048, "speed/FLOPS": 202054454317160.25, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.37159237265586853, "optim/lr": 0.0005970000000000001, "optim/total_tokens": 261095424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501228, "loss/out": 4.484995365142822, "created_at": "2025-01-14T14:23:24.673951+00:00"} {"global_step": 997, "acc_step": 0, "speed/wps": 12852.118340717376, "speed/FLOPS": 201860179202271.56, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2821286916732788, "optim/lr": 0.0005976, "optim/total_tokens": 261357568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 386757, "loss/out": 4.493447780609131, "created_at": "2025-01-14T14:23:29.781866+00:00"} {"global_step": 998, "acc_step": 0, "speed/wps": 12856.41651820718, "speed/FLOPS": 201927687986063.75, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.36143219470977783, "optim/lr": 0.0005982, "optim/total_tokens": 261619712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482761, "loss/out": 4.441871166229248, "created_at": "2025-01-14T14:23:34.883232+00:00"} {"global_step": 999, "acc_step": 0, "speed/wps": 12844.506295923049, "speed/FLOPS": 201740621578731.0, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3019419312477112, "optim/lr": 0.0005988, "optim/total_tokens": 261881856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473524, "loss/out": 4.394381999969482, "created_at": "2025-01-14T14:23:39.994260+00:00"} {"global_step": 1000, "acc_step": 0, "speed/wps": 12860.512592285673, "speed/FLOPS": 201992022458062.12, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3271893262863159, "optim/lr": 0.0005994, "optim/total_tokens": 262144000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493013, "loss/out": 4.533898830413818, "created_at": "2025-01-14T14:23:45.095416+00:00"} {"global_step": 1001, "acc_step": 0, "speed/wps": 11884.432165737635, "speed/FLOPS": 186661338084063.5, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.421, "optim/grad_norm": 0.290772944688797, "optim/lr": 0.0006000000000000001, "optim/total_tokens": 262406144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497808, "loss/out": 4.498537063598633, "created_at": "2025-01-14T14:23:50.615689+00:00"} {"global_step": 1002, "acc_step": 0, "speed/wps": 12852.504676577235, "speed/FLOPS": 201866247137831.75, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.30135276913642883, "optim/lr": 0.0006006, "optim/total_tokens": 262668288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483565, "loss/out": 4.472476005554199, "created_at": "2025-01-14T14:23:55.721589+00:00"} {"global_step": 1003, "acc_step": 0, "speed/wps": 12847.76729001954, "speed/FLOPS": 201791839972091.84, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.27784600853919983, "optim/lr": 0.0006012, "optim/total_tokens": 262930432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476914, "loss/out": 4.335964679718018, "created_at": "2025-01-14T14:24:00.833263+00:00"} {"global_step": 1004, "acc_step": 0, "speed/wps": 12847.922031763655, "speed/FLOPS": 201794270403820.62, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2696533203125, "optim/lr": 0.0006018, "optim/total_tokens": 263192576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352628, "loss/out": 4.354565620422363, "created_at": "2025-01-14T14:24:05.938228+00:00"} {"global_step": 1005, "acc_step": 0, "speed/wps": 12846.397990260415, "speed/FLOPS": 201770333237759.75, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.29631736874580383, "optim/lr": 0.0006024, "optim/total_tokens": 263454720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501464, "loss/out": 4.431206226348877, "created_at": "2025-01-14T14:24:11.046150+00:00"} {"global_step": 1006, "acc_step": 0, "speed/wps": 12842.409110661123, "speed/FLOPS": 201707682402357.9, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2932274341583252, "optim/lr": 0.000603, "optim/total_tokens": 263716864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488035, "loss/out": 4.50029182434082, "created_at": "2025-01-14T14:24:16.154718+00:00"} {"global_step": 1007, "acc_step": 0, "speed/wps": 12840.787592644874, "speed/FLOPS": 201682214233713.47, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2847248315811157, "optim/lr": 0.0006036, "optim/total_tokens": 263979008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478764, "loss/out": 4.320567607879639, "created_at": "2025-01-14T14:24:21.263262+00:00"} {"global_step": 1008, "acc_step": 0, "speed/wps": 12845.348221114587, "speed/FLOPS": 201753845170790.56, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3229124844074249, "optim/lr": 0.0006042, "optim/total_tokens": 264241152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504930, "loss/out": 4.291650772094727, "created_at": "2025-01-14T14:24:26.368802+00:00"} {"global_step": 1009, "acc_step": 0, "speed/wps": 12846.113765687644, "speed/FLOPS": 201765869100278.88, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.29931163787841797, "optim/lr": 0.0006048000000000001, "optim/total_tokens": 264503296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494106, "loss/out": 4.429737091064453, "created_at": "2025-01-14T14:24:31.474322+00:00"} {"global_step": 1010, "acc_step": 0, "speed/wps": 12840.903796109595, "speed/FLOPS": 201684039368807.2, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.26606252789497375, "optim/lr": 0.0006054000000000001, "optim/total_tokens": 264765440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 404476, "loss/out": 4.453780651092529, "created_at": "2025-01-14T14:24:36.582422+00:00"} {"global_step": 1011, "acc_step": 0, "speed/wps": 12841.453768925227, "speed/FLOPS": 201692677447618.53, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2821178734302521, "optim/lr": 0.0006060000000000001, "optim/total_tokens": 265027584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495671, "loss/out": 4.27570915222168, "created_at": "2025-01-14T14:24:41.690133+00:00"} {"global_step": 1012, "acc_step": 0, "speed/wps": 12839.123441687578, "speed/FLOPS": 201656076456146.06, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.3627918064594269, "optim/lr": 0.0006066, "optim/total_tokens": 265289728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472444, "loss/out": 4.28944730758667, "created_at": "2025-01-14T14:24:46.801796+00:00"} {"global_step": 1013, "acc_step": 0, "speed/wps": 12835.339196334544, "speed/FLOPS": 201596639682778.62, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.34035584330558777, "optim/lr": 0.0006072, "optim/total_tokens": 265551872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468765, "loss/out": 4.314891815185547, "created_at": "2025-01-14T14:24:51.913478+00:00"} {"global_step": 1014, "acc_step": 0, "speed/wps": 12836.425340805166, "speed/FLOPS": 201613699074208.66, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.3515172600746155, "optim/lr": 0.0006078, "optim/total_tokens": 265814016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486468, "loss/out": 4.463203430175781, "created_at": "2025-01-14T14:24:57.027330+00:00"} {"global_step": 1015, "acc_step": 0, "speed/wps": 12840.15762286285, "speed/FLOPS": 201672319692617.56, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2774505019187927, "optim/lr": 0.0006084, "optim/total_tokens": 266076160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360380, "loss/out": 4.464747905731201, "created_at": "2025-01-14T14:25:02.137495+00:00"} {"global_step": 1016, "acc_step": 0, "speed/wps": 12845.63892136055, "speed/FLOPS": 201758411017617.4, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.29979047179222107, "optim/lr": 0.0006090000000000001, "optim/total_tokens": 266338304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494106, "loss/out": 4.434101581573486, "created_at": "2025-01-14T14:25:07.244718+00:00"} {"global_step": 1017, "acc_step": 0, "speed/wps": 12845.030649440656, "speed/FLOPS": 201748857271262.25, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.27558380365371704, "optim/lr": 0.0006096, "optim/total_tokens": 266600448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479095, "loss/out": 4.327096939086914, "created_at": "2025-01-14T14:25:12.352656+00:00"} {"global_step": 1018, "acc_step": 0, "speed/wps": 12839.982984936953, "speed/FLOPS": 201669576764013.7, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.28759482502937317, "optim/lr": 0.0006102, "optim/total_tokens": 266862592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483235, "loss/out": 4.350541591644287, "created_at": "2025-01-14T14:25:17.461641+00:00"} {"global_step": 1019, "acc_step": 0, "speed/wps": 12843.06763795709, "speed/FLOPS": 201718025478452.4, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2965977191925049, "optim/lr": 0.0006108, "optim/total_tokens": 267124736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487982, "loss/out": 4.347005844116211, "created_at": "2025-01-14T14:25:22.572377+00:00"} {"global_step": 1020, "acc_step": 0, "speed/wps": 12841.08446490045, "speed/FLOPS": 201686877020432.5, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3077232539653778, "optim/lr": 0.0006114, "optim/total_tokens": 267386880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313100, "loss/out": 4.338085174560547, "created_at": "2025-01-14T14:25:27.688295+00:00"} {"global_step": 1021, "acc_step": 0, "speed/wps": 12846.59066675863, "speed/FLOPS": 201773359486929.47, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.28721147775650024, "optim/lr": 0.000612, "optim/total_tokens": 267649024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469784, "loss/out": 4.440831661224365, "created_at": "2025-01-14T14:25:32.801294+00:00"} {"global_step": 1022, "acc_step": 0, "speed/wps": 12840.862103492935, "speed/FLOPS": 201683384528970.88, "speed/curr_iter_time": 1.2877, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2916995584964752, "optim/lr": 0.0006126, "optim/total_tokens": 267911168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338776, "loss/out": 4.2575531005859375, "created_at": "2025-01-14T14:25:37.916767+00:00"} {"global_step": 1023, "acc_step": 0, "speed/wps": 12842.053684555105, "speed/FLOPS": 201702099946956.22, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.26652517914772034, "optim/lr": 0.0006132, "optim/total_tokens": 268173312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483045, "loss/out": 4.4958038330078125, "created_at": "2025-01-14T14:25:43.026859+00:00"} {"global_step": 1024, "acc_step": 0, "speed/wps": 12841.406905550073, "speed/FLOPS": 201691941393914.72, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2886286973953247, "optim/lr": 0.0006138000000000001, "optim/total_tokens": 268435456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 387117, "loss/out": 4.304127216339111, "created_at": "2025-01-14T14:25:48.137991+00:00"} {"global_step": 1025, "acc_step": 0, "speed/wps": 12838.057777237309, "speed/FLOPS": 201639338731578.2, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.2860719561576843, "optim/lr": 0.0006144000000000001, "optim/total_tokens": 268697600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504930, "loss/out": 4.340085983276367, "created_at": "2025-01-14T14:25:53.247815+00:00"} {"global_step": 1026, "acc_step": 0, "speed/wps": 12841.652779213477, "speed/FLOPS": 201695803177664.28, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3066851794719696, "optim/lr": 0.000615, "optim/total_tokens": 268959744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456300, "loss/out": 4.295355796813965, "created_at": "2025-01-14T14:25:58.356080+00:00"} {"global_step": 1027, "acc_step": 0, "speed/wps": 12845.004489482933, "speed/FLOPS": 201748446393178.66, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2666165828704834, "optim/lr": 0.0006156, "optim/total_tokens": 269221888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472681, "loss/out": 4.394792079925537, "created_at": "2025-01-14T14:26:03.462707+00:00"} {"global_step": 1028, "acc_step": 0, "speed/wps": 12836.689260664027, "speed/FLOPS": 201617844298255.94, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.35895317792892456, "optim/lr": 0.0006162, "optim/total_tokens": 269484032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486937, "loss/out": 4.504317760467529, "created_at": "2025-01-14T14:26:08.573840+00:00"} {"global_step": 1029, "acc_step": 0, "speed/wps": 12839.367236662261, "speed/FLOPS": 201659905591232.88, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3745431900024414, "optim/lr": 0.0006168, "optim/total_tokens": 269746176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294783, "loss/out": 4.4601898193359375, "created_at": "2025-01-14T14:26:13.685060+00:00"} {"global_step": 1030, "acc_step": 0, "speed/wps": 12840.996107567686, "speed/FLOPS": 201685489247105.75, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.33293667435646057, "optim/lr": 0.0006174, "optim/total_tokens": 270008320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480385, "loss/out": 4.335219860076904, "created_at": "2025-01-14T14:26:18.794861+00:00"} {"global_step": 1031, "acc_step": 0, "speed/wps": 12841.66878757458, "speed/FLOPS": 201696054610973.2, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.34396037459373474, "optim/lr": 0.000618, "optim/total_tokens": 270270464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457361, "loss/out": 4.317451477050781, "created_at": "2025-01-14T14:26:23.906508+00:00"} {"global_step": 1032, "acc_step": 0, "speed/wps": 12844.0201206157, "speed/FLOPS": 201732985527454.2, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4065030515193939, "optim/lr": 0.0006186, "optim/total_tokens": 270532608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472681, "loss/out": 4.432880401611328, "created_at": "2025-01-14T14:26:29.017381+00:00"} {"global_step": 1033, "acc_step": 0, "speed/wps": 12846.13625770337, "speed/FLOPS": 201766222368293.12, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.33834126591682434, "optim/lr": 0.0006192, "optim/total_tokens": 270794752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491159, "loss/out": 4.453081130981445, "created_at": "2025-01-14T14:26:34.123107+00:00"} {"global_step": 1034, "acc_step": 0, "speed/wps": 12844.327326689052, "speed/FLOPS": 201737810620981.56, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3361625373363495, "optim/lr": 0.0006198, "optim/total_tokens": 271056896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323123, "loss/out": 4.449516296386719, "created_at": "2025-01-14T14:26:39.230440+00:00"} {"global_step": 1035, "acc_step": 0, "speed/wps": 12844.919461563675, "speed/FLOPS": 201747110912866.12, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.283591091632843, "optim/lr": 0.0006204, "optim/total_tokens": 271319040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487982, "loss/out": 4.4053850173950195, "created_at": "2025-01-14T14:26:44.340051+00:00"} {"global_step": 1036, "acc_step": 0, "speed/wps": 12839.818001807844, "speed/FLOPS": 201666985477260.47, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2808929979801178, "optim/lr": 0.000621, "optim/total_tokens": 271581184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486468, "loss/out": 4.435977458953857, "created_at": "2025-01-14T14:26:49.449012+00:00"} {"global_step": 1037, "acc_step": 0, "speed/wps": 12843.8856455107, "speed/FLOPS": 201730873411142.0, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31637418270111084, "optim/lr": 0.0006216, "optim/total_tokens": 271843328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498044, "loss/out": 4.424154281616211, "created_at": "2025-01-14T14:26:54.555988+00:00"} {"global_step": 1038, "acc_step": 0, "speed/wps": 12841.014762165187, "speed/FLOPS": 201685782243193.6, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.001, "optim/grad_norm": 0.3469458520412445, "optim/lr": 0.0006222, "optim/total_tokens": 272105472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321163, "loss/out": 4.558429718017578, "created_at": "2025-01-14T14:26:59.687738+00:00"} {"global_step": 1039, "acc_step": 0, "speed/wps": 12846.397912868271, "speed/FLOPS": 201770332022209.78, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31395211815834045, "optim/lr": 0.0006228000000000001, "optim/total_tokens": 272367616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493297, "loss/out": 4.3528571128845215, "created_at": "2025-01-14T14:27:04.795044+00:00"} {"global_step": 1040, "acc_step": 0, "speed/wps": 12841.63601629227, "speed/FLOPS": 201695539892952.06, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.35046425461769104, "optim/lr": 0.0006234000000000001, "optim/total_tokens": 272629760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487456, "loss/out": 4.3293609619140625, "created_at": "2025-01-14T14:27:09.902421+00:00"} {"global_step": 1041, "acc_step": 0, "speed/wps": 12838.754437995545, "speed/FLOPS": 201650280746098.47, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.40681666135787964, "optim/lr": 0.000624, "optim/total_tokens": 272891904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306548, "loss/out": 4.417003631591797, "created_at": "2025-01-14T14:27:15.019613+00:00"} {"global_step": 1042, "acc_step": 0, "speed/wps": 12841.202611663715, "speed/FLOPS": 201688732677700.25, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31238722801208496, "optim/lr": 0.0006246, "optim/total_tokens": 273154048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488318, "loss/out": 4.319253921508789, "created_at": "2025-01-14T14:27:20.128575+00:00"} {"global_step": 1043, "acc_step": 0, "speed/wps": 12842.073504669763, "speed/FLOPS": 201702411249092.62, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3393770754337311, "optim/lr": 0.0006252, "optim/total_tokens": 273416192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462417, "loss/out": 4.360869407653809, "created_at": "2025-01-14T14:27:25.240849+00:00"} {"global_step": 1044, "acc_step": 0, "speed/wps": 12837.782285125597, "speed/FLOPS": 201635011749397.9, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.31208300590515137, "optim/lr": 0.0006258, "optim/total_tokens": 273678336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471924, "loss/out": 4.302133083343506, "created_at": "2025-01-14T14:27:30.350531+00:00"} {"global_step": 1045, "acc_step": 0, "speed/wps": 12838.749197989464, "speed/FLOPS": 201650198444602.47, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.28711768984794617, "optim/lr": 0.0006264, "optim/total_tokens": 273940480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479858, "loss/out": 4.340253829956055, "created_at": "2025-01-14T14:27:35.461875+00:00"} {"global_step": 1046, "acc_step": 0, "speed/wps": 12847.61636604712, "speed/FLOPS": 201789469503713.72, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.34019696712493896, "optim/lr": 0.000627, "optim/total_tokens": 274202624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301696, "loss/out": 4.308907508850098, "created_at": "2025-01-14T14:27:40.566807+00:00"} {"global_step": 1047, "acc_step": 0, "speed/wps": 12836.464510820622, "speed/FLOPS": 201614314293126.62, "speed/curr_iter_time": 1.2876, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.44340959191322327, "optim/lr": 0.0006276, "optim/total_tokens": 274464768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504121, "loss/out": 4.405536651611328, "created_at": "2025-01-14T14:27:45.677021+00:00"} {"global_step": 1048, "acc_step": 0, "speed/wps": 12847.991703945516, "speed/FLOPS": 201795364701176.34, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.34150174260139465, "optim/lr": 0.0006282, "optim/total_tokens": 274726912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 412691, "loss/out": 4.5103654861450195, "created_at": "2025-01-14T14:27:50.781447+00:00"} {"global_step": 1049, "acc_step": 0, "speed/wps": 12841.520296383213, "speed/FLOPS": 201693722352764.5, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3103654682636261, "optim/lr": 0.0006288, "optim/total_tokens": 274989056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473254, "loss/out": 4.396233558654785, "created_at": "2025-01-14T14:27:55.892708+00:00"} {"global_step": 1050, "acc_step": 0, "speed/wps": 12844.288232928418, "speed/FLOPS": 201737196599750.38, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3227156102657318, "optim/lr": 0.0006294, "optim/total_tokens": 275251200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358407, "loss/out": 4.398542881011963, "created_at": "2025-01-14T14:28:01.002013+00:00"} {"global_step": 1051, "acc_step": 0, "speed/wps": 12840.025705059645, "speed/FLOPS": 201670247742244.2, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.001, "optim/grad_norm": 0.3655067980289459, "optim/lr": 0.00063, "optim/total_tokens": 275513344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502792, "loss/out": 4.349592685699463, "created_at": "2025-01-14T14:28:06.113072+00:00"} {"global_step": 1052, "acc_step": 0, "speed/wps": 12844.813222795714, "speed/FLOPS": 201745442287027.9, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3454742133617401, "optim/lr": 0.0006306, "optim/total_tokens": 275775488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478008, "loss/out": 4.375301361083984, "created_at": "2025-01-14T14:28:11.220017+00:00"} {"global_step": 1053, "acc_step": 0, "speed/wps": 12841.210578956607, "speed/FLOPS": 201688857814983.3, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.33009976148605347, "optim/lr": 0.0006312000000000001, "optim/total_tokens": 276037632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481188, "loss/out": 4.428462982177734, "created_at": "2025-01-14T14:28:16.332103+00:00"} {"global_step": 1054, "acc_step": 0, "speed/wps": 12842.387196539341, "speed/FLOPS": 201707338210962.16, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3346034288406372, "optim/lr": 0.0006318000000000001, "optim/total_tokens": 276299776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481380, "loss/out": 4.384425163269043, "created_at": "2025-01-14T14:28:21.441551+00:00"} {"global_step": 1055, "acc_step": 0, "speed/wps": 12838.549727232672, "speed/FLOPS": 201647065482265.5, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3687923550605774, "optim/lr": 0.0006324, "optim/total_tokens": 276561920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465083, "loss/out": 4.450437068939209, "created_at": "2025-01-14T14:28:26.550005+00:00"} {"global_step": 1056, "acc_step": 0, "speed/wps": 12835.43585738214, "speed/FLOPS": 201598157877357.3, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2744784951210022, "optim/lr": 0.000633, "optim/total_tokens": 276824064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481380, "loss/out": 4.2726263999938965, "created_at": "2025-01-14T14:28:31.664468+00:00"} {"global_step": 1057, "acc_step": 0, "speed/wps": 12843.794148160416, "speed/FLOPS": 201729436319525.25, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2787644863128662, "optim/lr": 0.0006336, "optim/total_tokens": 277086208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498854, "loss/out": 4.347965240478516, "created_at": "2025-01-14T14:28:36.772982+00:00"} {"global_step": 1058, "acc_step": 0, "speed/wps": 12836.688328678225, "speed/FLOPS": 201617829660138.25, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.25001877546310425, "optim/lr": 0.0006342, "optim/total_tokens": 277348352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294783, "loss/out": 4.556581020355225, "created_at": "2025-01-14T14:28:41.885763+00:00"} {"global_step": 1059, "acc_step": 0, "speed/wps": 12843.213741183436, "speed/FLOPS": 201720320230389.03, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2803844213485718, "optim/lr": 0.0006348, "optim/total_tokens": 277610496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499373, "loss/out": 4.247040748596191, "created_at": "2025-01-14T14:28:46.992157+00:00"} {"global_step": 1060, "acc_step": 0, "speed/wps": 12838.825903485338, "speed/FLOPS": 201651403209819.56, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.26385387778282166, "optim/lr": 0.0006353999999999999, "optim/total_tokens": 277872640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483235, "loss/out": 4.333377361297607, "created_at": "2025-01-14T14:28:52.105013+00:00"} {"global_step": 1061, "acc_step": 0, "speed/wps": 12842.542160282737, "speed/FLOPS": 201709772129494.7, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.29070988297462463, "optim/lr": 0.000636, "optim/total_tokens": 278134784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482709, "loss/out": 4.312595367431641, "created_at": "2025-01-14T14:28:57.212889+00:00"} {"global_step": 1062, "acc_step": 0, "speed/wps": 12846.102361414485, "speed/FLOPS": 201765689980497.8, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2904601991176605, "optim/lr": 0.0006366, "optim/total_tokens": 278396928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487798, "loss/out": 4.45140266418457, "created_at": "2025-01-14T14:29:02.328124+00:00"} {"global_step": 1063, "acc_step": 0, "speed/wps": 12846.510463138988, "speed/FLOPS": 201772099778871.66, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2663748860359192, "optim/lr": 0.0006372, "optim/total_tokens": 278659072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487515, "loss/out": 4.370692253112793, "created_at": "2025-01-14T14:29:07.435620+00:00"} {"global_step": 1064, "acc_step": 0, "speed/wps": 12843.07850944159, "speed/FLOPS": 201718196230055.44, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2707778811454773, "optim/lr": 0.0006378, "optim/total_tokens": 278921216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483754, "loss/out": 4.35949182510376, "created_at": "2025-01-14T14:29:12.545486+00:00"} {"global_step": 1065, "acc_step": 0, "speed/wps": 12844.325305764378, "speed/FLOPS": 201737778879582.53, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2859427034854889, "optim/lr": 0.0006384, "optim/total_tokens": 279183360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323801, "loss/out": 4.142602443695068, "created_at": "2025-01-14T14:29:17.653996+00:00"} {"global_step": 1066, "acc_step": 0, "speed/wps": 12833.554792975214, "speed/FLOPS": 201568613175993.16, "speed/curr_iter_time": 1.2879, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2767430245876312, "optim/lr": 0.000639, "optim/total_tokens": 279445504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476431, "loss/out": 4.398394584655762, "created_at": "2025-01-14T14:29:22.767396+00:00"} {"global_step": 1067, "acc_step": 0, "speed/wps": 12842.94766140152, "speed/FLOPS": 201716141081782.1, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24277788400650024, "optim/lr": 0.0006396, "optim/total_tokens": 279707648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447832, "loss/out": 4.269869804382324, "created_at": "2025-01-14T14:29:27.874374+00:00"} {"global_step": 1068, "acc_step": 0, "speed/wps": 12845.112297705726, "speed/FLOPS": 201750139669461.66, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.24489742517471313, "optim/lr": 0.0006402000000000001, "optim/total_tokens": 279969792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489068, "loss/out": 4.311575889587402, "created_at": "2025-01-14T14:29:32.989002+00:00"} {"global_step": 1069, "acc_step": 0, "speed/wps": 12844.460036249484, "speed/FLOPS": 201739895006990.2, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.25971120595932007, "optim/lr": 0.0006408000000000001, "optim/total_tokens": 280231936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496999, "loss/out": 4.172187805175781, "created_at": "2025-01-14T14:29:38.095757+00:00"} {"global_step": 1070, "acc_step": 0, "speed/wps": 12838.784428152878, "speed/FLOPS": 201650751782730.2, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21550148725509644, "optim/lr": 0.0006414, "optim/total_tokens": 280494080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 286807, "loss/out": 4.317715644836426, "created_at": "2025-01-14T14:29:43.207572+00:00"} {"global_step": 1071, "acc_step": 0, "speed/wps": 12839.775660257903, "speed/FLOPS": 201666320445035.25, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.28001511096954346, "optim/lr": 0.000642, "optim/total_tokens": 280756224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500202, "loss/out": 4.413408279418945, "created_at": "2025-01-14T14:29:48.315973+00:00"} {"global_step": 1072, "acc_step": 0, "speed/wps": 12839.328245711116, "speed/FLOPS": 201659293184765.97, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.24645082652568817, "optim/lr": 0.0006426, "optim/total_tokens": 281018368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467961, "loss/out": 4.3129143714904785, "created_at": "2025-01-14T14:29:53.426093+00:00"} {"global_step": 1073, "acc_step": 0, "speed/wps": 12839.48001048969, "speed/FLOPS": 201661676859160.1, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2235284447669983, "optim/lr": 0.0006432, "optim/total_tokens": 281280512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472964, "loss/out": 4.31363582611084, "created_at": "2025-01-14T14:29:58.537351+00:00"} {"global_step": 1074, "acc_step": 0, "speed/wps": 12837.760895019142, "speed/FLOPS": 201634675788382.97, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2313995063304901, "optim/lr": 0.0006438, "optim/total_tokens": 281542656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495151, "loss/out": 4.331372261047363, "created_at": "2025-01-14T14:30:03.650783+00:00"} {"global_step": 1075, "acc_step": 0, "speed/wps": 12843.53223335996, "speed/FLOPS": 201725322587675.44, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.24516542255878448, "optim/lr": 0.0006443999999999999, "optim/total_tokens": 281804800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322220, "loss/out": 4.2613420486450195, "created_at": "2025-01-14T14:30:08.759923+00:00"} {"global_step": 1076, "acc_step": 0, "speed/wps": 12842.325199901705, "speed/FLOPS": 201706364468575.78, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.27435824275016785, "optim/lr": 0.000645, "optim/total_tokens": 282066944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501464, "loss/out": 4.270073890686035, "created_at": "2025-01-14T14:30:13.870072+00:00"} {"global_step": 1077, "acc_step": 0, "speed/wps": 12843.490221277247, "speed/FLOPS": 201724662730185.78, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4721555709838867, "optim/lr": 0.0006456, "optim/total_tokens": 282329088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288341, "loss/out": 4.277255058288574, "created_at": "2025-01-14T14:30:18.977989+00:00"} {"global_step": 1078, "acc_step": 0, "speed/wps": 12841.481245485613, "speed/FLOPS": 201693109004756.16, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.461775004863739, "optim/lr": 0.0006462, "optim/total_tokens": 282591232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475057, "loss/out": 4.204251289367676, "created_at": "2025-01-14T14:30:24.087648+00:00"} {"global_step": 1079, "acc_step": 0, "speed/wps": 12839.13074814016, "speed/FLOPS": 201656191214024.0, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8595333099365234, "optim/lr": 0.0006468000000000001, "optim/total_tokens": 282853376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471924, "loss/out": 4.3837056159973145, "created_at": "2025-01-14T14:30:29.196610+00:00"} {"global_step": 1080, "acc_step": 0, "speed/wps": 12837.622473501118, "speed/FLOPS": 201632501688231.28, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6218728423118591, "optim/lr": 0.0006474, "optim/total_tokens": 283115520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487798, "loss/out": 4.361275672912598, "created_at": "2025-01-14T14:30:34.305204+00:00"} {"global_step": 1081, "acc_step": 0, "speed/wps": 12847.731215346772, "speed/FLOPS": 201791273369784.25, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7454707026481628, "optim/lr": 0.000648, "optim/total_tokens": 283377664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494106, "loss/out": 4.514736175537109, "created_at": "2025-01-14T14:30:39.411086+00:00"} {"global_step": 1082, "acc_step": 0, "speed/wps": 12840.774583386443, "speed/FLOPS": 201682009905433.2, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7504900693893433, "optim/lr": 0.0006486, "optim/total_tokens": 283639808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293250, "loss/out": 4.514212608337402, "created_at": "2025-01-14T14:30:44.522154+00:00"} {"global_step": 1083, "acc_step": 0, "speed/wps": 12841.318194618976, "speed/FLOPS": 201690548066840.25, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0220692157745361, "optim/lr": 0.0006492000000000001, "optim/total_tokens": 283901952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484320, "loss/out": 4.486527919769287, "created_at": "2025-01-14T14:30:49.630686+00:00"} {"global_step": 1084, "acc_step": 0, "speed/wps": 12842.196966354684, "speed/FLOPS": 201704350384508.44, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8522510528564453, "optim/lr": 0.0006498, "optim/total_tokens": 284164096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493882, "loss/out": 4.510671138763428, "created_at": "2025-01-14T14:30:54.741909+00:00"} {"global_step": 1085, "acc_step": 0, "speed/wps": 12844.802071707396, "speed/FLOPS": 201745267143862.7, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9042350053787231, "optim/lr": 0.0006504, "optim/total_tokens": 284426240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473254, "loss/out": 4.758752822875977, "created_at": "2025-01-14T14:30:59.850000+00:00"} {"global_step": 1086, "acc_step": 0, "speed/wps": 12838.387779067663, "speed/FLOPS": 201644521863793.94, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.08413565158844, "optim/lr": 0.000651, "optim/total_tokens": 284688384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463434, "loss/out": 4.513358116149902, "created_at": "2025-01-14T14:31:04.958942+00:00"} {"global_step": 1087, "acc_step": 0, "speed/wps": 12844.08195310243, "speed/FLOPS": 201733956691625.03, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7882614731788635, "optim/lr": 0.0006516, "optim/total_tokens": 284950528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312171, "loss/out": 4.6526994705200195, "created_at": "2025-01-14T14:31:10.065141+00:00"} {"global_step": 1088, "acc_step": 0, "speed/wps": 12838.348445065676, "speed/FLOPS": 201643904069242.12, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.728906512260437, "optim/lr": 0.0006522, "optim/total_tokens": 285212672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483235, "loss/out": 4.632657051086426, "created_at": "2025-01-14T14:31:15.177502+00:00"} {"global_step": 1089, "acc_step": 0, "speed/wps": 12844.01080652626, "speed/FLOPS": 201732839236880.62, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5693209171295166, "optim/lr": 0.0006527999999999999, "optim/total_tokens": 285474816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305449, "loss/out": 4.5803022384643555, "created_at": "2025-01-14T14:31:20.285870+00:00"} {"global_step": 1090, "acc_step": 0, "speed/wps": 12842.492818607272, "speed/FLOPS": 201708997150678.94, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5213007926940918, "optim/lr": 0.0006533999999999999, "optim/total_tokens": 285736960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471388, "loss/out": 4.695590496063232, "created_at": "2025-01-14T14:31:25.397510+00:00"} {"global_step": 1091, "acc_step": 0, "speed/wps": 12846.75516963142, "speed/FLOPS": 201775943230597.28, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4420478641986847, "optim/lr": 0.000654, "optim/total_tokens": 285999104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 389222, "loss/out": 4.597036361694336, "created_at": "2025-01-14T14:31:30.502361+00:00"} {"global_step": 1092, "acc_step": 0, "speed/wps": 12844.1807175222, "speed/FLOPS": 201735507922554.06, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.38671961426734924, "optim/lr": 0.0006546000000000001, "optim/total_tokens": 286261248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492777, "loss/out": 4.4031877517700195, "created_at": "2025-01-14T14:31:35.608758+00:00"} {"global_step": 1093, "acc_step": 0, "speed/wps": 12836.620589127793, "speed/FLOPS": 201616765717414.8, "speed/curr_iter_time": 1.2889, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.3905285596847534, "optim/lr": 0.0006552000000000001, "optim/total_tokens": 286523392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479858, "loss/out": 4.521501064300537, "created_at": "2025-01-14T14:31:40.721941+00:00"} {"global_step": 1094, "acc_step": 0, "speed/wps": 12840.24455584566, "speed/FLOPS": 201673685094571.03, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3486853837966919, "optim/lr": 0.0006558, "optim/total_tokens": 286785536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309769, "loss/out": 4.508117198944092, "created_at": "2025-01-14T14:31:45.834680+00:00"} {"global_step": 1095, "acc_step": 0, "speed/wps": 12843.046661528993, "speed/FLOPS": 201717696014824.5, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.3053434491157532, "optim/lr": 0.0006564, "optim/total_tokens": 287047680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 4.26158332824707, "created_at": "2025-01-14T14:31:50.941714+00:00"} {"global_step": 1096, "acc_step": 0, "speed/wps": 12840.280192523389, "speed/FLOPS": 201674244817564.7, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3121289014816284, "optim/lr": 0.000657, "optim/total_tokens": 287309824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418886, "loss/out": 4.417936325073242, "created_at": "2025-01-14T14:31:56.049456+00:00"} {"global_step": 1097, "acc_step": 0, "speed/wps": 12839.898021141658, "speed/FLOPS": 201668242290856.1, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.39852723479270935, "optim/lr": 0.0006576, "optim/total_tokens": 287571968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484611, "loss/out": 4.392707824707031, "created_at": "2025-01-14T14:32:01.158524+00:00"} {"global_step": 1098, "acc_step": 0, "speed/wps": 12850.962755769704, "speed/FLOPS": 201842029152728.28, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5283666253089905, "optim/lr": 0.0006582000000000001, "optim/total_tokens": 287834112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489311, "loss/out": 4.43520975112915, "created_at": "2025-01-14T14:32:06.263867+00:00"} {"global_step": 1099, "acc_step": 0, "speed/wps": 12842.277141903796, "speed/FLOPS": 201705609651680.0, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.4941917061805725, "optim/lr": 0.0006588, "optim/total_tokens": 288096256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310176, "loss/out": 4.387411117553711, "created_at": "2025-01-14T14:32:11.371209+00:00"} {"global_step": 1100, "acc_step": 0, "speed/wps": 12839.462948691402, "speed/FLOPS": 201661408880172.62, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.41438615322113037, "optim/lr": 0.0006594, "optim/total_tokens": 288358400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474301, "loss/out": 4.502269268035889, "created_at": "2025-01-14T14:32:16.484465+00:00"} {"global_step": 1101, "acc_step": 0, "speed/wps": 12846.389209775767, "speed/FLOPS": 201770195328182.7, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.34679660201072693, "optim/lr": 0.00066, "optim/total_tokens": 288620544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309482, "loss/out": 4.319178581237793, "created_at": "2025-01-14T14:32:21.589641+00:00"} {"global_step": 1102, "acc_step": 0, "speed/wps": 12850.006877864591, "speed/FLOPS": 201827015776714.56, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3622605502605438, "optim/lr": 0.0006606, "optim/total_tokens": 288882688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493533, "loss/out": 4.393314361572266, "created_at": "2025-01-14T14:32:26.697326+00:00"} {"global_step": 1103, "acc_step": 0, "speed/wps": 12842.632552253132, "speed/FLOPS": 201711191859601.3, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3020920753479004, "optim/lr": 0.0006612, "optim/total_tokens": 289144832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308833, "loss/out": 4.277711868286133, "created_at": "2025-01-14T14:32:31.805839+00:00"} {"global_step": 1104, "acc_step": 0, "speed/wps": 12846.012907870534, "speed/FLOPS": 201764284989668.0, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3396454453468323, "optim/lr": 0.0006617999999999999, "optim/total_tokens": 289406976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471114, "loss/out": 4.333615303039551, "created_at": "2025-01-14T14:32:36.913728+00:00"} {"global_step": 1105, "acc_step": 0, "speed/wps": 12843.502830499185, "speed/FLOPS": 201724860775343.22, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31766822934150696, "optim/lr": 0.0006624000000000001, "optim/total_tokens": 289669120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359993, "loss/out": 4.34700345993042, "created_at": "2025-01-14T14:32:42.020807+00:00"} {"global_step": 1106, "acc_step": 0, "speed/wps": 12839.922920934634, "speed/FLOPS": 201668633376320.7, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.249442458152771, "optim/lr": 0.0006630000000000001, "optim/total_tokens": 289931264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498328, "loss/out": 4.4116129875183105, "created_at": "2025-01-14T14:32:47.130488+00:00"} {"global_step": 1107, "acc_step": 0, "speed/wps": 12846.444915869171, "speed/FLOPS": 201771070268927.8, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.26499438285827637, "optim/lr": 0.0006636000000000001, "optim/total_tokens": 290193408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495954, "loss/out": 4.306848526000977, "created_at": "2025-01-14T14:32:52.246532+00:00"} {"global_step": 1108, "acc_step": 0, "speed/wps": 12843.53344529612, "speed/FLOPS": 201725341622798.22, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2566104233264923, "optim/lr": 0.0006642000000000001, "optim/total_tokens": 290455552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290218, "loss/out": 4.402801036834717, "created_at": "2025-01-14T14:32:57.355862+00:00"} {"global_step": 1109, "acc_step": 0, "speed/wps": 12846.209378991047, "speed/FLOPS": 201767370838595.28, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25748956203460693, "optim/lr": 0.0006648, "optim/total_tokens": 290717696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 4.2836503982543945, "created_at": "2025-01-14T14:33:02.461048+00:00"} {"global_step": 1110, "acc_step": 0, "speed/wps": 12843.21853477574, "speed/FLOPS": 201720395520343.25, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25830110907554626, "optim/lr": 0.0006654, "optim/total_tokens": 290979840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478081, "loss/out": 4.271620750427246, "created_at": "2025-01-14T14:33:07.567642+00:00"} {"global_step": 1111, "acc_step": 0, "speed/wps": 12841.403702130528, "speed/FLOPS": 201691891079808.8, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25664788484573364, "optim/lr": 0.000666, "optim/total_tokens": 291241984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479906, "loss/out": 4.3776350021362305, "created_at": "2025-01-14T14:33:12.676914+00:00"} {"global_step": 1112, "acc_step": 0, "speed/wps": 12840.230465237517, "speed/FLOPS": 201673463782207.62, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2757446765899658, "optim/lr": 0.0006666, "optim/total_tokens": 291504128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 4.236489295959473, "created_at": "2025-01-14T14:33:17.784498+00:00"} {"global_step": 1113, "acc_step": 0, "speed/wps": 12837.713051513374, "speed/FLOPS": 201633924340381.0, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2430713027715683, "optim/lr": 0.0006672, "optim/total_tokens": 291766272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306907, "loss/out": 4.428747177124023, "created_at": "2025-01-14T14:33:22.901521+00:00"} {"global_step": 1114, "acc_step": 0, "speed/wps": 12839.546843135851, "speed/FLOPS": 201662726557705.75, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.24941058456897736, "optim/lr": 0.0006678, "optim/total_tokens": 292028416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 4.332709312438965, "created_at": "2025-01-14T14:33:28.009387+00:00"} {"global_step": 1115, "acc_step": 0, "speed/wps": 12838.036435398259, "speed/FLOPS": 201639003528669.12, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.26028308272361755, "optim/lr": 0.0006684, "optim/total_tokens": 292290560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336633, "loss/out": 4.214041233062744, "created_at": "2025-01-14T14:33:33.119663+00:00"} {"global_step": 1116, "acc_step": 0, "speed/wps": 12843.709895796283, "speed/FLOPS": 201728113020371.66, "speed/curr_iter_time": 1.2874, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.22836042940616608, "optim/lr": 0.000669, "optim/total_tokens": 292552704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472161, "loss/out": 4.350527763366699, "created_at": "2025-01-14T14:33:38.228905+00:00"} {"global_step": 1117, "acc_step": 0, "speed/wps": 12845.832660829092, "speed/FLOPS": 201761453962196.78, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2447691708803177, "optim/lr": 0.0006696, "optim/total_tokens": 292814848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456840, "loss/out": 4.371964454650879, "created_at": "2025-01-14T14:33:43.340230+00:00"} {"global_step": 1118, "acc_step": 0, "speed/wps": 12844.674044251584, "speed/FLOPS": 201743256296741.0, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.24433737993240356, "optim/lr": 0.0006702, "optim/total_tokens": 293076992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485941, "loss/out": 4.197805404663086, "created_at": "2025-01-14T14:33:48.446088+00:00"} {"global_step": 1119, "acc_step": 0, "speed/wps": 12842.764221265483, "speed/FLOPS": 201713259902372.47, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2530551850795746, "optim/lr": 0.0006708, "optim/total_tokens": 293339136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502058, "loss/out": 4.3145856857299805, "created_at": "2025-01-14T14:33:53.553015+00:00"} {"global_step": 1120, "acc_step": 0, "speed/wps": 12846.133443558838, "speed/FLOPS": 201766178168286.22, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.29508066177368164, "optim/lr": 0.0006714000000000001, "optim/total_tokens": 293601280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304678, "loss/out": 4.330568313598633, "created_at": "2025-01-14T14:33:58.659339+00:00"} {"global_step": 1121, "acc_step": 0, "speed/wps": 12846.550196615897, "speed/FLOPS": 201772723847726.03, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3027353584766388, "optim/lr": 0.0006720000000000001, "optim/total_tokens": 293863424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478244, "loss/out": 4.239729404449463, "created_at": "2025-01-14T14:34:03.768263+00:00"} {"global_step": 1122, "acc_step": 0, "speed/wps": 12840.322865420818, "speed/FLOPS": 201674915054057.53, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24942968785762787, "optim/lr": 0.0006726000000000001, "optim/total_tokens": 294125568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 4.157497406005859, "created_at": "2025-01-14T14:34:08.879741+00:00"} {"global_step": 1123, "acc_step": 0, "speed/wps": 12837.07416363591, "speed/FLOPS": 201623889728341.5, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2792978882789612, "optim/lr": 0.0006732, "optim/total_tokens": 294387712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491159, "loss/out": 4.269883155822754, "created_at": "2025-01-14T14:34:13.992441+00:00"} {"global_step": 1124, "acc_step": 0, "speed/wps": 12837.610126417676, "speed/FLOPS": 201632307760319.16, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2806459665298462, "optim/lr": 0.0006738, "optim/total_tokens": 294649856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485138, "loss/out": 4.300049781799316, "created_at": "2025-01-14T14:34:19.105909+00:00"} {"global_step": 1125, "acc_step": 0, "speed/wps": 12837.808310280618, "speed/FLOPS": 201635420510219.8, "speed/curr_iter_time": 1.2874, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.271741658449173, "optim/lr": 0.0006744, "optim/total_tokens": 294912000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361434, "loss/out": 4.25691032409668, "created_at": "2025-01-14T14:34:24.216150+00:00"} {"global_step": 1126, "acc_step": 0, "speed/wps": 12835.621757289193, "speed/FLOPS": 201601077690850.8, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.30712082982063293, "optim/lr": 0.000675, "optim/total_tokens": 295174144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496716, "loss/out": 4.284162521362305, "created_at": "2025-01-14T14:34:29.330612+00:00"} {"global_step": 1127, "acc_step": 0, "speed/wps": 12832.998622987197, "speed/FLOPS": 201559877762073.72, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.30363771319389343, "optim/lr": 0.0006756, "optim/total_tokens": 295436288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500702, "loss/out": 4.304912567138672, "created_at": "2025-01-14T14:34:34.444554+00:00"} {"global_step": 1128, "acc_step": 0, "speed/wps": 12839.680503487185, "speed/FLOPS": 201664825877192.03, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25171688199043274, "optim/lr": 0.0006762, "optim/total_tokens": 295698432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470067, "loss/out": 4.229434013366699, "created_at": "2025-01-14T14:34:39.552590+00:00"} {"global_step": 1129, "acc_step": 0, "speed/wps": 12826.488110242126, "speed/FLOPS": 201457621213030.34, "speed/curr_iter_time": 1.2881, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23313377797603607, "optim/lr": 0.0006768, "optim/total_tokens": 295960576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494106, "loss/out": 4.198880195617676, "created_at": "2025-01-14T14:34:44.665680+00:00"} {"global_step": 1130, "acc_step": 0, "speed/wps": 12834.265119828187, "speed/FLOPS": 201579769835310.5, "speed/curr_iter_time": 1.2874, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.28144600987434387, "optim/lr": 0.0006774, "optim/total_tokens": 296222720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324619, "loss/out": 4.156915187835693, "created_at": "2025-01-14T14:34:49.776144+00:00"} {"global_step": 1131, "acc_step": 0, "speed/wps": 12832.436533889573, "speed/FLOPS": 201551049380404.2, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25389012694358826, "optim/lr": 0.000678, "optim/total_tokens": 296484864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469548, "loss/out": 4.229714393615723, "created_at": "2025-01-14T14:34:54.889303+00:00"} {"global_step": 1132, "acc_step": 0, "speed/wps": 12835.977230173035, "speed/FLOPS": 201606660880962.5, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25767797231674194, "optim/lr": 0.0006786, "optim/total_tokens": 296747008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481951, "loss/out": 4.131605625152588, "created_at": "2025-01-14T14:35:00.001649+00:00"} {"global_step": 1133, "acc_step": 0, "speed/wps": 12834.056444249385, "speed/FLOPS": 201576492298595.12, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2653448283672333, "optim/lr": 0.0006792, "optim/total_tokens": 297009152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485422, "loss/out": 4.172815322875977, "created_at": "2025-01-14T14:35:05.112257+00:00"} {"global_step": 1134, "acc_step": 0, "speed/wps": 12833.295636047167, "speed/FLOPS": 201564542760319.25, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22974000871181488, "optim/lr": 0.0006798, "optim/total_tokens": 297271296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495671, "loss/out": 4.265365123748779, "created_at": "2025-01-14T14:35:10.223650+00:00"} {"global_step": 1135, "acc_step": 0, "speed/wps": 12834.378588222777, "speed/FLOPS": 201581552012369.44, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22729086875915527, "optim/lr": 0.0006804000000000001, "optim/total_tokens": 297533440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296535, "loss/out": 4.312915802001953, "created_at": "2025-01-14T14:35:15.334824+00:00"} {"global_step": 1136, "acc_step": 0, "speed/wps": 12831.888058540855, "speed/FLOPS": 201542434821368.44, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23580840229988098, "optim/lr": 0.0006810000000000001, "optim/total_tokens": 297795584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 4.2115583419799805, "created_at": "2025-01-14T14:35:20.450542+00:00"} {"global_step": 1137, "acc_step": 0, "speed/wps": 12837.264242082581, "speed/FLOPS": 201626875171541.62, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.26643747091293335, "optim/lr": 0.0006816000000000001, "optim/total_tokens": 298057728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 4.099208354949951, "created_at": "2025-01-14T14:35:25.560100+00:00"} {"global_step": 1138, "acc_step": 0, "speed/wps": 12836.009192885183, "speed/FLOPS": 201607162899278.16, "speed/curr_iter_time": 1.2878, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.28343746066093445, "optim/lr": 0.0006822, "optim/total_tokens": 298319872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321540, "loss/out": 4.266324996948242, "created_at": "2025-01-14T14:35:30.681792+00:00"} {"global_step": 1139, "acc_step": 0, "speed/wps": 12834.371762365865, "speed/FLOPS": 201581444802906.94, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.30024948716163635, "optim/lr": 0.0006828, "optim/total_tokens": 298582016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 4.454287528991699, "created_at": "2025-01-14T14:35:35.791602+00:00"} {"global_step": 1140, "acc_step": 0, "speed/wps": 12832.488215686135, "speed/FLOPS": 201551861114037.4, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.27872371673583984, "optim/lr": 0.0006834, "optim/total_tokens": 298844160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 4.258761405944824, "created_at": "2025-01-14T14:35:40.909851+00:00"} {"global_step": 1141, "acc_step": 0, "speed/wps": 12836.620036498893, "speed/FLOPS": 201616757037618.53, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.29005223512649536, "optim/lr": 0.000684, "optim/total_tokens": 299106304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 4.232899188995361, "created_at": "2025-01-14T14:35:46.020248+00:00"} {"global_step": 1142, "acc_step": 0, "speed/wps": 12838.638146939784, "speed/FLOPS": 201648454235267.1, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.29079174995422363, "optim/lr": 0.0006845999999999999, "optim/total_tokens": 299368448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 4.148184299468994, "created_at": "2025-01-14T14:35:51.133808+00:00"} {"global_step": 1143, "acc_step": 0, "speed/wps": 12830.876674350911, "speed/FLOPS": 201526549642877.88, "speed/curr_iter_time": 1.2884, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3028866648674011, "optim/lr": 0.0006852, "optim/total_tokens": 299630592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466361, "loss/out": 4.234348297119141, "created_at": "2025-01-14T14:35:56.248139+00:00"} {"global_step": 1144, "acc_step": 0, "speed/wps": 12836.940014253096, "speed/FLOPS": 201621782727943.8, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.27740687131881714, "optim/lr": 0.0006858, "optim/total_tokens": 299892736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472404, "loss/out": 4.227123737335205, "created_at": "2025-01-14T14:36:01.359389+00:00"} {"global_step": 1145, "acc_step": 0, "speed/wps": 12832.279218794698, "speed/FLOPS": 201548578530665.56, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3100893497467041, "optim/lr": 0.0006864, "optim/total_tokens": 300154880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 403424, "loss/out": 4.239360332489014, "created_at": "2025-01-14T14:36:06.472961+00:00"} {"global_step": 1146, "acc_step": 0, "speed/wps": 12828.340433715382, "speed/FLOPS": 201486714498537.03, "speed/curr_iter_time": 1.2892, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3377194404602051, "optim/lr": 0.000687, "optim/total_tokens": 300417024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499682, "loss/out": 4.1725006103515625, "created_at": "2025-01-14T14:36:11.588509+00:00"} {"global_step": 1147, "acc_step": 0, "speed/wps": 12831.788982341432, "speed/FLOPS": 201540878693511.53, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.27578920125961304, "optim/lr": 0.0006876, "optim/total_tokens": 300679168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496999, "loss/out": 4.176952838897705, "created_at": "2025-01-14T14:36:16.701778+00:00"} {"global_step": 1148, "acc_step": 0, "speed/wps": 12832.38615828684, "speed/FLOPS": 201550258162339.9, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.26302313804626465, "optim/lr": 0.0006882, "optim/total_tokens": 300941312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288679, "loss/out": 4.146848678588867, "created_at": "2025-01-14T14:36:21.816708+00:00"} {"global_step": 1149, "acc_step": 0, "speed/wps": 12839.10148958398, "speed/FLOPS": 201655731668193.72, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2721782922744751, "optim/lr": 0.0006888, "optim/total_tokens": 301203456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495387, "loss/out": 4.220407962799072, "created_at": "2025-01-14T14:36:26.928263+00:00"} {"global_step": 1150, "acc_step": 0, "speed/wps": 12835.992039629058, "speed/FLOPS": 201606893483819.66, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.27762797474861145, "optim/lr": 0.0006894000000000001, "optim/total_tokens": 301465600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 4.085305213928223, "created_at": "2025-01-14T14:36:32.039539+00:00"} {"global_step": 1151, "acc_step": 0, "speed/wps": 12832.827722902957, "speed/FLOPS": 201557193541409.4, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2264261543750763, "optim/lr": 0.0006900000000000001, "optim/total_tokens": 301727744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468454, "loss/out": 4.160910129547119, "created_at": "2025-01-14T14:36:37.155066+00:00"} {"global_step": 1152, "acc_step": 0, "speed/wps": 12829.037604927536, "speed/FLOPS": 201497664530437.53, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2512263059616089, "optim/lr": 0.0006906, "optim/total_tokens": 301989888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465291, "loss/out": 4.220462799072266, "created_at": "2025-01-14T14:36:42.268356+00:00"} {"global_step": 1153, "acc_step": 0, "speed/wps": 12827.664874731796, "speed/FLOPS": 201476103916384.47, "speed/curr_iter_time": 1.2878, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2378026843070984, "optim/lr": 0.0006912, "optim/total_tokens": 302252032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493362, "loss/out": 4.105224132537842, "created_at": "2025-01-14T14:36:47.380882+00:00"} {"global_step": 1154, "acc_step": 0, "speed/wps": 12832.970071064494, "speed/FLOPS": 201559429314893.22, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24898578226566315, "optim/lr": 0.0006918, "optim/total_tokens": 302514176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502273, "loss/out": 4.230471611022949, "created_at": "2025-01-14T14:36:52.491488+00:00"} {"global_step": 1155, "acc_step": 0, "speed/wps": 12835.048912904924, "speed/FLOPS": 201592080382625.2, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22840973734855652, "optim/lr": 0.0006924, "optim/total_tokens": 302776320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448597, "loss/out": 4.083991050720215, "created_at": "2025-01-14T14:36:57.601877+00:00"} {"global_step": 1156, "acc_step": 0, "speed/wps": 12832.496574673733, "speed/FLOPS": 201551992403424.06, "speed/curr_iter_time": 1.2881, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.24376097321510315, "optim/lr": 0.000693, "optim/total_tokens": 303038464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484037, "loss/out": 4.133752346038818, "created_at": "2025-01-14T14:37:02.715406+00:00"} {"global_step": 1157, "acc_step": 0, "speed/wps": 12835.542041760644, "speed/FLOPS": 201599825648934.34, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22730720043182373, "optim/lr": 0.0006935999999999999, "optim/total_tokens": 303300608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495151, "loss/out": 4.146599292755127, "created_at": "2025-01-14T14:37:07.826541+00:00"} {"global_step": 1158, "acc_step": 0, "speed/wps": 12831.647598654074, "speed/FLOPS": 201538658068420.03, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.22944724559783936, "optim/lr": 0.0006942, "optim/total_tokens": 303562752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 285414, "loss/out": 4.343517303466797, "created_at": "2025-01-14T14:37:12.939485+00:00"} {"global_step": 1159, "acc_step": 0, "speed/wps": 12829.641598244867, "speed/FLOPS": 201507151075460.03, "speed/curr_iter_time": 1.2886, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.2644946575164795, "optim/lr": 0.0006948, "optim/total_tokens": 303824896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 506495, "loss/out": 4.287865161895752, "created_at": "2025-01-14T14:37:18.055878+00:00"} {"global_step": 1160, "acc_step": 0, "speed/wps": 12834.004246571121, "speed/FLOPS": 201575672462330.88, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3191869854927063, "optim/lr": 0.0006954, "optim/total_tokens": 304087040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488549, "loss/out": 4.195244789123535, "created_at": "2025-01-14T14:37:23.172700+00:00"} {"global_step": 1161, "acc_step": 0, "speed/wps": 12836.537409646995, "speed/FLOPS": 201615459269368.3, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.36559125781059265, "optim/lr": 0.000696, "optim/total_tokens": 304349184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304391, "loss/out": 4.189497947692871, "created_at": "2025-01-14T14:37:28.282056+00:00"} {"global_step": 1162, "acc_step": 0, "speed/wps": 12836.292893224312, "speed/FLOPS": 201611618802949.28, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3152409493923187, "optim/lr": 0.0006966, "optim/total_tokens": 304611328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481715, "loss/out": 4.216701984405518, "created_at": "2025-01-14T14:37:33.397399+00:00"} {"global_step": 1163, "acc_step": 0, "speed/wps": 12851.648502430733, "speed/FLOPS": 201852799746354.38, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.304429829120636, "optim/lr": 0.0006972, "optim/total_tokens": 304873472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 4.204524993896484, "created_at": "2025-01-14T14:37:38.500947+00:00"} {"global_step": 1164, "acc_step": 0, "speed/wps": 12839.644933392396, "speed/FLOPS": 201664267199974.94, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3313232958316803, "optim/lr": 0.0006978, "optim/total_tokens": 305135616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485422, "loss/out": 4.26689338684082, "created_at": "2025-01-14T14:37:43.608830+00:00"} {"global_step": 1165, "acc_step": 0, "speed/wps": 12834.467985146586, "speed/FLOPS": 201582956113902.12, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2566334903240204, "optim/lr": 0.0006984000000000001, "optim/total_tokens": 305397760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501228, "loss/out": 4.169419765472412, "created_at": "2025-01-14T14:37:48.719260+00:00"} {"global_step": 1166, "acc_step": 0, "speed/wps": 12839.891646266642, "speed/FLOPS": 201668142164683.97, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2724165916442871, "optim/lr": 0.0006990000000000001, "optim/total_tokens": 305659904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290218, "loss/out": 4.185037612915039, "created_at": "2025-01-14T14:37:53.828015+00:00"} {"global_step": 1167, "acc_step": 0, "speed/wps": 12844.677854206786, "speed/FLOPS": 201743316137322.88, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2968865931034088, "optim/lr": 0.0006996, "optim/total_tokens": 305922048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467691, "loss/out": 4.246690273284912, "created_at": "2025-01-14T14:37:58.940450+00:00"} {"global_step": 1168, "acc_step": 0, "speed/wps": 12839.877530661839, "speed/FLOPS": 201667920459713.7, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.32809948921203613, "optim/lr": 0.0007002, "optim/total_tokens": 306184192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307319, "loss/out": 4.144842147827148, "created_at": "2025-01-14T14:38:04.051508+00:00"} {"global_step": 1169, "acc_step": 0, "speed/wps": 12844.246061167274, "speed/FLOPS": 201736534234291.53, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3113659918308258, "optim/lr": 0.0007008, "optim/total_tokens": 306446336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473815, "loss/out": 4.2314558029174805, "created_at": "2025-01-14T14:38:09.164557+00:00"} {"global_step": 1170, "acc_step": 0, "speed/wps": 12836.193854360352, "speed/FLOPS": 201610063261497.12, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.34777605533599854, "optim/lr": 0.0007014, "optim/total_tokens": 306708480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456567, "loss/out": 4.140158653259277, "created_at": "2025-01-14T14:38:14.274032+00:00"} {"global_step": 1171, "acc_step": 0, "speed/wps": 12836.490251982588, "speed/FLOPS": 201614718593447.38, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.30585312843322754, "optim/lr": 0.000702, "optim/total_tokens": 306970624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475341, "loss/out": 4.237933158874512, "created_at": "2025-01-14T14:38:19.384749+00:00"} {"global_step": 1172, "acc_step": 0, "speed/wps": 12842.00288421896, "speed/FLOPS": 201701302057870.1, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3401503264904022, "optim/lr": 0.0007026, "optim/total_tokens": 307232768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496480, "loss/out": 4.186361312866211, "created_at": "2025-01-14T14:38:24.492261+00:00"} {"global_step": 1173, "acc_step": 0, "speed/wps": 12840.718120417354, "speed/FLOPS": 201681123076915.4, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2971452474594116, "optim/lr": 0.0007032, "optim/total_tokens": 307494912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306022, "loss/out": 4.114754676818848, "created_at": "2025-01-14T14:38:29.600400+00:00"} {"global_step": 1174, "acc_step": 0, "speed/wps": 12840.32941879213, "speed/FLOPS": 201675017983759.25, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4131370484828949, "optim/lr": 0.0007038, "optim/total_tokens": 307757056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480904, "loss/out": 4.228046894073486, "created_at": "2025-01-14T14:38:34.713020+00:00"} {"global_step": 1175, "acc_step": 0, "speed/wps": 12845.79756550362, "speed/FLOPS": 201760902741884.22, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.36987683176994324, "optim/lr": 0.0007044, "optim/total_tokens": 308019200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471956, "loss/out": 4.31910514831543, "created_at": "2025-01-14T14:38:39.820620+00:00"} {"global_step": 1176, "acc_step": 0, "speed/wps": 12839.193203539566, "speed/FLOPS": 201657172161894.38, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2965953052043915, "optim/lr": 0.000705, "optim/total_tokens": 308281344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480621, "loss/out": 4.148116111755371, "created_at": "2025-01-14T14:38:44.931667+00:00"} {"global_step": 1177, "acc_step": 0, "speed/wps": 12844.966200519391, "speed/FLOPS": 201747845012391.7, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2719250023365021, "optim/lr": 0.0007056, "optim/total_tokens": 308543488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490923, "loss/out": 4.223272800445557, "created_at": "2025-01-14T14:38:50.043609+00:00"} {"global_step": 1178, "acc_step": 0, "speed/wps": 12841.104296083184, "speed/FLOPS": 201687188496408.25, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2955496311187744, "optim/lr": 0.0007062, "optim/total_tokens": 308805632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 340643, "loss/out": 4.2532572746276855, "created_at": "2025-01-14T14:38:55.166636+00:00"} {"global_step": 1179, "acc_step": 0, "speed/wps": 12842.201201391095, "speed/FLOPS": 201704416901574.84, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2812265455722809, "optim/lr": 0.0007068, "optim/total_tokens": 309067776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479054, "loss/out": 4.166440010070801, "created_at": "2025-01-14T14:39:00.277555+00:00"} {"global_step": 1180, "acc_step": 0, "speed/wps": 12842.922700652827, "speed/FLOPS": 201715749038924.22, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2820758819580078, "optim/lr": 0.0007074000000000001, "optim/total_tokens": 309329920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497808, "loss/out": 4.22792387008667, "created_at": "2025-01-14T14:39:05.387377+00:00"} {"global_step": 1181, "acc_step": 0, "speed/wps": 12843.956767257017, "speed/FLOPS": 201731990475899.44, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2977502942085266, "optim/lr": 0.000708, "optim/total_tokens": 309592064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483235, "loss/out": 4.194660663604736, "created_at": "2025-01-14T14:39:10.496830+00:00"} {"global_step": 1182, "acc_step": 0, "speed/wps": 12837.852278324079, "speed/FLOPS": 201636111088760.8, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.27846306562423706, "optim/lr": 0.0007086, "optim/total_tokens": 309854208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485941, "loss/out": 4.211625576019287, "created_at": "2025-01-14T14:39:15.608293+00:00"} {"global_step": 1183, "acc_step": 0, "speed/wps": 12838.880686658991, "speed/FLOPS": 201652263654841.75, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2825428247451782, "optim/lr": 0.0007092, "optim/total_tokens": 310116352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317421, "loss/out": 4.245269298553467, "created_at": "2025-01-14T14:39:20.719742+00:00"} {"global_step": 1184, "acc_step": 0, "speed/wps": 12834.958978812727, "speed/FLOPS": 201590667844124.84, "speed/curr_iter_time": 1.2877, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.28518420457839966, "optim/lr": 0.0007098, "optim/total_tokens": 310378496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466597, "loss/out": 4.199513912200928, "created_at": "2025-01-14T14:39:25.830814+00:00"} {"global_step": 1185, "acc_step": 0, "speed/wps": 12848.543065306381, "speed/FLOPS": 201804024588998.66, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24351471662521362, "optim/lr": 0.0007104, "optim/total_tokens": 310640640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425730, "loss/out": 4.1335225105285645, "created_at": "2025-01-14T14:39:30.940008+00:00"} {"global_step": 1186, "acc_step": 0, "speed/wps": 12844.673415902267, "speed/FLOPS": 201743246427651.53, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2446274757385254, "optim/lr": 0.0007109999999999999, "optim/total_tokens": 310902784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497525, "loss/out": 4.154821395874023, "created_at": "2025-01-14T14:39:36.050008+00:00"} {"global_step": 1187, "acc_step": 0, "speed/wps": 12838.501950459495, "speed/FLOPS": 201646315082390.56, "speed/curr_iter_time": 1.2875, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24316668510437012, "optim/lr": 0.0007116, "optim/total_tokens": 311164928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487982, "loss/out": 4.140872955322266, "created_at": "2025-01-14T14:39:41.158554+00:00"} {"global_step": 1188, "acc_step": 0, "speed/wps": 12840.640284016426, "speed/FLOPS": 201679900549280.4, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.282733678817749, "optim/lr": 0.0007122, "optim/total_tokens": 311427072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320877, "loss/out": 4.185974597930908, "created_at": "2025-01-14T14:39:46.272827+00:00"} {"global_step": 1189, "acc_step": 0, "speed/wps": 12841.588036099562, "speed/FLOPS": 201694786298093.9, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.365829199552536, "optim/lr": 0.0007128, "optim/total_tokens": 311689216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479338, "loss/out": 4.1845316886901855, "created_at": "2025-01-14T14:39:51.384616+00:00"} {"global_step": 1190, "acc_step": 0, "speed/wps": 12836.109413679013, "speed/FLOPS": 201608737004563.03, "speed/curr_iter_time": 1.2878, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.31917300820350647, "optim/lr": 0.0007134000000000001, "optim/total_tokens": 311951360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460513, "loss/out": 4.254024505615234, "created_at": "2025-01-14T14:39:56.495481+00:00"} {"global_step": 1191, "acc_step": 0, "speed/wps": 12842.665731764751, "speed/FLOPS": 201711712989424.72, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3631565272808075, "optim/lr": 0.000714, "optim/total_tokens": 312213504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485658, "loss/out": 4.195870876312256, "created_at": "2025-01-14T14:40:01.602178+00:00"} {"global_step": 1192, "acc_step": 0, "speed/wps": 12838.358701911064, "speed/FLOPS": 201644065167093.34, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3558043837547302, "optim/lr": 0.0007146, "optim/total_tokens": 312475648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496542, "loss/out": 4.323206901550293, "created_at": "2025-01-14T14:40:06.712138+00:00"} {"global_step": 1193, "acc_step": 0, "speed/wps": 12839.478675155771, "speed/FLOPS": 201661655885905.97, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.35377687215805054, "optim/lr": 0.0007152, "optim/total_tokens": 312737792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318716, "loss/out": 4.165643692016602, "created_at": "2025-01-14T14:40:11.820195+00:00"} {"global_step": 1194, "acc_step": 0, "speed/wps": 12839.917375404071, "speed/FLOPS": 201668546276142.84, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2980208992958069, "optim/lr": 0.0007158, "optim/total_tokens": 312999936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494106, "loss/out": 4.1244330406188965, "created_at": "2025-01-14T14:40:16.951585+00:00"} {"global_step": 1195, "acc_step": 0, "speed/wps": 12842.039437640433, "speed/FLOPS": 201701876179584.0, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2698635756969452, "optim/lr": 0.0007164000000000001, "optim/total_tokens": 313262080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372878, "loss/out": 4.1384687423706055, "created_at": "2025-01-14T14:40:22.058366+00:00"} {"global_step": 1196, "acc_step": 0, "speed/wps": 12838.369361646028, "speed/FLOPS": 201644232592878.97, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.269845187664032, "optim/lr": 0.000717, "optim/total_tokens": 313524224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503075, "loss/out": 4.194545745849609, "created_at": "2025-01-14T14:40:27.169823+00:00"} {"global_step": 1197, "acc_step": 0, "speed/wps": 12836.646232566127, "speed/FLOPS": 201617168482851.62, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2515663504600525, "optim/lr": 0.0007176, "optim/total_tokens": 313786368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494118, "loss/out": 4.194518089294434, "created_at": "2025-01-14T14:40:32.281014+00:00"} {"global_step": 1198, "acc_step": 0, "speed/wps": 12845.847314814006, "speed/FLOPS": 201761684123166.53, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.27055588364601135, "optim/lr": 0.0007182, "optim/total_tokens": 314048512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317660, "loss/out": 4.108572006225586, "created_at": "2025-01-14T14:40:37.391651+00:00"} {"global_step": 1199, "acc_step": 0, "speed/wps": 12840.83429919283, "speed/FLOPS": 201682947824230.62, "speed/curr_iter_time": 1.2876, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2447109967470169, "optim/lr": 0.0007188, "optim/total_tokens": 314310656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484895, "loss/out": 4.083837985992432, "created_at": "2025-01-14T14:40:42.501662+00:00"} {"global_step": 1200, "acc_step": 0, "speed/wps": 12844.184297432728, "speed/FLOPS": 201735564149968.22, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.230894535779953, "optim/lr": 0.0007194, "optim/total_tokens": 314572800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 4.1081109046936035, "created_at": "2025-01-14T14:40:47.610364+00:00"} {"global_step": 1201, "acc_step": 0, "speed/wps": 12834.43856993332, "speed/FLOPS": 201582494107557.03, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2693174481391907, "optim/lr": 0.0007199999999999999, "optim/total_tokens": 314834944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491685, "loss/out": 4.106167316436768, "created_at": "2025-01-14T14:40:52.722899+00:00"} {"global_step": 1202, "acc_step": 0, "speed/wps": 12844.302707556035, "speed/FLOPS": 201737423943667.3, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.27586236596107483, "optim/lr": 0.0007206, "optim/total_tokens": 315097088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493816, "loss/out": 4.141815185546875, "created_at": "2025-01-14T14:40:57.856011+00:00"} {"global_step": 1203, "acc_step": 0, "speed/wps": 12846.135788679196, "speed/FLOPS": 201766215001623.97, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25351306796073914, "optim/lr": 0.0007212000000000001, "optim/total_tokens": 315359232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467407, "loss/out": 4.249550819396973, "created_at": "2025-01-14T14:41:02.962769+00:00"} {"global_step": 1204, "acc_step": 0, "speed/wps": 12838.867179976012, "speed/FLOPS": 201652051513825.44, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2500813901424408, "optim/lr": 0.0007218000000000001, "optim/total_tokens": 315621376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470587, "loss/out": 4.197220325469971, "created_at": "2025-01-14T14:41:08.072164+00:00"} {"global_step": 1205, "acc_step": 0, "speed/wps": 12839.022182935172, "speed/FLOPS": 201654486048295.44, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.28471410274505615, "optim/lr": 0.0007224, "optim/total_tokens": 315883520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498564, "loss/out": 4.164783000946045, "created_at": "2025-01-14T14:41:13.180783+00:00"} {"global_step": 1206, "acc_step": 0, "speed/wps": 12841.669464843262, "speed/FLOPS": 201696065248408.53, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2836779057979584, "optim/lr": 0.000723, "optim/total_tokens": 316145664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328075, "loss/out": 4.078447341918945, "created_at": "2025-01-14T14:41:18.290423+00:00"} {"global_step": 1207, "acc_step": 0, "speed/wps": 12840.192530400123, "speed/FLOPS": 201672867963458.44, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3090285658836365, "optim/lr": 0.0007236, "optim/total_tokens": 316407808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500418, "loss/out": 4.221528053283691, "created_at": "2025-01-14T14:41:23.399084+00:00"} {"global_step": 1208, "acc_step": 0, "speed/wps": 12837.967927450141, "speed/FLOPS": 201637927517204.16, "speed/curr_iter_time": 1.2875, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.27647003531455994, "optim/lr": 0.0007242, "optim/total_tokens": 316669952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323038, "loss/out": 4.00012731552124, "created_at": "2025-01-14T14:41:28.509727+00:00"} {"global_step": 1209, "acc_step": 0, "speed/wps": 12820.638194126419, "speed/FLOPS": 201365740241806.03, "speed/curr_iter_time": 1.2876, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.27598193287849426, "optim/lr": 0.0007248, "optim/total_tokens": 316932096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488318, "loss/out": 4.134312152862549, "created_at": "2025-01-14T14:41:33.625179+00:00"} {"global_step": 1210, "acc_step": 0, "speed/wps": 12838.059034983988, "speed/FLOPS": 201639358486218.1, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.25139695405960083, "optim/lr": 0.0007254, "optim/total_tokens": 317194240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476914, "loss/out": 4.146534442901611, "created_at": "2025-01-14T14:41:38.737749+00:00"} {"global_step": 1211, "acc_step": 0, "speed/wps": 12837.612496531245, "speed/FLOPS": 201632344986209.62, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2637876868247986, "optim/lr": 0.000726, "optim/total_tokens": 317456384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474011, "loss/out": 4.19441032409668, "created_at": "2025-01-14T14:41:43.847421+00:00"} {"global_step": 1212, "acc_step": 0, "speed/wps": 12834.072962679329, "speed/FLOPS": 201576751743236.1, "speed/curr_iter_time": 1.2883, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.240087628364563, "optim/lr": 0.0007266, "optim/total_tokens": 317718528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486468, "loss/out": 4.05484676361084, "created_at": "2025-01-14T14:41:48.960160+00:00"} {"global_step": 1213, "acc_step": 0, "speed/wps": 12835.077998481836, "speed/FLOPS": 201592537211578.38, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.238872691988945, "optim/lr": 0.0007272, "optim/total_tokens": 317980672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411640, "loss/out": 4.106629848480225, "created_at": "2025-01-14T14:41:54.070749+00:00"} {"global_step": 1214, "acc_step": 0, "speed/wps": 12842.279813736228, "speed/FLOPS": 201705651616479.8, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3048771321773529, "optim/lr": 0.0007278, "optim/total_tokens": 318242816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497525, "loss/out": 4.212069511413574, "created_at": "2025-01-14T14:41:59.178316+00:00"} {"global_step": 1215, "acc_step": 0, "speed/wps": 12840.429939021118, "speed/FLOPS": 201676596792084.78, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.35348713397979736, "optim/lr": 0.0007284, "optim/total_tokens": 318504960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490175, "loss/out": 4.137969493865967, "created_at": "2025-01-14T14:42:04.287665+00:00"} {"global_step": 1216, "acc_step": 0, "speed/wps": 12846.38845227191, "speed/FLOPS": 201770183430543.72, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.34248289465904236, "optim/lr": 0.000729, "optim/total_tokens": 318767104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493598, "loss/out": 4.183494567871094, "created_at": "2025-01-14T14:42:09.394653+00:00"} {"global_step": 1217, "acc_step": 0, "speed/wps": 12834.056570647865, "speed/FLOPS": 201576494283856.94, "speed/curr_iter_time": 1.2879, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2648186683654785, "optim/lr": 0.0007296000000000001, "optim/total_tokens": 319029248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475341, "loss/out": 4.08425760269165, "created_at": "2025-01-14T14:42:14.507893+00:00"} {"global_step": 1218, "acc_step": 0, "speed/wps": 12841.06136960521, "speed/FLOPS": 201686514277084.5, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2825905978679657, "optim/lr": 0.0007302000000000001, "optim/total_tokens": 319291392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 393825, "loss/out": 4.058852195739746, "created_at": "2025-01-14T14:42:19.615452+00:00"} {"global_step": 1219, "acc_step": 0, "speed/wps": 12843.482403525197, "speed/FLOPS": 201724539941647.12, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.29617494344711304, "optim/lr": 0.0007308000000000001, "optim/total_tokens": 319553536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505002, "loss/out": 4.13262939453125, "created_at": "2025-01-14T14:42:24.722764+00:00"} {"global_step": 1220, "acc_step": 0, "speed/wps": 12847.230171545429, "speed/FLOPS": 201783403788377.8, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.33243998885154724, "optim/lr": 0.0007314, "optim/total_tokens": 319815680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 392890, "loss/out": 4.206699371337891, "created_at": "2025-01-14T14:42:29.831549+00:00"} {"global_step": 1221, "acc_step": 0, "speed/wps": 12842.633283531844, "speed/FLOPS": 201711203345338.38, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.32362157106399536, "optim/lr": 0.000732, "optim/total_tokens": 320077824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495739, "loss/out": 4.205981254577637, "created_at": "2025-01-14T14:42:34.938555+00:00"} {"global_step": 1222, "acc_step": 0, "speed/wps": 12842.139961208377, "speed/FLOPS": 201703455040352.4, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.29996761679649353, "optim/lr": 0.0007326, "optim/total_tokens": 320339968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423105, "loss/out": 4.264058589935303, "created_at": "2025-01-14T14:42:40.047863+00:00"} {"global_step": 1223, "acc_step": 0, "speed/wps": 12842.07318124775, "speed/FLOPS": 201702406169305.47, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.3086898922920227, "optim/lr": 0.0007332, "optim/total_tokens": 320602112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503311, "loss/out": 4.130751132965088, "created_at": "2025-01-14T14:42:45.158382+00:00"} {"global_step": 1224, "acc_step": 0, "speed/wps": 12835.965413071139, "speed/FLOPS": 201606475277138.8, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2867456078529358, "optim/lr": 0.0007338000000000001, "optim/total_tokens": 320864256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496190, "loss/out": 4.199334144592285, "created_at": "2025-01-14T14:42:50.269193+00:00"} {"global_step": 1225, "acc_step": 0, "speed/wps": 12841.84841626195, "speed/FLOPS": 201698875926344.03, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.27813196182250977, "optim/lr": 0.0007344, "optim/total_tokens": 321126400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303047, "loss/out": 4.1302595138549805, "created_at": "2025-01-14T14:42:55.379163+00:00"} {"global_step": 1226, "acc_step": 0, "speed/wps": 12844.52886905791, "speed/FLOPS": 201740976120832.72, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2751787602901459, "optim/lr": 0.000735, "optim/total_tokens": 321388544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478811, "loss/out": 4.155167102813721, "created_at": "2025-01-14T14:43:00.487891+00:00"} {"global_step": 1227, "acc_step": 0, "speed/wps": 12839.940187784172, "speed/FLOPS": 201668904575919.8, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3113340139389038, "optim/lr": 0.0007356, "optim/total_tokens": 321650688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470839, "loss/out": 4.217464923858643, "created_at": "2025-01-14T14:43:05.599444+00:00"} {"global_step": 1228, "acc_step": 0, "speed/wps": 12841.472047562114, "speed/FLOPS": 201692964538728.34, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2360697090625763, "optim/lr": 0.0007362, "optim/total_tokens": 321912832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491741, "loss/out": 4.076966285705566, "created_at": "2025-01-14T14:43:10.710400+00:00"} {"global_step": 1229, "acc_step": 0, "speed/wps": 12844.60314901753, "speed/FLOPS": 201742142789669.9, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22944673895835876, "optim/lr": 0.0007368, "optim/total_tokens": 322174976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493297, "loss/out": 4.06364631652832, "created_at": "2025-01-14T14:43:15.820933+00:00"} {"global_step": 1230, "acc_step": 0, "speed/wps": 12841.89608904875, "speed/FLOPS": 201699624692970.03, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22720767557621002, "optim/lr": 0.0007374, "optim/total_tokens": 322437120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289212, "loss/out": 4.207949161529541, "created_at": "2025-01-14T14:43:20.927782+00:00"} {"global_step": 1231, "acc_step": 0, "speed/wps": 12840.707756703507, "speed/FLOPS": 201680960300548.2, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2518357038497925, "optim/lr": 0.000738, "optim/total_tokens": 322699264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492025, "loss/out": 4.140360355377197, "created_at": "2025-01-14T14:43:26.041145+00:00"} {"global_step": 1232, "acc_step": 0, "speed/wps": 12837.836227950826, "speed/FLOPS": 201635858995593.34, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23683986067771912, "optim/lr": 0.0007386000000000001, "optim/total_tokens": 322961408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473524, "loss/out": 4.085765838623047, "created_at": "2025-01-14T14:43:31.150020+00:00"} {"global_step": 1233, "acc_step": 0, "speed/wps": 12844.759950498234, "speed/FLOPS": 201744605572392.2, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2519865334033966, "optim/lr": 0.0007392000000000001, "optim/total_tokens": 323223552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488785, "loss/out": 4.318150520324707, "created_at": "2025-01-14T14:43:36.257785+00:00"} {"global_step": 1234, "acc_step": 0, "speed/wps": 12839.456295469152, "speed/FLOPS": 201661304382174.72, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2516918182373047, "optim/lr": 0.0007398000000000001, "optim/total_tokens": 323485696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 386832, "loss/out": 4.049402713775635, "created_at": "2025-01-14T14:43:41.368076+00:00"} {"global_step": 1235, "acc_step": 0, "speed/wps": 12843.357653649844, "speed/FLOPS": 201722580573430.4, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2932405471801758, "optim/lr": 0.0007404, "optim/total_tokens": 323747840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495151, "loss/out": 4.164251804351807, "created_at": "2025-01-14T14:43:46.474484+00:00"} {"global_step": 1236, "acc_step": 0, "speed/wps": 12840.068027172589, "speed/FLOPS": 201670912469184.62, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.28518441319465637, "optim/lr": 0.000741, "optim/total_tokens": 324009984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470830, "loss/out": 4.206303596496582, "created_at": "2025-01-14T14:43:51.582748+00:00"} {"global_step": 1237, "acc_step": 0, "speed/wps": 12845.585972847615, "speed/FLOPS": 201757579388462.88, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24660034477710724, "optim/lr": 0.0007416, "optim/total_tokens": 324272128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479054, "loss/out": 4.219311714172363, "created_at": "2025-01-14T14:43:56.688109+00:00"} {"global_step": 1238, "acc_step": 0, "speed/wps": 12842.577640959944, "speed/FLOPS": 201710329402285.78, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2576132118701935, "optim/lr": 0.0007422, "optim/total_tokens": 324534272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500702, "loss/out": 4.168554306030273, "created_at": "2025-01-14T14:44:01.794822+00:00"} {"global_step": 1239, "acc_step": 0, "speed/wps": 12843.171735872393, "speed/FLOPS": 201719660479257.75, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.28290051221847534, "optim/lr": 0.0007428, "optim/total_tokens": 324796416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301170, "loss/out": 4.162154197692871, "created_at": "2025-01-14T14:44:06.903995+00:00"} {"global_step": 1240, "acc_step": 0, "speed/wps": 12844.31403830042, "speed/FLOPS": 201737601908577.94, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.28036409616470337, "optim/lr": 0.0007434, "optim/total_tokens": 325058560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484328, "loss/out": 4.151783466339111, "created_at": "2025-01-14T14:44:12.009915+00:00"} {"global_step": 1241, "acc_step": 0, "speed/wps": 12830.835102970948, "speed/FLOPS": 201525896707230.47, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25438401103019714, "optim/lr": 0.000744, "optim/total_tokens": 325320704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470067, "loss/out": 4.10375452041626, "created_at": "2025-01-14T14:44:17.122706+00:00"} {"global_step": 1242, "acc_step": 0, "speed/wps": 12842.31163909591, "speed/FLOPS": 201706151477486.25, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2559124827384949, "optim/lr": 0.0007446, "optim/total_tokens": 325582848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480904, "loss/out": 4.095188140869141, "created_at": "2025-01-14T14:44:22.232014+00:00"} {"global_step": 1243, "acc_step": 0, "speed/wps": 12841.417120415024, "speed/FLOPS": 201692101832405.47, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2689483165740967, "optim/lr": 0.0007452, "optim/total_tokens": 325844992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492487, "loss/out": 4.111169815063477, "created_at": "2025-01-14T14:44:27.339104+00:00"} {"global_step": 1244, "acc_step": 0, "speed/wps": 12843.40070331137, "speed/FLOPS": 201723256727521.12, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2763098478317261, "optim/lr": 0.0007458, "optim/total_tokens": 326107136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300113, "loss/out": 4.141256332397461, "created_at": "2025-01-14T14:44:32.445673+00:00"} {"global_step": 1245, "acc_step": 0, "speed/wps": 12838.688426411998, "speed/FLOPS": 201649243943469.47, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.26018226146698, "optim/lr": 0.0007464, "optim/total_tokens": 326369280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481951, "loss/out": 4.073925495147705, "created_at": "2025-01-14T14:44:37.555902+00:00"} {"global_step": 1246, "acc_step": 0, "speed/wps": 12837.68241527522, "speed/FLOPS": 201633443156162.2, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22009415924549103, "optim/lr": 0.000747, "optim/total_tokens": 326631424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349591, "loss/out": 4.18433141708374, "created_at": "2025-01-14T14:44:42.664788+00:00"} {"global_step": 1247, "acc_step": 0, "speed/wps": 12842.25755787476, "speed/FLOPS": 201705302057592.72, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2528358995914459, "optim/lr": 0.0007476000000000001, "optim/total_tokens": 326893568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480621, "loss/out": 4.176400661468506, "created_at": "2025-01-14T14:44:47.777990+00:00"} {"global_step": 1248, "acc_step": 0, "speed/wps": 12843.154966641365, "speed/FLOPS": 201719397095441.12, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2874998450279236, "optim/lr": 0.0007482000000000001, "optim/total_tokens": 327155712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361049, "loss/out": 4.186230659484863, "created_at": "2025-01-14T14:44:52.884646+00:00"} {"global_step": 1249, "acc_step": 0, "speed/wps": 12841.292762159728, "speed/FLOPS": 201690148615119.9, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.24641436338424683, "optim/lr": 0.0007488, "optim/total_tokens": 327417856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496480, "loss/out": 4.287100791931152, "created_at": "2025-01-14T14:44:57.991733+00:00"} {"global_step": 1250, "acc_step": 0, "speed/wps": 12841.574401859949, "speed/FLOPS": 201694572153625.38, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24668821692466736, "optim/lr": 0.0007494, "optim/total_tokens": 327680000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476914, "loss/out": 4.085331439971924, "created_at": "2025-01-14T14:45:03.099071+00:00"} {"global_step": 1251, "acc_step": 0, "speed/wps": 12844.614046547416, "speed/FLOPS": 201742313950351.56, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.31008633971214294, "optim/lr": 0.00075, "optim/total_tokens": 327942144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470067, "loss/out": 4.102134704589844, "created_at": "2025-01-14T14:45:08.205264+00:00"} {"global_step": 1252, "acc_step": 0, "speed/wps": 12840.451342708848, "speed/FLOPS": 201676932966412.34, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2718829810619354, "optim/lr": 0.0007505999999999999, "optim/total_tokens": 328204288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495387, "loss/out": 4.118716716766357, "created_at": "2025-01-14T14:45:13.319613+00:00"} {"global_step": 1253, "acc_step": 0, "speed/wps": 12840.611726197283, "speed/FLOPS": 201679452009488.22, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22203843295574188, "optim/lr": 0.0007512, "optim/total_tokens": 328466432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327836, "loss/out": 4.058236598968506, "created_at": "2025-01-14T14:45:18.430205+00:00"} {"global_step": 1254, "acc_step": 0, "speed/wps": 12841.085144450777, "speed/FLOPS": 201686887693704.2, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2324933111667633, "optim/lr": 0.0007518, "optim/total_tokens": 328728576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484895, "loss/out": 4.134615421295166, "created_at": "2025-01-14T14:45:23.538769+00:00"} {"global_step": 1255, "acc_step": 0, "speed/wps": 12832.282899716984, "speed/FLOPS": 201548636344608.2, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2451556921005249, "optim/lr": 0.0007524000000000001, "optim/total_tokens": 328990720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493598, "loss/out": 4.162521839141846, "created_at": "2025-01-14T14:45:28.649704+00:00"} {"global_step": 1256, "acc_step": 0, "speed/wps": 12841.802182761769, "speed/FLOPS": 201698149765692.22, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.28880399465560913, "optim/lr": 0.000753, "optim/total_tokens": 329252864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475620, "loss/out": 4.161846160888672, "created_at": "2025-01-14T14:45:33.759387+00:00"} {"global_step": 1257, "acc_step": 0, "speed/wps": 12848.860564341143, "speed/FLOPS": 201809011347628.97, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.37545517086982727, "optim/lr": 0.0007536, "optim/total_tokens": 329515008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496999, "loss/out": 4.196111679077148, "created_at": "2025-01-14T14:45:38.891321+00:00"} {"global_step": 1258, "acc_step": 0, "speed/wps": 12841.473059918908, "speed/FLOPS": 201692980439182.9, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31154510378837585, "optim/lr": 0.0007542, "optim/total_tokens": 329777152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472161, "loss/out": 4.140561103820801, "created_at": "2025-01-14T14:45:44.001207+00:00"} {"global_step": 1259, "acc_step": 0, "speed/wps": 12836.455498043024, "speed/FLOPS": 201614172735069.6, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.25855588912963867, "optim/lr": 0.0007548, "optim/total_tokens": 330039296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482235, "loss/out": 3.9712209701538086, "created_at": "2025-01-14T14:45:49.113634+00:00"} {"global_step": 1260, "acc_step": 0, "speed/wps": 12842.642923761903, "speed/FLOPS": 201711354758398.53, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.27662625908851624, "optim/lr": 0.0007554, "optim/total_tokens": 330301440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487798, "loss/out": 4.2001142501831055, "created_at": "2025-01-14T14:45:54.222539+00:00"} {"global_step": 1261, "acc_step": 0, "speed/wps": 12842.00545984914, "speed/FLOPS": 201701342511681.38, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.33728352189064026, "optim/lr": 0.000756, "optim/total_tokens": 330563584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484320, "loss/out": 4.167404651641846, "created_at": "2025-01-14T14:45:59.329365+00:00"} {"global_step": 1262, "acc_step": 0, "speed/wps": 12842.029411618758, "speed/FLOPS": 201701718707136.4, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.3905685245990753, "optim/lr": 0.0007566, "optim/total_tokens": 330825728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493882, "loss/out": 4.171884536743164, "created_at": "2025-01-14T14:46:04.439514+00:00"} {"global_step": 1263, "acc_step": 0, "speed/wps": 12842.753964411217, "speed/FLOPS": 201713098804381.84, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2928973138332367, "optim/lr": 0.0007572000000000001, "optim/total_tokens": 331087872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324333, "loss/out": 4.149930953979492, "created_at": "2025-01-14T14:46:09.550646+00:00"} {"global_step": 1264, "acc_step": 0, "speed/wps": 12841.496883157515, "speed/FLOPS": 201693354615881.78, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2956469655036926, "optim/lr": 0.0007578, "optim/total_tokens": 331350016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477718, "loss/out": 4.110779285430908, "created_at": "2025-01-14T14:46:14.659026+00:00"} {"global_step": 1265, "acc_step": 0, "speed/wps": 12844.672231885996, "speed/FLOPS": 201743227831048.97, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.258783757686615, "optim/lr": 0.0007584000000000001, "optim/total_tokens": 331612160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420458, "loss/out": 4.247766971588135, "created_at": "2025-01-14T14:46:19.766761+00:00"} {"global_step": 1266, "acc_step": 0, "speed/wps": 12837.457984288656, "speed/FLOPS": 201629918159117.72, "speed/curr_iter_time": 1.2879, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25164708495140076, "optim/lr": 0.000759, "optim/total_tokens": 331874304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494342, "loss/out": 4.169016361236572, "created_at": "2025-01-14T14:46:24.877198+00:00"} {"global_step": 1267, "acc_step": 0, "speed/wps": 12842.571263423844, "speed/FLOPS": 201710229234317.7, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.26246780157089233, "optim/lr": 0.0007595999999999999, "optim/total_tokens": 332136448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495671, "loss/out": 4.1092529296875, "created_at": "2025-01-14T14:46:29.986510+00:00"} {"global_step": 1268, "acc_step": 0, "speed/wps": 12843.12412862269, "speed/FLOPS": 201718912741981.97, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23213569819927216, "optim/lr": 0.0007602, "optim/total_tokens": 332398592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300645, "loss/out": 4.199310302734375, "created_at": "2025-01-14T14:46:35.102206+00:00"} {"global_step": 1269, "acc_step": 0, "speed/wps": 12845.6793881069, "speed/FLOPS": 201759046603475.97, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23486538231372833, "optim/lr": 0.0007608, "optim/total_tokens": 332660736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491505, "loss/out": 4.074358940124512, "created_at": "2025-01-14T14:46:40.208190+00:00"} {"global_step": 1270, "acc_step": 0, "speed/wps": 12840.155083117897, "speed/FLOPS": 201672279802433.12, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3317265808582306, "optim/lr": 0.0007614000000000001, "optim/total_tokens": 332922880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478528, "loss/out": 3.9576048851013184, "created_at": "2025-01-14T14:46:45.318856+00:00"} {"global_step": 1271, "acc_step": 0, "speed/wps": 12839.76611804377, "speed/FLOPS": 201666170571449.9, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3023267686367035, "optim/lr": 0.000762, "optim/total_tokens": 333185024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478764, "loss/out": 4.1970133781433105, "created_at": "2025-01-14T14:46:50.426793+00:00"} {"global_step": 1272, "acc_step": 0, "speed/wps": 12842.936344787498, "speed/FLOPS": 201715963338808.25, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31633374094963074, "optim/lr": 0.0007626, "optim/total_tokens": 333447168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502273, "loss/out": 4.065725326538086, "created_at": "2025-01-14T14:46:55.533719+00:00"} {"global_step": 1273, "acc_step": 0, "speed/wps": 12846.470908146339, "speed/FLOPS": 201771478513357.84, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2977364957332611, "optim/lr": 0.0007632, "optim/total_tokens": 333709312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502030, "loss/out": 4.120262145996094, "created_at": "2025-01-14T14:47:00.642115+00:00"} {"global_step": 1274, "acc_step": 0, "speed/wps": 12841.532929858746, "speed/FLOPS": 201693920778857.97, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22744767367839813, "optim/lr": 0.0007638, "optim/total_tokens": 333971456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471114, "loss/out": 4.041078567504883, "created_at": "2025-01-14T14:47:05.749232+00:00"} {"global_step": 1275, "acc_step": 0, "speed/wps": 12840.532124407197, "speed/FLOPS": 201678201753990.6, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22256512939929962, "optim/lr": 0.0007644000000000001, "optim/total_tokens": 334233600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488318, "loss/out": 4.128570556640625, "created_at": "2025-01-14T14:47:10.856684+00:00"} {"global_step": 1276, "acc_step": 0, "speed/wps": 12836.004610718228, "speed/FLOPS": 201607090930049.56, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23383359611034393, "optim/lr": 0.0007650000000000001, "optim/total_tokens": 334495744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494861, "loss/out": 4.229945182800293, "created_at": "2025-01-14T14:47:15.968626+00:00"} {"global_step": 1277, "acc_step": 0, "speed/wps": 12836.613039659123, "speed/FLOPS": 201616647142635.3, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.24914653599262238, "optim/lr": 0.0007656, "optim/total_tokens": 334757888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305449, "loss/out": 4.146387100219727, "created_at": "2025-01-14T14:47:21.079074+00:00"} {"global_step": 1278, "acc_step": 0, "speed/wps": 12836.193980800937, "speed/FLOPS": 201610065247420.28, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.001, "optim/grad_norm": 0.23513874411582947, "optim/lr": 0.0007662000000000001, "optim/total_tokens": 335020032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466077, "loss/out": 4.0076727867126465, "created_at": "2025-01-14T14:47:26.192985+00:00"} {"global_step": 1279, "acc_step": 0, "speed/wps": 12842.872397386001, "speed/FLOPS": 201714958956994.78, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2186981588602066, "optim/lr": 0.0007668, "optim/total_tokens": 335282176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482522, "loss/out": 4.109147548675537, "created_at": "2025-01-14T14:47:31.302593+00:00"} {"global_step": 1280, "acc_step": 0, "speed/wps": 12840.408064455696, "speed/FLOPS": 201676253221976.03, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.26032623648643494, "optim/lr": 0.0007674000000000001, "optim/total_tokens": 335544320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486694, "loss/out": 4.123739242553711, "created_at": "2025-01-14T14:47:36.410121+00:00"} {"global_step": 1281, "acc_step": 0, "speed/wps": 12839.654538680792, "speed/FLOPS": 201664418064228.25, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.26658815145492554, "optim/lr": 0.000768, "optim/total_tokens": 335806464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504404, "loss/out": 4.122982025146484, "created_at": "2025-01-14T14:47:41.518264+00:00"} {"global_step": 1282, "acc_step": 0, "speed/wps": 12839.043256318033, "speed/FLOPS": 201654817034731.12, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2550624907016754, "optim/lr": 0.0007685999999999999, "optim/total_tokens": 336068608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303047, "loss/out": 4.1171875, "created_at": "2025-01-14T14:47:46.630830+00:00"} {"global_step": 1283, "acc_step": 0, "speed/wps": 12843.104028671234, "speed/FLOPS": 201718597044624.1, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.26544082164764404, "optim/lr": 0.0007692, "optim/total_tokens": 336330752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498399, "loss/out": 4.072838306427002, "created_at": "2025-01-14T14:47:51.739317+00:00"} {"global_step": 1284, "acc_step": 0, "speed/wps": 12837.051141308963, "speed/FLOPS": 201623528131060.78, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2869080901145935, "optim/lr": 0.0007698, "optim/total_tokens": 336592896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474821, "loss/out": 4.223908424377441, "created_at": "2025-01-14T14:47:56.849392+00:00"} {"global_step": 1285, "acc_step": 0, "speed/wps": 12846.086750045626, "speed/FLOPS": 201765444782497.62, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3095954954624176, "optim/lr": 0.0007704, "optim/total_tokens": 336855040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488318, "loss/out": 3.9992051124572754, "created_at": "2025-01-14T14:48:01.955175+00:00"} {"global_step": 1286, "acc_step": 0, "speed/wps": 12844.368767507387, "speed/FLOPS": 201738461505979.16, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2775881886482239, "optim/lr": 0.000771, "optim/total_tokens": 337117184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486185, "loss/out": 4.115455627441406, "created_at": "2025-01-14T14:48:07.063858+00:00"} {"global_step": 1287, "acc_step": 0, "speed/wps": 12842.74571862481, "speed/FLOPS": 201712969292975.4, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2506205141544342, "optim/lr": 0.0007716, "optim/total_tokens": 337379328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494861, "loss/out": 3.960073232650757, "created_at": "2025-01-14T14:48:12.172799+00:00"} {"global_step": 1288, "acc_step": 0, "speed/wps": 12824.705187537038, "speed/FLOPS": 201429617961954.4, "speed/curr_iter_time": 1.2887, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.28449928760528564, "optim/lr": 0.0007722000000000001, "optim/total_tokens": 337641472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500938, "loss/out": 4.134477138519287, "created_at": "2025-01-14T14:48:17.294947+00:00"} {"global_step": 1289, "acc_step": 0, "speed/wps": 12842.560976408184, "speed/FLOPS": 201710067662600.88, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2615678310394287, "optim/lr": 0.0007728, "optim/total_tokens": 337903616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310462, "loss/out": 4.171828746795654, "created_at": "2025-01-14T14:48:22.409519+00:00"} {"global_step": 1290, "acc_step": 0, "speed/wps": 12839.772136688842, "speed/FLOPS": 201666265102541.25, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25382667779922485, "optim/lr": 0.0007733999999999999, "optim/total_tokens": 338165760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467407, "loss/out": 4.152002334594727, "created_at": "2025-01-14T14:48:27.518096+00:00"} {"global_step": 1291, "acc_step": 0, "speed/wps": 12847.202926038559, "speed/FLOPS": 201782975860253.16, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25787249207496643, "optim/lr": 0.0007740000000000001, "optim/total_tokens": 338427904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415924, "loss/out": 4.043933868408203, "created_at": "2025-01-14T14:48:32.622901+00:00"} {"global_step": 1292, "acc_step": 0, "speed/wps": 12839.856566995993, "speed/FLOPS": 201667591196534.47, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.24125656485557556, "optim/lr": 0.0007746, "optim/total_tokens": 338690048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494625, "loss/out": 4.044689655303955, "created_at": "2025-01-14T14:48:37.733265+00:00"} {"global_step": 1293, "acc_step": 0, "speed/wps": 12838.17829484735, "speed/FLOPS": 201641231626252.44, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23985733091831207, "optim/lr": 0.0007752000000000001, "optim/total_tokens": 338952192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490175, "loss/out": 4.1322736740112305, "created_at": "2025-01-14T14:48:42.844236+00:00"} {"global_step": 1294, "acc_step": 0, "speed/wps": 12838.88885486625, "speed/FLOPS": 201652391947761.06, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2543260157108307, "optim/lr": 0.0007758, "optim/total_tokens": 339214336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316842, "loss/out": 4.2046427726745605, "created_at": "2025-01-14T14:48:47.955457+00:00"} {"global_step": 1295, "acc_step": 0, "speed/wps": 12841.056605742246, "speed/FLOPS": 201686439454070.4, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2272314876317978, "optim/lr": 0.0007763999999999999, "optim/total_tokens": 339476480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475341, "loss/out": 4.1158857345581055, "created_at": "2025-01-14T14:48:53.065420+00:00"} {"global_step": 1296, "acc_step": 0, "speed/wps": 12843.596569436146, "speed/FLOPS": 201726333074158.97, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21608440577983856, "optim/lr": 0.000777, "optim/total_tokens": 339738624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 4.067721366882324, "created_at": "2025-01-14T14:48:58.171766+00:00"} {"global_step": 1297, "acc_step": 0, "speed/wps": 12841.64441063655, "speed/FLOPS": 201695671737664.03, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21276842057704926, "optim/lr": 0.0007775999999999999, "optim/total_tokens": 340000768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495387, "loss/out": 4.094351768493652, "created_at": "2025-01-14T14:49:03.279780+00:00"} {"global_step": 1298, "acc_step": 0, "speed/wps": 12846.592636799902, "speed/FLOPS": 201773390429134.75, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23660537600517273, "optim/lr": 0.0007782, "optim/total_tokens": 340262912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486694, "loss/out": 4.218020439147949, "created_at": "2025-01-14T14:49:08.387566+00:00"} {"global_step": 1299, "acc_step": 0, "speed/wps": 12838.452871713787, "speed/FLOPS": 201645544233247.97, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2754230499267578, "optim/lr": 0.0007788, "optim/total_tokens": 340525056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498328, "loss/out": 4.055298805236816, "created_at": "2025-01-14T14:49:13.496605+00:00"} {"global_step": 1300, "acc_step": 0, "speed/wps": 12844.53099086164, "speed/FLOPS": 201741009446675.97, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22633767127990723, "optim/lr": 0.0007794, "optim/total_tokens": 340787200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423342, "loss/out": 4.11628532409668, "created_at": "2025-01-14T14:49:18.603155+00:00"} {"global_step": 1301, "acc_step": 0, "speed/wps": 12843.5724617925, "speed/FLOPS": 201725954430488.16, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22032701969146729, "optim/lr": 0.0007800000000000001, "optim/total_tokens": 341049344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497525, "loss/out": 4.023491859436035, "created_at": "2025-01-14T14:49:23.709374+00:00"} {"global_step": 1302, "acc_step": 0, "speed/wps": 12829.324498638467, "speed/FLOPS": 201502170590400.97, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22778506577014923, "optim/lr": 0.0007806, "optim/total_tokens": 341311488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481431, "loss/out": 4.158929824829102, "created_at": "2025-01-14T14:49:28.822566+00:00"} {"global_step": 1303, "acc_step": 0, "speed/wps": 12835.660802994376, "speed/FLOPS": 201601690957305.16, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.24911215901374817, "optim/lr": 0.0007812000000000001, "optim/total_tokens": 341573632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469009, "loss/out": 4.021832466125488, "created_at": "2025-01-14T14:49:33.932872+00:00"} {"global_step": 1304, "acc_step": 0, "speed/wps": 12831.953008076087, "speed/FLOPS": 201543454943069.28, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.321454256772995, "optim/lr": 0.0007818, "optim/total_tokens": 341835776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487220, "loss/out": 4.042318820953369, "created_at": "2025-01-14T14:49:39.046088+00:00"} {"global_step": 1305, "acc_step": 0, "speed/wps": 12843.72240699149, "speed/FLOPS": 201728309525884.6, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3108188509941101, "optim/lr": 0.0007823999999999999, "optim/total_tokens": 342097920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331285, "loss/out": 4.182007789611816, "created_at": "2025-01-14T14:49:44.152548+00:00"} {"global_step": 1306, "acc_step": 0, "speed/wps": 12821.385296161714, "speed/FLOPS": 201377474505896.12, "speed/curr_iter_time": 1.2876, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3143859803676605, "optim/lr": 0.0007830000000000001, "optim/total_tokens": 342360064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492777, "loss/out": 4.033458709716797, "created_at": "2025-01-14T14:49:49.270450+00:00"} {"global_step": 1307, "acc_step": 0, "speed/wps": 12829.897136245094, "speed/FLOPS": 201511164650899.6, "speed/curr_iter_time": 1.2884, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.37137678265571594, "optim/lr": 0.0007836, "optim/total_tokens": 342622208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490923, "loss/out": 4.053498268127441, "created_at": "2025-01-14T14:49:54.384307+00:00"} {"global_step": 1308, "acc_step": 0, "speed/wps": 12832.304573442081, "speed/FLOPS": 201548976760243.66, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3478035628795624, "optim/lr": 0.0007842000000000001, "optim/total_tokens": 342884352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317421, "loss/out": 4.152408599853516, "created_at": "2025-01-14T14:49:59.498470+00:00"} {"global_step": 1309, "acc_step": 0, "speed/wps": 12843.841038338338, "speed/FLOPS": 201730172794203.38, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3053646683692932, "optim/lr": 0.0007848, "optim/total_tokens": 343146496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476431, "loss/out": 3.99857759475708, "created_at": "2025-01-14T14:50:04.604704+00:00"} {"global_step": 1310, "acc_step": 0, "speed/wps": 12844.95646770289, "speed/FLOPS": 201747692145134.25, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.27409833669662476, "optim/lr": 0.0007853999999999999, "optim/total_tokens": 343408640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369954, "loss/out": 4.132613182067871, "created_at": "2025-01-14T14:50:09.714226+00:00"} {"global_step": 1311, "acc_step": 0, "speed/wps": 12843.380276662238, "speed/FLOPS": 201722935898927.34, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2855895459651947, "optim/lr": 0.000786, "optim/total_tokens": 343670784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502273, "loss/out": 4.179725170135498, "created_at": "2025-01-14T14:50:14.821446+00:00"} {"global_step": 1312, "acc_step": 0, "speed/wps": 12841.225353624443, "speed/FLOPS": 201689089871444.2, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2939405143260956, "optim/lr": 0.0007865999999999999, "optim/total_tokens": 343932928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411683, "loss/out": 4.180782318115234, "created_at": "2025-01-14T14:50:19.929298+00:00"} {"global_step": 1313, "acc_step": 0, "speed/wps": 12844.000800869591, "speed/FLOPS": 201732682084293.44, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.29414457082748413, "optim/lr": 0.0007872, "optim/total_tokens": 344195072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486185, "loss/out": 4.092168807983398, "created_at": "2025-01-14T14:50:25.039873+00:00"} {"global_step": 1314, "acc_step": 0, "speed/wps": 12841.606547121955, "speed/FLOPS": 201695077039137.4, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.24613623321056366, "optim/lr": 0.0007878000000000001, "optim/total_tokens": 344457216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431605, "loss/out": 4.077401638031006, "created_at": "2025-01-14T14:50:30.150304+00:00"} {"global_step": 1315, "acc_step": 0, "speed/wps": 12848.773770375908, "speed/FLOPS": 201807648129137.56, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3375994861125946, "optim/lr": 0.0007884, "optim/total_tokens": 344719360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496716, "loss/out": 4.178339004516602, "created_at": "2025-01-14T14:50:35.275906+00:00"} {"global_step": 1316, "acc_step": 0, "speed/wps": 12844.72368638655, "speed/FLOPS": 201744035994686.06, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3451440930366516, "optim/lr": 0.0007890000000000001, "optim/total_tokens": 344981504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485941, "loss/out": 4.093037128448486, "created_at": "2025-01-14T14:50:40.381697+00:00"} {"global_step": 1317, "acc_step": 0, "speed/wps": 12841.443767252094, "speed/FLOPS": 201692520357598.25, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.24493730068206787, "optim/lr": 0.0007896, "optim/total_tokens": 345243648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332634, "loss/out": 4.064677715301514, "created_at": "2025-01-14T14:50:45.488824+00:00"} {"global_step": 1318, "acc_step": 0, "speed/wps": 12842.235857546622, "speed/FLOPS": 201704961224119.94, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.24392667412757874, "optim/lr": 0.0007902000000000001, "optim/total_tokens": 345505792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468454, "loss/out": 4.062453746795654, "created_at": "2025-01-14T14:50:50.598821+00:00"} {"global_step": 1319, "acc_step": 0, "speed/wps": 12843.054682678232, "speed/FLOPS": 201717821997995.5, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2086455374956131, "optim/lr": 0.0007908, "optim/total_tokens": 345767936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 4.147523403167725, "created_at": "2025-01-14T14:50:55.708677+00:00"} {"global_step": 1320, "acc_step": 0, "speed/wps": 12847.605770703358, "speed/FLOPS": 201789303089280.4, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22961054742336273, "optim/lr": 0.0007913999999999999, "optim/total_tokens": 346030080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500938, "loss/out": 3.959628105163574, "created_at": "2025-01-14T14:51:00.813478+00:00"} {"global_step": 1321, "acc_step": 0, "speed/wps": 12838.440349701872, "speed/FLOPS": 201645347557843.8, "speed/curr_iter_time": 1.2874, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.28900378942489624, "optim/lr": 0.0007920000000000001, "optim/total_tokens": 346292224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500418, "loss/out": 4.1586737632751465, "created_at": "2025-01-14T14:51:05.921921+00:00"} {"global_step": 1322, "acc_step": 0, "speed/wps": 12844.683333512476, "speed/FLOPS": 201743402197348.06, "speed/curr_iter_time": 1.2876, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.44984713196754456, "optim/lr": 0.0007926, "optim/total_tokens": 346554368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478487, "loss/out": 4.051786422729492, "created_at": "2025-01-14T14:51:11.032297+00:00"} {"global_step": 1323, "acc_step": 0, "speed/wps": 12834.045946162045, "speed/FLOPS": 201576327411707.56, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23165354132652283, "optim/lr": 0.0007932000000000001, "optim/total_tokens": 346816512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468974, "loss/out": 3.9997849464416504, "created_at": "2025-01-14T14:51:16.144618+00:00"} {"global_step": 1324, "acc_step": 0, "speed/wps": 12841.238715296224, "speed/FLOPS": 201689299734860.38, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.23052841424942017, "optim/lr": 0.0007938, "optim/total_tokens": 347078656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352273, "loss/out": 4.119983196258545, "created_at": "2025-01-14T14:51:21.257852+00:00"} {"global_step": 1325, "acc_step": 0, "speed/wps": 12838.456474197608, "speed/FLOPS": 201645600815206.72, "speed/curr_iter_time": 1.2882, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2520074248313904, "optim/lr": 0.0007943999999999999, "optim/total_tokens": 347340800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480094, "loss/out": 4.065810680389404, "created_at": "2025-01-14T14:51:26.366771+00:00"} {"global_step": 1326, "acc_step": 0, "speed/wps": 12843.96904918887, "speed/FLOPS": 201732183380516.34, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2646602690219879, "optim/lr": 0.000795, "optim/total_tokens": 347602944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310826, "loss/out": 4.04359769821167, "created_at": "2025-01-14T14:51:31.474299+00:00"} {"global_step": 1327, "acc_step": 0, "speed/wps": 12832.98397255853, "speed/FLOPS": 201559647656959.7, "speed/curr_iter_time": 1.2875, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.22142229974269867, "optim/lr": 0.0007956, "optim/total_tokens": 347865088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494402, "loss/out": 4.193357944488525, "created_at": "2025-01-14T14:51:36.584756+00:00"} {"global_step": 1328, "acc_step": 0, "speed/wps": 12839.185990719261, "speed/FLOPS": 201657058874640.03, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2646392583847046, "optim/lr": 0.0007962000000000001, "optim/total_tokens": 348127232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461033, "loss/out": 4.236826419830322, "created_at": "2025-01-14T14:51:41.692993+00:00"} {"global_step": 1329, "acc_step": 0, "speed/wps": 12844.13247215122, "speed/FLOPS": 201734750162706.72, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2536284923553467, "optim/lr": 0.0007968000000000001, "optim/total_tokens": 348389376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502273, "loss/out": 4.056176662445068, "created_at": "2025-01-14T14:51:46.801794+00:00"} {"global_step": 1330, "acc_step": 0, "speed/wps": 12839.325204970375, "speed/FLOPS": 201659245425754.2, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2996363043785095, "optim/lr": 0.0007974, "optim/total_tokens": 348651520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494106, "loss/out": 4.138761043548584, "created_at": "2025-01-14T14:51:51.910329+00:00"} {"global_step": 1331, "acc_step": 0, "speed/wps": 12844.11171730527, "speed/FLOPS": 201734424179330.94, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3355262577533722, "optim/lr": 0.0007980000000000001, "optim/total_tokens": 348913664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466361, "loss/out": 4.04273796081543, "created_at": "2025-01-14T14:51:57.019325+00:00"} {"global_step": 1332, "acc_step": 0, "speed/wps": 12842.549172962807, "speed/FLOPS": 201709882273271.6, "speed/curr_iter_time": 1.2879, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3297550678253174, "optim/lr": 0.0007986, "optim/total_tokens": 349175808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480094, "loss/out": 4.265170574188232, "created_at": "2025-01-14T14:52:02.131832+00:00"} {"global_step": 1333, "acc_step": 0, "speed/wps": 12838.312847216983, "speed/FLOPS": 201643344956111.84, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3174521327018738, "optim/lr": 0.0007992000000000001, "optim/total_tokens": 349437952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 392890, "loss/out": 4.082796096801758, "created_at": "2025-01-14T14:52:07.244024+00:00"} {"global_step": 1334, "acc_step": 0, "speed/wps": 12838.629531656743, "speed/FLOPS": 201648318920408.3, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.26447680592536926, "optim/lr": 0.0007998, "optim/total_tokens": 349700096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489068, "loss/out": 4.158961772918701, "created_at": "2025-01-14T14:52:12.356623+00:00"} {"global_step": 1335, "acc_step": 0, "speed/wps": 12845.02820390097, "speed/FLOPS": 201748818860700.97, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2959447503089905, "optim/lr": 0.0008003999999999999, "optim/total_tokens": 349962240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482471, "loss/out": 3.953902244567871, "created_at": "2025-01-14T14:52:17.464754+00:00"} {"global_step": 1336, "acc_step": 0, "speed/wps": 12839.826599943448, "speed/FLOPS": 201667120522795.2, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2729155719280243, "optim/lr": 0.0008010000000000001, "optim/total_tokens": 350224384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323801, "loss/out": 3.9623055458068848, "created_at": "2025-01-14T14:52:22.573559+00:00"} {"global_step": 1337, "acc_step": 0, "speed/wps": 12845.639908580106, "speed/FLOPS": 201758426523257.1, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.24292133748531342, "optim/lr": 0.0008016, "optim/total_tokens": 350486528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463984, "loss/out": 4.019658088684082, "created_at": "2025-01-14T14:52:27.683269+00:00"} {"global_step": 1338, "acc_step": 0, "speed/wps": 12837.01713136401, "speed/FLOPS": 201622993958140.72, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.26018965244293213, "optim/lr": 0.0008022000000000001, "optim/total_tokens": 350748672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349067, "loss/out": 4.097731113433838, "created_at": "2025-01-14T14:52:32.791953+00:00"} {"global_step": 1339, "acc_step": 0, "speed/wps": 12841.833822957256, "speed/FLOPS": 201698646718440.53, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.29758143424987793, "optim/lr": 0.0008028, "optim/total_tokens": 351010816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479331, "loss/out": 4.046126365661621, "created_at": "2025-01-14T14:52:37.904569+00:00"} {"global_step": 1340, "acc_step": 0, "speed/wps": 12846.469620607983, "speed/FLOPS": 201771458290798.72, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22638261318206787, "optim/lr": 0.0008034, "optim/total_tokens": 351272960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312410, "loss/out": 4.040905952453613, "created_at": "2025-01-14T14:52:43.009798+00:00"} {"global_step": 1341, "acc_step": 0, "speed/wps": 12837.825981308602, "speed/FLOPS": 201635698057996.88, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2357504963874817, "optim/lr": 0.000804, "optim/total_tokens": 351535104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495975, "loss/out": 4.014890670776367, "created_at": "2025-01-14T14:52:48.121249+00:00"} {"global_step": 1342, "acc_step": 0, "speed/wps": 12846.493710985898, "speed/FLOPS": 201771836663287.53, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3136906325817108, "optim/lr": 0.0008046, "optim/total_tokens": 351797248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 396157, "loss/out": 3.919376850128174, "created_at": "2025-01-14T14:52:53.228639+00:00"} {"global_step": 1343, "acc_step": 0, "speed/wps": 12834.022354201903, "speed/FLOPS": 201575956867554.78, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3104746639728546, "optim/lr": 0.0008052000000000001, "optim/total_tokens": 352059392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499609, "loss/out": 4.090664386749268, "created_at": "2025-01-14T14:52:58.338937+00:00"} {"global_step": 1344, "acc_step": 0, "speed/wps": 12841.87491724437, "speed/FLOPS": 201699292160689.1, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2819781005382538, "optim/lr": 0.0008058000000000001, "optim/total_tokens": 352321536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487740, "loss/out": 4.048408508300781, "created_at": "2025-01-14T14:53:03.450153+00:00"} {"global_step": 1345, "acc_step": 0, "speed/wps": 12840.21194178141, "speed/FLOPS": 201673172845875.56, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3013049364089966, "optim/lr": 0.0008064, "optim/total_tokens": 352583680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477718, "loss/out": 4.088981628417969, "created_at": "2025-01-14T14:53:08.563267+00:00"} {"global_step": 1346, "acc_step": 0, "speed/wps": 12848.57239740869, "speed/FLOPS": 201804485289971.94, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2963888943195343, "optim/lr": 0.0008070000000000001, "optim/total_tokens": 352845824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468737, "loss/out": 4.030386924743652, "created_at": "2025-01-14T14:53:13.669860+00:00"} {"global_step": 1347, "acc_step": 0, "speed/wps": 12843.861461764547, "speed/FLOPS": 201730493572176.72, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2897557318210602, "optim/lr": 0.0008076, "optim/total_tokens": 353107968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351701, "loss/out": 3.9504427909851074, "created_at": "2025-01-14T14:53:18.779945+00:00"} {"global_step": 1348, "acc_step": 0, "speed/wps": 12845.424782721271, "speed/FLOPS": 201755047676030.2, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.28662770986557007, "optim/lr": 0.0008081999999999999, "optim/total_tokens": 353370112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493533, "loss/out": 3.9512131214141846, "created_at": "2025-01-14T14:53:23.891009+00:00"} {"global_step": 1349, "acc_step": 0, "speed/wps": 12843.068172389276, "speed/FLOPS": 201718033872444.25, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2833317220211029, "optim/lr": 0.0008088, "optim/total_tokens": 353632256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474052, "loss/out": 4.040330410003662, "created_at": "2025-01-14T14:53:28.997341+00:00"} {"global_step": 1350, "acc_step": 0, "speed/wps": 12845.947396163676, "speed/FLOPS": 201763256038289.28, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2348366528749466, "optim/lr": 0.0008093999999999999, "optim/total_tokens": 353894400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475867, "loss/out": 4.135111331939697, "created_at": "2025-01-14T14:53:34.104168+00:00"} {"global_step": 1351, "acc_step": 0, "speed/wps": 12841.965447468918, "speed/FLOPS": 201700714062267.2, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.266964316368103, "optim/lr": 0.0008100000000000001, "optim/total_tokens": 354156544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486411, "loss/out": 4.027251243591309, "created_at": "2025-01-14T14:53:39.214582+00:00"} {"global_step": 1352, "acc_step": 0, "speed/wps": 12839.97406327561, "speed/FLOPS": 201669436637062.66, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.26026540994644165, "optim/lr": 0.0008106, "optim/total_tokens": 354418688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461324, "loss/out": 4.083229064941406, "created_at": "2025-01-14T14:53:44.325499+00:00"} {"global_step": 1353, "acc_step": 0, "speed/wps": 12838.486788546632, "speed/FLOPS": 201646076943714.4, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22552849352359772, "optim/lr": 0.0008112, "optim/total_tokens": 354680832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502342, "loss/out": 4.131058692932129, "created_at": "2025-01-14T14:53:49.433810+00:00"} {"global_step": 1354, "acc_step": 0, "speed/wps": 12844.82526017813, "speed/FLOPS": 201745631350659.84, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24829265475273132, "optim/lr": 0.0008118, "optim/total_tokens": 354942976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473491, "loss/out": 4.051204681396484, "created_at": "2025-01-14T14:53:54.540072+00:00"} {"global_step": 1355, "acc_step": 0, "speed/wps": 12845.47745086461, "speed/FLOPS": 201755874901594.94, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23559513688087463, "optim/lr": 0.0008124, "optim/total_tokens": 355205120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489830, "loss/out": 4.093020439147949, "created_at": "2025-01-14T14:53:59.649332+00:00"} {"global_step": 1356, "acc_step": 0, "speed/wps": 12845.970452756355, "speed/FLOPS": 201763618173760.4, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23129957914352417, "optim/lr": 0.000813, "optim/total_tokens": 355467264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327543, "loss/out": 4.079368591308594, "created_at": "2025-01-14T14:54:04.756005+00:00"} {"global_step": 1357, "acc_step": 0, "speed/wps": 12848.623223927212, "speed/FLOPS": 201805283590288.28, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20521888136863708, "optim/lr": 0.0008136, "optim/total_tokens": 355729408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484328, "loss/out": 4.0056843757629395, "created_at": "2025-01-14T14:54:09.860671+00:00"} {"global_step": 1358, "acc_step": 0, "speed/wps": 12841.194700620323, "speed/FLOPS": 201688608423892.84, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.23352308571338654, "optim/lr": 0.0008142, "optim/total_tokens": 355991552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299587, "loss/out": 4.100898742675781, "created_at": "2025-01-14T14:54:14.971492+00:00"} {"global_step": 1359, "acc_step": 0, "speed/wps": 12841.88676632177, "speed/FLOPS": 201699478266732.0, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21762214601039886, "optim/lr": 0.0008148000000000001, "optim/total_tokens": 356253696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484895, "loss/out": 4.017331123352051, "created_at": "2025-01-14T14:54:20.078641+00:00"} {"global_step": 1360, "acc_step": 0, "speed/wps": 12845.04487954576, "speed/FLOPS": 201749080774617.12, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23430177569389343, "optim/lr": 0.0008154, "optim/total_tokens": 356515840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359278, "loss/out": 4.218703269958496, "created_at": "2025-01-14T14:54:25.185082+00:00"} {"global_step": 1361, "acc_step": 0, "speed/wps": 12838.852293611026, "speed/FLOPS": 201651817703007.3, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2218713015317917, "optim/lr": 0.0008160000000000001, "optim/total_tokens": 356777984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497235, "loss/out": 4.120281219482422, "created_at": "2025-01-14T14:54:30.297236+00:00"} {"global_step": 1362, "acc_step": 0, "speed/wps": 12834.868673760422, "speed/FLOPS": 201589249479186.25, "speed/curr_iter_time": 1.2882, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24937985837459564, "optim/lr": 0.0008166, "optim/total_tokens": 357040128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502792, "loss/out": 3.9953157901763916, "created_at": "2025-01-14T14:54:35.413481+00:00"} {"global_step": 1363, "acc_step": 0, "speed/wps": 12849.242136204502, "speed/FLOPS": 201815004458077.72, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2410997599363327, "optim/lr": 0.0008171999999999999, "optim/total_tokens": 357302272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479338, "loss/out": 3.976595163345337, "created_at": "2025-01-14T14:54:40.521580+00:00"} {"global_step": 1364, "acc_step": 0, "speed/wps": 12839.217292403493, "speed/FLOPS": 201657550510603.9, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2290053367614746, "optim/lr": 0.0008178, "optim/total_tokens": 357564416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490403, "loss/out": 4.0655436515808105, "created_at": "2025-01-14T14:54:45.629438+00:00"} {"global_step": 1365, "acc_step": 0, "speed/wps": 12846.519548673301, "speed/FLOPS": 201772242479672.88, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24862982332706451, "optim/lr": 0.0008183999999999999, "optim/total_tokens": 357826560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328599, "loss/out": 3.9628186225891113, "created_at": "2025-01-14T14:54:50.738622+00:00"} {"global_step": 1366, "acc_step": 0, "speed/wps": 12841.328055494556, "speed/FLOPS": 201690702945441.5, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.27896648645401, "optim/lr": 0.0008190000000000001, "optim/total_tokens": 358088704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502556, "loss/out": 4.137759208679199, "created_at": "2025-01-14T14:54:55.853363+00:00"} {"global_step": 1367, "acc_step": 0, "speed/wps": 12842.652601508022, "speed/FLOPS": 201711506760699.94, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2749612033367157, "optim/lr": 0.0008196, "optim/total_tokens": 358350848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306793, "loss/out": 4.001486301422119, "created_at": "2025-01-14T14:55:00.959927+00:00"} {"global_step": 1368, "acc_step": 0, "speed/wps": 12845.796915939402, "speed/FLOPS": 201760892539585.56, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.23035989701747894, "optim/lr": 0.0008202, "optim/total_tokens": 358612992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484895, "loss/out": 4.060258865356445, "created_at": "2025-01-14T14:55:06.068452+00:00"} {"global_step": 1369, "acc_step": 0, "speed/wps": 12846.569756214936, "speed/FLOPS": 201773031058106.66, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.25792622566223145, "optim/lr": 0.0008208, "optim/total_tokens": 358875136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358407, "loss/out": 3.962264060974121, "created_at": "2025-01-14T14:55:11.175964+00:00"} {"global_step": 1370, "acc_step": 0, "speed/wps": 12844.981441047037, "speed/FLOPS": 201748084385821.03, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20599815249443054, "optim/lr": 0.0008214, "optim/total_tokens": 359137280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493297, "loss/out": 4.026721000671387, "created_at": "2025-01-14T14:55:16.283036+00:00"} {"global_step": 1371, "acc_step": 0, "speed/wps": 12844.360444624132, "speed/FLOPS": 201738330783660.8, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2184571921825409, "optim/lr": 0.000822, "optim/total_tokens": 359399424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 3.9961929321289062, "created_at": "2025-01-14T14:55:21.389422+00:00"} {"global_step": 1372, "acc_step": 0, "speed/wps": 12848.14428118004, "speed/FLOPS": 201797761136307.66, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22066807746887207, "optim/lr": 0.0008226, "optim/total_tokens": 359661568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491159, "loss/out": 3.891312599182129, "created_at": "2025-01-14T14:55:26.493785+00:00"} {"global_step": 1373, "acc_step": 0, "speed/wps": 12843.24461244937, "speed/FLOPS": 201720805106042.1, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2521495521068573, "optim/lr": 0.0008232, "optim/total_tokens": 359923712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486988, "loss/out": 4.013645172119141, "created_at": "2025-01-14T14:55:31.602352+00:00"} {"global_step": 1374, "acc_step": 0, "speed/wps": 12843.46248763974, "speed/FLOPS": 201724227135298.72, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3163416385650635, "optim/lr": 0.0008238000000000001, "optim/total_tokens": 360185856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325437, "loss/out": 3.9482579231262207, "created_at": "2025-01-14T14:55:36.714906+00:00"} {"global_step": 1375, "acc_step": 0, "speed/wps": 12846.527646864304, "speed/FLOPS": 201772369672890.66, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2680876851081848, "optim/lr": 0.0008244, "optim/total_tokens": 360448000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472930, "loss/out": 4.037059783935547, "created_at": "2025-01-14T14:55:41.820267+00:00"} {"global_step": 1376, "acc_step": 0, "speed/wps": 12843.955053555472, "speed/FLOPS": 201731963559861.78, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.30722272396087646, "optim/lr": 0.0008250000000000001, "optim/total_tokens": 360710144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366127, "loss/out": 4.074921607971191, "created_at": "2025-01-14T14:55:46.926273+00:00"} {"global_step": 1377, "acc_step": 0, "speed/wps": 12850.025058838612, "speed/FLOPS": 201827301333894.84, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2463962286710739, "optim/lr": 0.0008256, "optim/total_tokens": 360972288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490985, "loss/out": 3.9925408363342285, "created_at": "2025-01-14T14:55:52.036238+00:00"} {"global_step": 1378, "acc_step": 0, "speed/wps": 12840.69170856194, "speed/FLOPS": 201680708242432.44, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2789955139160156, "optim/lr": 0.0008261999999999999, "optim/total_tokens": 361234432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422843, "loss/out": 4.076170444488525, "created_at": "2025-01-14T14:55:57.145503+00:00"} {"global_step": 1379, "acc_step": 0, "speed/wps": 12844.377444417723, "speed/FLOPS": 201738597788779.3, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3003056049346924, "optim/lr": 0.0008268, "optim/total_tokens": 361496576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480094, "loss/out": 3.9413681030273438, "created_at": "2025-01-14T14:56:02.256266+00:00"} {"global_step": 1380, "acc_step": 0, "speed/wps": 12849.565645999795, "speed/FLOPS": 201820085623961.84, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24102404713630676, "optim/lr": 0.0008273999999999999, "optim/total_tokens": 361758720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 405290, "loss/out": 4.033342361450195, "created_at": "2025-01-14T14:56:07.363090+00:00"} {"global_step": 1381, "acc_step": 0, "speed/wps": 12846.27210261649, "speed/FLOPS": 201768355999324.16, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.29862314462661743, "optim/lr": 0.0008280000000000001, "optim/total_tokens": 362020864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493533, "loss/out": 4.033965110778809, "created_at": "2025-01-14T14:56:12.476075+00:00"} {"global_step": 1382, "acc_step": 0, "speed/wps": 12847.289522867133, "speed/FLOPS": 201784335982442.44, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.26802587509155273, "optim/lr": 0.0008286, "optim/total_tokens": 362283008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487515, "loss/out": 4.021241188049316, "created_at": "2025-01-14T14:56:17.584126+00:00"} {"global_step": 1383, "acc_step": 0, "speed/wps": 12842.49060138147, "speed/FLOPS": 201708962326100.94, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23847459256649017, "optim/lr": 0.0008292, "optim/total_tokens": 362545152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471601, "loss/out": 3.9849016666412354, "created_at": "2025-01-14T14:56:22.692275+00:00"} {"global_step": 1384, "acc_step": 0, "speed/wps": 12841.97548510873, "speed/FLOPS": 201700871717193.6, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22669318318367004, "optim/lr": 0.0008298, "optim/total_tokens": 362807296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489594, "loss/out": 4.086407661437988, "created_at": "2025-01-14T14:56:27.804272+00:00"} {"global_step": 1385, "acc_step": 0, "speed/wps": 12844.232876101536, "speed/FLOPS": 201736327144715.9, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2416089028120041, "optim/lr": 0.0008304, "optim/total_tokens": 363069440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303047, "loss/out": 4.062778949737549, "created_at": "2025-01-14T14:56:32.912668+00:00"} {"global_step": 1386, "acc_step": 0, "speed/wps": 12845.284029280394, "speed/FLOPS": 201752836949825.72, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24861951172351837, "optim/lr": 0.0008310000000000001, "optim/total_tokens": 363331584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478244, "loss/out": 4.0128703117370605, "created_at": "2025-01-14T14:56:38.018424+00:00"} {"global_step": 1387, "acc_step": 0, "speed/wps": 12844.718591561375, "speed/FLOPS": 201743955973455.7, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23279568552970886, "optim/lr": 0.0008316, "optim/total_tokens": 363593728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368376, "loss/out": 3.9240994453430176, "created_at": "2025-01-14T14:56:43.129198+00:00"} {"global_step": 1388, "acc_step": 0, "speed/wps": 12843.743695067687, "speed/FLOPS": 201728643884374.16, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19734089076519012, "optim/lr": 0.0008322, "optim/total_tokens": 363855872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502792, "loss/out": 4.037587642669678, "created_at": "2025-01-14T14:56:48.241478+00:00"} {"global_step": 1389, "acc_step": 0, "speed/wps": 12845.267629701508, "speed/FLOPS": 201752579371903.94, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.24325694143772125, "optim/lr": 0.0008328000000000001, "optim/total_tokens": 364118016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484895, "loss/out": 3.8536922931671143, "created_at": "2025-01-14T14:56:53.349945+00:00"} {"global_step": 1390, "acc_step": 0, "speed/wps": 12842.593679752923, "speed/FLOPS": 201710581313569.12, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25851738452911377, "optim/lr": 0.0008334, "optim/total_tokens": 364380160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322311, "loss/out": 4.1424078941345215, "created_at": "2025-01-14T14:56:58.457376+00:00"} {"global_step": 1391, "acc_step": 0, "speed/wps": 12842.804873821597, "speed/FLOPS": 201713898406628.72, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2744190990924835, "optim/lr": 0.0008340000000000001, "optim/total_tokens": 364642304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496542, "loss/out": 4.097365379333496, "created_at": "2025-01-14T14:57:03.567712+00:00"} {"global_step": 1392, "acc_step": 0, "speed/wps": 12845.026593080995, "speed/FLOPS": 201748793560559.75, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.38537487387657166, "optim/lr": 0.0008346, "optim/total_tokens": 364904448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309195, "loss/out": 4.035802841186523, "created_at": "2025-01-14T14:57:08.677305+00:00"} {"global_step": 1393, "acc_step": 0, "speed/wps": 12832.607170814777, "speed/FLOPS": 201553729467794.22, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2700258195400238, "optim/lr": 0.0008351999999999999, "optim/total_tokens": 365166592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470304, "loss/out": 4.0443034172058105, "created_at": "2025-01-14T14:57:13.790917+00:00"} {"global_step": 1394, "acc_step": 0, "speed/wps": 12845.118530046584, "speed/FLOPS": 201750237556938.97, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24694141745567322, "optim/lr": 0.0008358, "optim/total_tokens": 365428736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365596, "loss/out": 3.9789435863494873, "created_at": "2025-01-14T14:57:18.897676+00:00"} {"global_step": 1395, "acc_step": 0, "speed/wps": 12845.790607900399, "speed/FLOPS": 201760793463164.88, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2913143038749695, "optim/lr": 0.0008364, "optim/total_tokens": 365690880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499373, "loss/out": 4.039824485778809, "created_at": "2025-01-14T14:57:24.003357+00:00"} {"global_step": 1396, "acc_step": 0, "speed/wps": 12844.893487287545, "speed/FLOPS": 201746702951167.1, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2510920464992523, "optim/lr": 0.0008370000000000001, "optim/total_tokens": 365953024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476151, "loss/out": 3.86757755279541, "created_at": "2025-01-14T14:57:29.109635+00:00"} {"global_step": 1397, "acc_step": 0, "speed/wps": 12835.814293291844, "speed/FLOPS": 201604101733344.66, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2636006474494934, "optim/lr": 0.0008376, "optim/total_tokens": 366215168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480904, "loss/out": 4.042704105377197, "created_at": "2025-01-14T14:57:34.223068+00:00"} {"global_step": 1398, "acc_step": 0, "speed/wps": 12842.356568456793, "speed/FLOPS": 201706857154838.34, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.268426775932312, "optim/lr": 0.0008382, "optim/total_tokens": 366477312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494861, "loss/out": 4.079061508178711, "created_at": "2025-01-14T14:57:39.334380+00:00"} {"global_step": 1399, "acc_step": 0, "speed/wps": 12843.884440540967, "speed/FLOPS": 201730854485436.56, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2630178928375244, "optim/lr": 0.0008388000000000001, "optim/total_tokens": 366739456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308424, "loss/out": 4.032220840454102, "created_at": "2025-01-14T14:57:44.446567+00:00"} {"global_step": 1400, "acc_step": 0, "speed/wps": 12844.120644717514, "speed/FLOPS": 201734564396607.78, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23777717351913452, "optim/lr": 0.0008394, "optim/total_tokens": 367001600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487272, "loss/out": 3.927701234817505, "created_at": "2025-01-14T14:57:49.552537+00:00"} {"global_step": 1401, "acc_step": 0, "speed/wps": 12841.50998996026, "speed/FLOPS": 201693560476229.5, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2255140095949173, "optim/lr": 0.0008400000000000001, "optim/total_tokens": 367263744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423656, "loss/out": 4.066821098327637, "created_at": "2025-01-14T14:57:54.662152+00:00"} {"global_step": 1402, "acc_step": 0, "speed/wps": 12847.054536690011, "speed/FLOPS": 201780645201625.9, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24036723375320435, "optim/lr": 0.0008406, "optim/total_tokens": 367525888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497305, "loss/out": 3.9848828315734863, "created_at": "2025-01-14T14:57:59.766945+00:00"} {"global_step": 1403, "acc_step": 0, "speed/wps": 12842.487882585549, "speed/FLOPS": 201708919623674.97, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22168241441249847, "optim/lr": 0.0008412, "optim/total_tokens": 367788032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492835, "loss/out": 3.981476306915283, "created_at": "2025-01-14T14:58:04.875082+00:00"} {"global_step": 1404, "acc_step": 0, "speed/wps": 12843.971672494405, "speed/FLOPS": 201732224583134.56, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2574385702610016, "optim/lr": 0.0008418000000000001, "optim/total_tokens": 368050176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497525, "loss/out": 4.015039920806885, "created_at": "2025-01-14T14:58:09.983108+00:00"} {"global_step": 1405, "acc_step": 0, "speed/wps": 12844.521835459393, "speed/FLOPS": 201740865648502.8, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23003564774990082, "optim/lr": 0.0008424, "optim/total_tokens": 368312320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479906, "loss/out": 4.035258769989014, "created_at": "2025-01-14T14:58:15.089198+00:00"} {"global_step": 1406, "acc_step": 0, "speed/wps": 12840.440527218532, "speed/FLOPS": 201676763094274.88, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20545834302902222, "optim/lr": 0.0008430000000000001, "optim/total_tokens": 368574464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489891, "loss/out": 3.966825485229492, "created_at": "2025-01-14T14:58:20.199326+00:00"} {"global_step": 1407, "acc_step": 0, "speed/wps": 12847.354171022273, "speed/FLOPS": 201785351370554.78, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2286030501127243, "optim/lr": 0.0008436, "optim/total_tokens": 368836608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478811, "loss/out": 4.076422691345215, "created_at": "2025-01-14T14:58:25.305259+00:00"} {"global_step": 1408, "acc_step": 0, "speed/wps": 12836.655709260389, "speed/FLOPS": 201617317327357.34, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2525654733181, "optim/lr": 0.0008441999999999999, "optim/total_tokens": 369098752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473774, "loss/out": 3.938401699066162, "created_at": "2025-01-14T14:58:30.415643+00:00"} {"global_step": 1409, "acc_step": 0, "speed/wps": 12846.160166256033, "speed/FLOPS": 201766597884966.5, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2811174690723419, "optim/lr": 0.0008448, "optim/total_tokens": 369360896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494928, "loss/out": 3.8780336380004883, "created_at": "2025-01-14T14:58:35.521345+00:00"} {"global_step": 1410, "acc_step": 0, "speed/wps": 12835.682087768271, "speed/FLOPS": 201602025263927.5, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.28743988275527954, "optim/lr": 0.0008454, "optim/total_tokens": 369623040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316842, "loss/out": 3.9344043731689453, "created_at": "2025-01-14T14:58:40.630931+00:00"} {"global_step": 1411, "acc_step": 0, "speed/wps": 12841.219000864605, "speed/FLOPS": 201688990092621.34, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2817995846271515, "optim/lr": 0.000846, "optim/total_tokens": 369885184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474301, "loss/out": 4.003325462341309, "created_at": "2025-01-14T14:58:45.740029+00:00"} {"global_step": 1412, "acc_step": 0, "speed/wps": 12841.263193801087, "speed/FLOPS": 201689684203416.25, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23622776567935944, "optim/lr": 0.0008466000000000001, "optim/total_tokens": 370147328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353444, "loss/out": 4.140383720397949, "created_at": "2025-01-14T14:58:50.855985+00:00"} {"global_step": 1413, "acc_step": 0, "speed/wps": 12842.314922655141, "speed/FLOPS": 201706203050296.06, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3013400733470917, "optim/lr": 0.0008472, "optim/total_tokens": 370409472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499609, "loss/out": 4.006617069244385, "created_at": "2025-01-14T14:58:55.963227+00:00"} {"global_step": 1414, "acc_step": 0, "speed/wps": 12841.408904465932, "speed/FLOPS": 201691972789635.1, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2888130843639374, "optim/lr": 0.0008478000000000001, "optim/total_tokens": 370671616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471145, "loss/out": 4.019827365875244, "created_at": "2025-01-14T14:59:01.073566+00:00"} {"global_step": 1415, "acc_step": 0, "speed/wps": 12844.077948918059, "speed/FLOPS": 201733893800407.16, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2301941215991974, "optim/lr": 0.0008484, "optim/total_tokens": 370933760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482709, "loss/out": 3.844944953918457, "created_at": "2025-01-14T14:59:06.182848+00:00"} {"global_step": 1416, "acc_step": 0, "speed/wps": 12844.019187565105, "speed/FLOPS": 201732970872612.47, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21008867025375366, "optim/lr": 0.0008489999999999999, "optim/total_tokens": 371195904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498044, "loss/out": 4.026289463043213, "created_at": "2025-01-14T14:59:11.292999+00:00"} {"global_step": 1417, "acc_step": 0, "speed/wps": 12841.494085105825, "speed/FLOPS": 201693310668635.06, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2421329766511917, "optim/lr": 0.0008496, "optim/total_tokens": 371458048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314443, "loss/out": 3.93428897857666, "created_at": "2025-01-14T14:59:16.403752+00:00"} {"global_step": 1418, "acc_step": 0, "speed/wps": 12841.862308875103, "speed/FLOPS": 201699094128924.06, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22410069406032562, "optim/lr": 0.0008502, "optim/total_tokens": 371720192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486752, "loss/out": 4.088537216186523, "created_at": "2025-01-14T14:59:21.512367+00:00"} {"global_step": 1419, "acc_step": 0, "speed/wps": 12836.115843335496, "speed/FLOPS": 201608837991153.38, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24334204196929932, "optim/lr": 0.0008508000000000001, "optim/total_tokens": 371982336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314736, "loss/out": 3.956218719482422, "created_at": "2025-01-14T14:59:26.621483+00:00"} {"global_step": 1420, "acc_step": 0, "speed/wps": 12843.880706073953, "speed/FLOPS": 201730795830500.4, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2585943341255188, "optim/lr": 0.0008514, "optim/total_tokens": 372244480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488845, "loss/out": 3.9578466415405273, "created_at": "2025-01-14T14:59:31.729479+00:00"} {"global_step": 1421, "acc_step": 0, "speed/wps": 12845.970326123093, "speed/FLOPS": 201763616184811.0, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24093352258205414, "optim/lr": 0.0008519999999999999, "optim/total_tokens": 372506624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297534, "loss/out": 3.951529026031494, "created_at": "2025-01-14T14:59:36.834881+00:00"} {"global_step": 1422, "acc_step": 0, "speed/wps": 12840.683364668883, "speed/FLOPS": 201680577190126.25, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2305237501859665, "optim/lr": 0.0008526, "optim/total_tokens": 372768768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453620, "loss/out": 4.016726493835449, "created_at": "2025-01-14T14:59:41.943562+00:00"} {"global_step": 1423, "acc_step": 0, "speed/wps": 12847.13316155534, "speed/FLOPS": 201781880113177.78, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23375965654850006, "optim/lr": 0.0008531999999999999, "optim/total_tokens": 373030912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402562, "loss/out": 4.044307231903076, "created_at": "2025-01-14T14:59:47.050027+00:00"} {"global_step": 1424, "acc_step": 0, "speed/wps": 12846.393782943356, "speed/FLOPS": 201770267156063.88, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2501448690891266, "optim/lr": 0.0008538, "optim/total_tokens": 373293056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502030, "loss/out": 3.9173362255096436, "created_at": "2025-01-14T14:59:52.155530+00:00"} {"global_step": 1425, "acc_step": 0, "speed/wps": 12844.136307573446, "speed/FLOPS": 201734810403283.22, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2298952341079712, "optim/lr": 0.0008544000000000001, "optim/total_tokens": 373555200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481474, "loss/out": 3.980424404144287, "created_at": "2025-01-14T14:59:57.261787+00:00"} {"global_step": 1426, "acc_step": 0, "speed/wps": 12842.027421884199, "speed/FLOPS": 201701687455620.97, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.323462575674057, "optim/lr": 0.000855, "optim/total_tokens": 373817344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480904, "loss/out": 3.9997901916503906, "created_at": "2025-01-14T15:00:02.373095+00:00"} {"global_step": 1427, "acc_step": 0, "speed/wps": 12845.4406011744, "speed/FLOPS": 201755296126573.4, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2681821882724762, "optim/lr": 0.0008556000000000001, "optim/total_tokens": 374079488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488266, "loss/out": 4.062782287597656, "created_at": "2025-01-14T15:00:07.483939+00:00"} {"global_step": 1428, "acc_step": 0, "speed/wps": 12833.556382195007, "speed/FLOPS": 201568638136873.84, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2958658039569855, "optim/lr": 0.0008562, "optim/total_tokens": 374341632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302515, "loss/out": 4.0254645347595215, "created_at": "2025-01-14T15:00:12.602578+00:00"} {"global_step": 1429, "acc_step": 0, "speed/wps": 12843.024878838489, "speed/FLOPS": 201717353887737.56, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.304340124130249, "optim/lr": 0.0008568000000000001, "optim/total_tokens": 374603776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494118, "loss/out": 4.005529403686523, "created_at": "2025-01-14T15:00:17.708930+00:00"} {"global_step": 1430, "acc_step": 0, "speed/wps": 12840.393973488688, "speed/FLOPS": 201676031903976.16, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24354013800621033, "optim/lr": 0.0008574, "optim/total_tokens": 374865920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431855, "loss/out": 3.934943199157715, "created_at": "2025-01-14T15:00:22.817556+00:00"} {"global_step": 1431, "acc_step": 0, "speed/wps": 12847.08703541391, "speed/FLOPS": 201781155638742.66, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22998234629631042, "optim/lr": 0.0008579999999999999, "optim/total_tokens": 375128064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474821, "loss/out": 3.9902777671813965, "created_at": "2025-01-14T15:00:27.926102+00:00"} {"global_step": 1432, "acc_step": 0, "speed/wps": 12842.477743361942, "speed/FLOPS": 201708760373235.47, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22408652305603027, "optim/lr": 0.0008586, "optim/total_tokens": 375390208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294251, "loss/out": 3.9907007217407227, "created_at": "2025-01-14T15:00:33.037049+00:00"} {"global_step": 1433, "acc_step": 0, "speed/wps": 12845.48872974937, "speed/FLOPS": 201756052051978.9, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21810169517993927, "optim/lr": 0.0008592, "optim/total_tokens": 375652352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500202, "loss/out": 3.8114681243896484, "created_at": "2025-01-14T15:00:38.143497+00:00"} {"global_step": 1434, "acc_step": 0, "speed/wps": 12841.039619424428, "speed/FLOPS": 201686172660607.44, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18512845039367676, "optim/lr": 0.0008598000000000001, "optim/total_tokens": 375914496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434294, "loss/out": 4.037075042724609, "created_at": "2025-01-14T15:00:43.253000+00:00"} {"global_step": 1435, "acc_step": 0, "speed/wps": 12837.46388367226, "speed/FLOPS": 201630010817043.8, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2128286063671112, "optim/lr": 0.0008604, "optim/total_tokens": 376176640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498328, "loss/out": 3.933058738708496, "created_at": "2025-01-14T15:00:48.363691+00:00"} {"global_step": 1436, "acc_step": 0, "speed/wps": 12840.203699258927, "speed/FLOPS": 201673043385733.56, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20395739376544952, "optim/lr": 0.000861, "optim/total_tokens": 376438784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484328, "loss/out": 4.113076210021973, "created_at": "2025-01-14T15:00:53.471260+00:00"} {"global_step": 1437, "acc_step": 0, "speed/wps": 12847.730131631924, "speed/FLOPS": 201791256348553.38, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2246873825788498, "optim/lr": 0.0008616, "optim/total_tokens": 376700928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482992, "loss/out": 3.848194122314453, "created_at": "2025-01-14T15:00:58.578696+00:00"} {"global_step": 1438, "acc_step": 0, "speed/wps": 12842.117098825289, "speed/FLOPS": 201703095955209.8, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.254316508769989, "optim/lr": 0.0008622, "optim/total_tokens": 376963072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484846, "loss/out": 3.9548349380493164, "created_at": "2025-01-14T15:01:03.685495+00:00"} {"global_step": 1439, "acc_step": 0, "speed/wps": 12845.635291395418, "speed/FLOPS": 201758354004026.9, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2505335211753845, "optim/lr": 0.0008628, "optim/total_tokens": 377225216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310826, "loss/out": 3.9689550399780273, "created_at": "2025-01-14T15:01:08.791020+00:00"} {"global_step": 1440, "acc_step": 0, "speed/wps": 12843.80489897288, "speed/FLOPS": 201729605175808.16, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25445812940597534, "optim/lr": 0.0008634000000000001, "optim/total_tokens": 377487360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472720, "loss/out": 4.159793853759766, "created_at": "2025-01-14T15:01:13.897315+00:00"} {"global_step": 1441, "acc_step": 0, "speed/wps": 12845.503915239637, "speed/FLOPS": 201756290560970.6, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.32945430278778076, "optim/lr": 0.000864, "optim/total_tokens": 377749504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370572, "loss/out": 4.070425987243652, "created_at": "2025-01-14T15:01:19.004706+00:00"} {"global_step": 1442, "acc_step": 0, "speed/wps": 12836.03078320079, "speed/FLOPS": 201607502004852.94, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.309096097946167, "optim/lr": 0.0008646000000000001, "optim/total_tokens": 378011648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495151, "loss/out": 3.998281478881836, "created_at": "2025-01-14T15:01:24.119983+00:00"} {"global_step": 1443, "acc_step": 0, "speed/wps": 12848.108976171901, "speed/FLOPS": 201797206622640.66, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2580823004245758, "optim/lr": 0.0008652, "optim/total_tokens": 378273792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475057, "loss/out": 4.016047477722168, "created_at": "2025-01-14T15:01:29.227814+00:00"} {"global_step": 1444, "acc_step": 0, "speed/wps": 12843.105517120293, "speed/FLOPS": 201718620422761.94, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.28329935669898987, "optim/lr": 0.0008658000000000001, "optim/total_tokens": 378535936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492487, "loss/out": 3.927445411682129, "created_at": "2025-01-14T15:01:34.339504+00:00"} {"global_step": 1445, "acc_step": 0, "speed/wps": 12844.41772034606, "speed/FLOPS": 201739230377578.78, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2621583640575409, "optim/lr": 0.0008664, "optim/total_tokens": 378798080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469069, "loss/out": 4.001547813415527, "created_at": "2025-01-14T15:01:39.447658+00:00"} {"global_step": 1446, "acc_step": 0, "speed/wps": 12835.83404030041, "speed/FLOPS": 201604411887249.4, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.189748615026474, "optim/lr": 0.0008669999999999999, "optim/total_tokens": 379060224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486752, "loss/out": 4.0186309814453125, "created_at": "2025-01-14T15:01:44.557074+00:00"} {"global_step": 1447, "acc_step": 0, "speed/wps": 12844.313255252264, "speed/FLOPS": 201737589609730.62, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22621996700763702, "optim/lr": 0.0008676000000000001, "optim/total_tokens": 379322368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485422, "loss/out": 3.9351959228515625, "created_at": "2025-01-14T15:01:49.666090+00:00"} {"global_step": 1448, "acc_step": 0, "speed/wps": 12844.582671652312, "speed/FLOPS": 201741821164510.3, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20696143805980682, "optim/lr": 0.0008682, "optim/total_tokens": 379584512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330862, "loss/out": 3.8526902198791504, "created_at": "2025-01-14T15:01:54.773952+00:00"} {"global_step": 1449, "acc_step": 0, "speed/wps": 12843.91329189198, "speed/FLOPS": 201731307635550.12, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.21837663650512695, "optim/lr": 0.0008688000000000001, "optim/total_tokens": 379846656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468454, "loss/out": 3.915865421295166, "created_at": "2025-01-14T15:01:59.883110+00:00"} {"global_step": 1450, "acc_step": 0, "speed/wps": 12841.356061119923, "speed/FLOPS": 201691142812271.62, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.21201613545417786, "optim/lr": 0.0008694, "optim/total_tokens": 380108800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354145, "loss/out": 3.9929895401000977, "created_at": "2025-01-14T15:02:04.991475+00:00"} {"global_step": 1451, "acc_step": 0, "speed/wps": 12844.105666458598, "speed/FLOPS": 201734329142469.22, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.24304422736167908, "optim/lr": 0.00087, "optim/total_tokens": 380370944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490639, "loss/out": 4.013955593109131, "created_at": "2025-01-14T15:02:10.099301+00:00"} {"global_step": 1452, "acc_step": 0, "speed/wps": 12839.594533976891, "speed/FLOPS": 201663475607898.38, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.23590032756328583, "optim/lr": 0.0008706, "optim/total_tokens": 380633088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480094, "loss/out": 3.842738151550293, "created_at": "2025-01-14T15:02:15.210523+00:00"} {"global_step": 1453, "acc_step": 0, "speed/wps": 12838.393808132085, "speed/FLOPS": 201644616558535.47, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.29881808161735535, "optim/lr": 0.0008712, "optim/total_tokens": 380895232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474011, "loss/out": 3.887232542037964, "created_at": "2025-01-14T15:02:20.321428+00:00"} {"global_step": 1454, "acc_step": 0, "speed/wps": 12847.439329693241, "speed/FLOPS": 201786688904502.62, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2994272708892822, "optim/lr": 0.0008718, "optim/total_tokens": 381157376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493533, "loss/out": 3.889801025390625, "created_at": "2025-01-14T15:02:25.428488+00:00"} {"global_step": 1455, "acc_step": 0, "speed/wps": 12834.423492523496, "speed/FLOPS": 201582257296117.22, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.27641814947128296, "optim/lr": 0.0008724000000000001, "optim/total_tokens": 381419520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 343042, "loss/out": 3.960489273071289, "created_at": "2025-01-14T15:02:30.542958+00:00"} {"global_step": 1456, "acc_step": 0, "speed/wps": 12834.187565285645, "speed/FLOPS": 201578551734648.78, "speed/curr_iter_time": 1.2884, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31437814235687256, "optim/lr": 0.000873, "optim/total_tokens": 381681664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502556, "loss/out": 4.040684700012207, "created_at": "2025-01-14T15:02:35.655855+00:00"} {"global_step": 1457, "acc_step": 0, "speed/wps": 12841.85129414898, "speed/FLOPS": 201698921127514.5, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.3154464662075043, "optim/lr": 0.0008736000000000001, "optim/total_tokens": 381943808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465818, "loss/out": 4.18691349029541, "created_at": "2025-01-14T15:02:40.766299+00:00"} {"global_step": 1458, "acc_step": 0, "speed/wps": 12839.628002349526, "speed/FLOPS": 201664001274680.97, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.2956506609916687, "optim/lr": 0.0008742, "optim/total_tokens": 382205952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494342, "loss/out": 4.0114336013793945, "created_at": "2025-01-14T15:02:45.902983+00:00"} {"global_step": 1459, "acc_step": 0, "speed/wps": 12844.0522265473, "speed/FLOPS": 201733489795228.12, "speed/curr_iter_time": 1.2875, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3104873299598694, "optim/lr": 0.0008748000000000001, "optim/total_tokens": 382468096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504121, "loss/out": 3.9554641246795654, "created_at": "2025-01-14T15:02:51.012131+00:00"} {"global_step": 1460, "acc_step": 0, "speed/wps": 12842.018443478333, "speed/FLOPS": 201701546437419.22, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2795703709125519, "optim/lr": 0.0008754, "optim/total_tokens": 382730240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293678, "loss/out": 3.9114675521850586, "created_at": "2025-01-14T15:02:56.119036+00:00"} {"global_step": 1461, "acc_step": 0, "speed/wps": 12845.156198556013, "speed/FLOPS": 201750829192641.5, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.26722827553749084, "optim/lr": 0.0008759999999999999, "optim/total_tokens": 382992384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472161, "loss/out": 4.035360336303711, "created_at": "2025-01-14T15:03:01.225881+00:00"} {"global_step": 1462, "acc_step": 0, "speed/wps": 12845.744725896811, "speed/FLOPS": 201760072823250.03, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21201741695404053, "optim/lr": 0.0008766000000000001, "optim/total_tokens": 383254528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438536, "loss/out": 4.111055850982666, "created_at": "2025-01-14T15:03:06.332510+00:00"} {"global_step": 1463, "acc_step": 0, "speed/wps": 12842.821996068966, "speed/FLOPS": 201714167335052.06, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2147233933210373, "optim/lr": 0.0008772, "optim/total_tokens": 383516672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497825, "loss/out": 4.091217041015625, "created_at": "2025-01-14T15:03:11.439003+00:00"} {"global_step": 1464, "acc_step": 0, "speed/wps": 12846.762766214631, "speed/FLOPS": 201776062545375.4, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21634690463542938, "optim/lr": 0.0008778000000000001, "optim/total_tokens": 383778816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502273, "loss/out": 3.9300436973571777, "created_at": "2025-01-14T15:03:16.544322+00:00"} {"global_step": 1465, "acc_step": 0, "speed/wps": 12845.565668217014, "speed/FLOPS": 201757260476337.78, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25070273876190186, "optim/lr": 0.0008784, "optim/total_tokens": 384040960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481715, "loss/out": 3.956355094909668, "created_at": "2025-01-14T15:03:21.650347+00:00"} {"global_step": 1466, "acc_step": 0, "speed/wps": 12847.468804529568, "speed/FLOPS": 201787151847309.78, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24016033113002777, "optim/lr": 0.000879, "optim/total_tokens": 384303104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492487, "loss/out": 4.082815170288086, "created_at": "2025-01-14T15:03:26.759961+00:00"} {"global_step": 1467, "acc_step": 0, "speed/wps": 12847.136135630946, "speed/FLOPS": 201781926825121.97, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2634146809577942, "optim/lr": 0.0008796, "optim/total_tokens": 384565248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291562, "loss/out": 3.9689815044403076, "created_at": "2025-01-14T15:03:31.867783+00:00"} {"global_step": 1468, "acc_step": 0, "speed/wps": 12843.892308012797, "speed/FLOPS": 201730978054892.7, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.24769872426986694, "optim/lr": 0.0008802, "optim/total_tokens": 384827392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498399, "loss/out": 3.8843026161193848, "created_at": "2025-01-14T15:03:36.979142+00:00"} {"global_step": 1469, "acc_step": 0, "speed/wps": 12841.789827637025, "speed/FLOPS": 201697955711480.38, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.24221274256706238, "optim/lr": 0.0008808, "optim/total_tokens": 385089536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473491, "loss/out": 3.9150516986846924, "created_at": "2025-01-14T15:03:42.086386+00:00"} {"global_step": 1470, "acc_step": 0, "speed/wps": 12845.526062764719, "speed/FLOPS": 201756638418285.9, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22261327505111694, "optim/lr": 0.0008814000000000001, "optim/total_tokens": 385351680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496480, "loss/out": 4.079231262207031, "created_at": "2025-01-14T15:03:47.193072+00:00"} {"global_step": 1471, "acc_step": 0, "speed/wps": 12840.927420405627, "speed/FLOPS": 201684410420839.56, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.22902141511440277, "optim/lr": 0.000882, "optim/total_tokens": 385613824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495739, "loss/out": 3.9128546714782715, "created_at": "2025-01-14T15:03:52.301439+00:00"} {"global_step": 1472, "acc_step": 0, "speed/wps": 12843.293256676938, "speed/FLOPS": 201721569130480.22, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2288612574338913, "optim/lr": 0.0008826000000000001, "optim/total_tokens": 385875968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504647, "loss/out": 3.8310182094573975, "created_at": "2025-01-14T15:03:57.411214+00:00"} {"global_step": 1473, "acc_step": 0, "speed/wps": 12846.769279254071, "speed/FLOPS": 201776164841609.66, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19664297997951508, "optim/lr": 0.0008832, "optim/total_tokens": 386138112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491505, "loss/out": 4.002926826477051, "created_at": "2025-01-14T15:04:02.518768+00:00"} {"global_step": 1474, "acc_step": 0, "speed/wps": 12844.213206396955, "speed/FLOPS": 201736018204976.47, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20455345511436462, "optim/lr": 0.0008837999999999999, "optim/total_tokens": 386400256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470030, "loss/out": 4.002253532409668, "created_at": "2025-01-14T15:04:07.632774+00:00"} {"global_step": 1475, "acc_step": 0, "speed/wps": 12837.970955837325, "speed/FLOPS": 201637975082186.34, "speed/curr_iter_time": 1.2875, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20735809206962585, "optim/lr": 0.0008844, "optim/total_tokens": 386662400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475904, "loss/out": 3.9263811111450195, "created_at": "2025-01-14T15:04:12.742575+00:00"} {"global_step": 1476, "acc_step": 0, "speed/wps": 12841.977676377037, "speed/FLOPS": 201700906134073.5, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20459170639514923, "optim/lr": 0.0008849999999999999, "optim/total_tokens": 386924544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425228, "loss/out": 3.9700846672058105, "created_at": "2025-01-14T15:04:17.849788+00:00"} {"global_step": 1477, "acc_step": 0, "speed/wps": 12831.15441570075, "speed/FLOPS": 201530911952433.7, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.26053622364997864, "optim/lr": 0.0008856000000000001, "optim/total_tokens": 387186688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491505, "loss/out": 4.1055216789245605, "created_at": "2025-01-14T15:04:22.968844+00:00"} {"global_step": 1478, "acc_step": 0, "speed/wps": 12836.904602187606, "speed/FLOPS": 201621226532794.0, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.26975834369659424, "optim/lr": 0.0008862, "optim/total_tokens": 387448832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487798, "loss/out": 3.8630716800689697, "created_at": "2025-01-14T15:04:28.077685+00:00"} {"global_step": 1479, "acc_step": 0, "speed/wps": 12838.399223520342, "speed/FLOPS": 201644701614649.6, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22919616103172302, "optim/lr": 0.0008868, "optim/total_tokens": 387710976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336083, "loss/out": 3.835935354232788, "created_at": "2025-01-14T15:04:33.189783+00:00"} {"global_step": 1480, "acc_step": 0, "speed/wps": 12839.060492691284, "speed/FLOPS": 201655087755658.28, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.24611902236938477, "optim/lr": 0.0008874, "optim/total_tokens": 387973120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496999, "loss/out": 4.011736869812012, "created_at": "2025-01-14T15:04:38.300463+00:00"} {"global_step": 1481, "acc_step": 0, "speed/wps": 12838.242771702377, "speed/FLOPS": 201642244323861.1, "speed/curr_iter_time": 1.2874, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.26127514243125916, "optim/lr": 0.000888, "optim/total_tokens": 388235264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497305, "loss/out": 3.915653944015503, "created_at": "2025-01-14T15:04:43.411917+00:00"} {"global_step": 1482, "acc_step": 0, "speed/wps": 12834.946138211264, "speed/FLOPS": 201590466164834.06, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2316238284111023, "optim/lr": 0.0008886, "optim/total_tokens": 388497408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310462, "loss/out": 3.8846259117126465, "created_at": "2025-01-14T15:04:48.521827+00:00"} {"global_step": 1483, "acc_step": 0, "speed/wps": 12836.635079261894, "speed/FLOPS": 201616993304882.4, "speed/curr_iter_time": 1.2876, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23303307592868805, "optim/lr": 0.0008892, "optim/total_tokens": 388759552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463700, "loss/out": 3.9916512966156006, "created_at": "2025-01-14T15:04:53.633265+00:00"} {"global_step": 1484, "acc_step": 0, "speed/wps": 12838.927662777005, "speed/FLOPS": 201653001479327.1, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22031956911087036, "optim/lr": 0.0008897999999999999, "optim/total_tokens": 389021696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 286758, "loss/out": 4.059222221374512, "created_at": "2025-01-14T15:04:58.743925+00:00"} {"global_step": 1485, "acc_step": 0, "speed/wps": 12839.202323235952, "speed/FLOPS": 201657315399257.88, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21957707405090332, "optim/lr": 0.0008904000000000001, "optim/total_tokens": 389283840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470067, "loss/out": 3.993152618408203, "created_at": "2025-01-14T15:05:03.858289+00:00"} {"global_step": 1486, "acc_step": 0, "speed/wps": 12837.242303361794, "speed/FLOPS": 201626530593784.7, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24471421539783478, "optim/lr": 0.000891, "optim/total_tokens": 389545984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411887, "loss/out": 3.9617021083831787, "created_at": "2025-01-14T15:05:08.970221+00:00"} {"global_step": 1487, "acc_step": 0, "speed/wps": 12826.135860578954, "speed/FLOPS": 201452088648028.47, "speed/curr_iter_time": 1.2875, "speed/data_load_time": 0.001, "optim/grad_norm": 0.19305658340454102, "optim/lr": 0.0008916000000000001, "optim/total_tokens": 389808128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486988, "loss/out": 3.9595437049865723, "created_at": "2025-01-14T15:05:14.085734+00:00"} {"global_step": 1488, "acc_step": 0, "speed/wps": 12835.569626480534, "speed/FLOPS": 201600258904865.6, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23437510430812836, "optim/lr": 0.0008922, "optim/total_tokens": 390070272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480621, "loss/out": 3.8847079277038574, "created_at": "2025-01-14T15:05:19.196647+00:00"} {"global_step": 1489, "acc_step": 0, "speed/wps": 12842.853866306556, "speed/FLOPS": 201714667900927.72, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.27405399084091187, "optim/lr": 0.0008927999999999999, "optim/total_tokens": 390332416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355021, "loss/out": 3.9542946815490723, "created_at": "2025-01-14T15:05:24.306095+00:00"} {"global_step": 1490, "acc_step": 0, "speed/wps": 12833.80457447445, "speed/FLOPS": 201572536337674.44, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3452337682247162, "optim/lr": 0.0008934, "optim/total_tokens": 390594560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504121, "loss/out": 3.860532760620117, "created_at": "2025-01-14T15:05:29.416508+00:00"} {"global_step": 1491, "acc_step": 0, "speed/wps": 12838.944219516066, "speed/FLOPS": 201653261525665.4, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2808831036090851, "optim/lr": 0.0008939999999999999, "optim/total_tokens": 390856704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484611, "loss/out": 3.944718360900879, "created_at": "2025-01-14T15:05:34.524638+00:00"} {"global_step": 1492, "acc_step": 0, "speed/wps": 12841.893065840177, "speed/FLOPS": 201699577209325.03, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.315168559551239, "optim/lr": 0.0008946000000000001, "optim/total_tokens": 391118848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469257, "loss/out": 3.972672462463379, "created_at": "2025-01-14T15:05:39.633640+00:00"} {"global_step": 1493, "acc_step": 0, "speed/wps": 12839.28508278886, "speed/FLOPS": 201658615251759.84, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2986621856689453, "optim/lr": 0.0008952, "optim/total_tokens": 391380992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484320, "loss/out": 3.9118330478668213, "created_at": "2025-01-14T15:05:44.742137+00:00"} {"global_step": 1494, "acc_step": 0, "speed/wps": 12837.54417344043, "speed/FLOPS": 201631271878182.62, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2916112542152405, "optim/lr": 0.0008958, "optim/total_tokens": 391643136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363018, "loss/out": 4.015806198120117, "created_at": "2025-01-14T15:05:49.853427+00:00"} {"global_step": 1495, "acc_step": 0, "speed/wps": 12832.666937051426, "speed/FLOPS": 201554668178668.2, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.3440952003002167, "optim/lr": 0.0008964, "optim/total_tokens": 391905280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503075, "loss/out": 3.9697580337524414, "created_at": "2025-01-14T15:05:54.966586+00:00"} {"global_step": 1496, "acc_step": 0, "speed/wps": 12836.234334206894, "speed/FLOPS": 201610699053112.22, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3599347472190857, "optim/lr": 0.000897, "optim/total_tokens": 392167424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499445, "loss/out": 3.858677625656128, "created_at": "2025-01-14T15:06:00.095677+00:00"} {"global_step": 1497, "acc_step": 0, "speed/wps": 12841.628606218585, "speed/FLOPS": 201695423507562.34, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.3343978524208069, "optim/lr": 0.0008976000000000001, "optim/total_tokens": 392429568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304748, "loss/out": 3.9272518157958984, "created_at": "2025-01-14T15:06:05.206156+00:00"} {"global_step": 1498, "acc_step": 0, "speed/wps": 12845.97068022722, "speed/FLOPS": 201763621746502.9, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.28812935948371887, "optim/lr": 0.0008982, "optim/total_tokens": 392691712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482471, "loss/out": 3.9967994689941406, "created_at": "2025-01-14T15:06:10.312705+00:00"} {"global_step": 1499, "acc_step": 0, "speed/wps": 12840.753438711401, "speed/FLOPS": 201681677799255.84, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22956490516662598, "optim/lr": 0.0008988, "optim/total_tokens": 392953856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 283057, "loss/out": 3.9234237670898438, "created_at": "2025-01-14T15:06:15.423382+00:00"} {"global_step": 1500, "acc_step": 0, "speed/wps": 12826.505111856883, "speed/FLOPS": 201457888246752.78, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23608629405498505, "optim/lr": 0.0008994000000000001, "optim/total_tokens": 393216000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495387, "loss/out": 3.99153470993042, "created_at": "2025-01-14T15:06:20.540165+00:00"} {"global_step": 1501, "acc_step": 0, "speed/wps": 12841.214161898308, "speed/FLOPS": 201688914090006.22, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.223192498087883, "optim/lr": 0.0009, "optim/total_tokens": 393478144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482235, "loss/out": 4.004619121551514, "created_at": "2025-01-14T15:06:25.648476+00:00"} {"global_step": 1502, "acc_step": 0, "speed/wps": 12840.197680209258, "speed/FLOPS": 201672948848287.47, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21454574167728424, "optim/lr": 0.0009006000000000001, "optim/total_tokens": 393740288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483281, "loss/out": 3.925856113433838, "created_at": "2025-01-14T15:06:30.757694+00:00"} {"global_step": 1503, "acc_step": 0, "speed/wps": 12842.143495451797, "speed/FLOPS": 201703510550501.88, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20219005644321442, "optim/lr": 0.0009012, "optim/total_tokens": 394002432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470587, "loss/out": 3.878668785095215, "created_at": "2025-01-14T15:06:35.865513+00:00"} {"global_step": 1504, "acc_step": 0, "speed/wps": 12837.234696956892, "speed/FLOPS": 201626411124743.4, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2664683163166046, "optim/lr": 0.0009017999999999999, "optim/total_tokens": 394264576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362152, "loss/out": 3.8876495361328125, "created_at": "2025-01-14T15:06:40.975280+00:00"} {"global_step": 1505, "acc_step": 0, "speed/wps": 12844.990142232191, "speed/FLOPS": 201748221049890.56, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22809229791164398, "optim/lr": 0.0009024, "optim/total_tokens": 394526720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498564, "loss/out": 4.013123989105225, "created_at": "2025-01-14T15:06:46.081723+00:00"} {"global_step": 1506, "acc_step": 0, "speed/wps": 12842.67176719166, "speed/FLOPS": 201711807784097.88, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3274937570095062, "optim/lr": 0.0009029999999999999, "optim/total_tokens": 394788864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495975, "loss/out": 3.8708558082580566, "created_at": "2025-01-14T15:06:51.188329+00:00"} {"global_step": 1507, "acc_step": 0, "speed/wps": 12844.636596615268, "speed/FLOPS": 201742668130154.2, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.35660889744758606, "optim/lr": 0.0009036000000000001, "optim/total_tokens": 395051008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468454, "loss/out": 4.022948265075684, "created_at": "2025-01-14T15:06:56.295304+00:00"} {"global_step": 1508, "acc_step": 0, "speed/wps": 12838.344629495268, "speed/FLOPS": 201643844140465.75, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.001, "optim/grad_norm": 0.27422669529914856, "optim/lr": 0.0009042, "optim/total_tokens": 395313152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484846, "loss/out": 4.000986576080322, "created_at": "2025-01-14T15:07:01.408526+00:00"} {"global_step": 1509, "acc_step": 0, "speed/wps": 12844.77425737615, "speed/FLOPS": 201744830281569.56, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.29316142201423645, "optim/lr": 0.0009048, "optim/total_tokens": 395575296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304917, "loss/out": 4.049122333526611, "created_at": "2025-01-14T15:07:06.514967+00:00"} {"global_step": 1510, "acc_step": 0, "speed/wps": 12843.614450971241, "speed/FLOPS": 201726613928239.47, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23782990872859955, "optim/lr": 0.0009054000000000001, "optim/total_tokens": 395837440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465054, "loss/out": 3.9595868587493896, "created_at": "2025-01-14T15:07:11.627731+00:00"} {"global_step": 1511, "acc_step": 0, "speed/wps": 12846.036546279325, "speed/FLOPS": 201764656263360.66, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.259622722864151, "optim/lr": 0.000906, "optim/total_tokens": 396099584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321933, "loss/out": 3.937072277069092, "created_at": "2025-01-14T15:07:16.732889+00:00"} {"global_step": 1512, "acc_step": 0, "speed/wps": 12846.47504750182, "speed/FLOPS": 201771543527623.78, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.27764472365379333, "optim/lr": 0.0009066000000000001, "optim/total_tokens": 396361728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470304, "loss/out": 3.896735429763794, "created_at": "2025-01-14T15:07:21.840689+00:00"} {"global_step": 1513, "acc_step": 0, "speed/wps": 12840.986052712906, "speed/FLOPS": 201685331321794.6, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21544234454631805, "optim/lr": 0.0009072, "optim/total_tokens": 396623872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360804, "loss/out": 3.9188575744628906, "created_at": "2025-01-14T15:07:26.949923+00:00"} {"global_step": 1514, "acc_step": 0, "speed/wps": 12845.303350542306, "speed/FLOPS": 201753140416794.25, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23511171340942383, "optim/lr": 0.0009078, "optim/total_tokens": 396886016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488785, "loss/out": 3.9339261054992676, "created_at": "2025-01-14T15:07:32.055761+00:00"} {"global_step": 1515, "acc_step": 0, "speed/wps": 12844.070786873197, "speed/FLOPS": 201733781310650.9, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2300996035337448, "optim/lr": 0.0009084000000000001, "optim/total_tokens": 397148160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471924, "loss/out": 3.887345552444458, "created_at": "2025-01-14T15:07:37.162166+00:00"} {"global_step": 1516, "acc_step": 0, "speed/wps": 12841.979366114925, "speed/FLOPS": 201700932673728.97, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21548672020435333, "optim/lr": 0.000909, "optim/total_tokens": 397410304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479815, "loss/out": 3.95538330078125, "created_at": "2025-01-14T15:07:42.273248+00:00"} {"global_step": 1517, "acc_step": 0, "speed/wps": 12844.436157469772, "speed/FLOPS": 201739519957941.94, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22940802574157715, "optim/lr": 0.0009096000000000001, "optim/total_tokens": 397672448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492268, "loss/out": 4.002301216125488, "created_at": "2025-01-14T15:07:47.379707+00:00"} {"global_step": 1518, "acc_step": 0, "speed/wps": 12848.314550223342, "speed/FLOPS": 201800435445606.88, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21244709193706512, "optim/lr": 0.0009102, "optim/total_tokens": 397934592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291276, "loss/out": 3.891324043273926, "created_at": "2025-01-14T15:07:52.484070+00:00"} {"global_step": 1519, "acc_step": 0, "speed/wps": 12842.43243580639, "speed/FLOPS": 201708048755816.4, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2286466658115387, "optim/lr": 0.0009107999999999999, "optim/total_tokens": 398196736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486468, "loss/out": 3.9452548027038574, "created_at": "2025-01-14T15:07:57.593588+00:00"} {"global_step": 1520, "acc_step": 0, "speed/wps": 12846.536908317963, "speed/FLOPS": 201772515136746.94, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19878548383712769, "optim/lr": 0.0009114, "optim/total_tokens": 398458880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486468, "loss/out": 3.8629865646362305, "created_at": "2025-01-14T15:08:02.700641+00:00"} {"global_step": 1521, "acc_step": 0, "speed/wps": 12846.7690236107, "speed/FLOPS": 201776160826379.22, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.001, "optim/grad_norm": 0.23226088285446167, "optim/lr": 0.000912, "optim/total_tokens": 398721024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482761, "loss/out": 3.9476053714752197, "created_at": "2025-01-14T15:08:07.805875+00:00"} {"global_step": 1522, "acc_step": 0, "speed/wps": 12845.491452160903, "speed/FLOPS": 201756094811193.0, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19953538477420807, "optim/lr": 0.0009126000000000001, "optim/total_tokens": 398983168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494402, "loss/out": 3.8994357585906982, "created_at": "2025-01-14T15:08:12.912554+00:00"} {"global_step": 1523, "acc_step": 0, "speed/wps": 12843.971196595287, "speed/FLOPS": 201732217108484.94, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19311553239822388, "optim/lr": 0.0009132000000000001, "optim/total_tokens": 399245312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 4.008150100708008, "created_at": "2025-01-14T15:08:18.018575+00:00"} {"global_step": 1524, "acc_step": 0, "speed/wps": 12846.73274111785, "speed/FLOPS": 201775590959971.66, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21692019701004028, "optim/lr": 0.0009138, "optim/total_tokens": 399507456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474937, "loss/out": 3.9565210342407227, "created_at": "2025-01-14T15:08:23.123469+00:00"} {"global_step": 1525, "acc_step": 0, "speed/wps": 12844.466500064033, "speed/FLOPS": 201739996530079.9, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19972549378871918, "optim/lr": 0.0009144000000000001, "optim/total_tokens": 399769600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498854, "loss/out": 3.8077313899993896, "created_at": "2025-01-14T15:08:28.230858+00:00"} {"global_step": 1526, "acc_step": 0, "speed/wps": 12844.617577463605, "speed/FLOPS": 201742369408242.28, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2317284345626831, "optim/lr": 0.000915, "optim/total_tokens": 400031744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502626, "loss/out": 3.9544055461883545, "created_at": "2025-01-14T15:08:33.340691+00:00"} {"global_step": 1527, "acc_step": 0, "speed/wps": 12847.49880726527, "speed/FLOPS": 201787623081502.1, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2797956168651581, "optim/lr": 0.0009156000000000001, "optim/total_tokens": 400293888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480621, "loss/out": 3.8786532878875732, "created_at": "2025-01-14T15:08:38.446215+00:00"} {"global_step": 1528, "acc_step": 0, "speed/wps": 12840.631438764312, "speed/FLOPS": 201679761622441.3, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.34502333402633667, "optim/lr": 0.0009162, "optim/total_tokens": 400556032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490695, "loss/out": 3.874093532562256, "created_at": "2025-01-14T15:08:43.555747+00:00"} {"global_step": 1529, "acc_step": 0, "speed/wps": 12842.176070111887, "speed/FLOPS": 201704022180300.88, "speed/curr_iter_time": 1.2874, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3223743438720703, "optim/lr": 0.0009168, "optim/total_tokens": 400818176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333311, "loss/out": 4.002134323120117, "created_at": "2025-01-14T15:08:48.665397+00:00"} {"global_step": 1530, "acc_step": 0, "speed/wps": 12845.993966720242, "speed/FLOPS": 201763987492874.8, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.29727527499198914, "optim/lr": 0.0009174000000000001, "optim/total_tokens": 401080320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481431, "loss/out": 3.8035550117492676, "created_at": "2025-01-14T15:08:53.773557+00:00"} {"global_step": 1531, "acc_step": 0, "speed/wps": 12846.328916969656, "speed/FLOPS": 201769248346811.72, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.37174874544143677, "optim/lr": 0.000918, "optim/total_tokens": 401342464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 3.9729723930358887, "created_at": "2025-01-14T15:08:58.880365+00:00"} {"global_step": 1532, "acc_step": 0, "speed/wps": 12835.78349285614, "speed/FLOPS": 201603617970177.7, "speed/curr_iter_time": 1.2878, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.30386555194854736, "optim/lr": 0.0009186000000000001, "optim/total_tokens": 401604608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483754, "loss/out": 3.8839619159698486, "created_at": "2025-01-14T15:09:03.992344+00:00"} {"global_step": 1533, "acc_step": 0, "speed/wps": 12842.909869895358, "speed/FLOPS": 201715547514246.88, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23945413529872894, "optim/lr": 0.0009192, "optim/total_tokens": 401866752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471884, "loss/out": 3.9873545169830322, "created_at": "2025-01-14T15:09:09.104594+00:00"} {"global_step": 1534, "acc_step": 0, "speed/wps": 12846.6355064548, "speed/FLOPS": 201774063755974.03, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19708700478076935, "optim/lr": 0.0009198, "optim/total_tokens": 402128896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313679, "loss/out": 3.933610677719116, "created_at": "2025-01-14T15:09:14.209578+00:00"} {"global_step": 1535, "acc_step": 0, "speed/wps": 12846.420727174818, "speed/FLOPS": 201770690352244.25, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24023590981960297, "optim/lr": 0.0009204, "optim/total_tokens": 402391040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472964, "loss/out": 4.035461902618408, "created_at": "2025-01-14T15:09:19.315080+00:00"} {"global_step": 1536, "acc_step": 0, "speed/wps": 12842.489999027446, "speed/FLOPS": 201708952865303.28, "speed/curr_iter_time": 1.2876, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2502189874649048, "optim/lr": 0.000921, "optim/total_tokens": 402653184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361964, "loss/out": 3.9471888542175293, "created_at": "2025-01-14T15:09:24.423119+00:00"} {"global_step": 1537, "acc_step": 0, "speed/wps": 12847.948607081007, "speed/FLOPS": 201794687805697.94, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2696141004562378, "optim/lr": 0.0009216, "optim/total_tokens": 402915328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465030, "loss/out": 3.958893299102783, "created_at": "2025-01-14T15:09:29.536490+00:00"} {"global_step": 1538, "acc_step": 0, "speed/wps": 12846.157415419499, "speed/FLOPS": 201766554679298.7, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2818089723587036, "optim/lr": 0.0009222000000000001, "optim/total_tokens": 403177472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309482, "loss/out": 3.9050116539001465, "created_at": "2025-01-14T15:09:34.642465+00:00"} {"global_step": 1539, "acc_step": 0, "speed/wps": 12849.045870272492, "speed/FLOPS": 201811921831917.84, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.26916301250457764, "optim/lr": 0.0009228, "optim/total_tokens": 403439616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474537, "loss/out": 3.8323869705200195, "created_at": "2025-01-14T15:09:39.747936+00:00"} {"global_step": 1540, "acc_step": 0, "speed/wps": 12844.156764760279, "speed/FLOPS": 201735131711513.53, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.3036695420742035, "optim/lr": 0.0009234000000000001, "optim/total_tokens": 403701760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443311, "loss/out": 4.051164627075195, "created_at": "2025-01-14T15:09:44.856403+00:00"} {"global_step": 1541, "acc_step": 0, "speed/wps": 12843.434847887394, "speed/FLOPS": 201723793015007.16, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2609504759311676, "optim/lr": 0.000924, "optim/total_tokens": 403963904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498564, "loss/out": 3.9465675354003906, "created_at": "2025-01-14T15:09:49.963078+00:00"} {"global_step": 1542, "acc_step": 0, "speed/wps": 12841.651220796617, "speed/FLOPS": 201695778700586.03, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23320147395133972, "optim/lr": 0.0009245999999999999, "optim/total_tokens": 404226048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496999, "loss/out": 3.9041879177093506, "created_at": "2025-01-14T15:09:55.070313+00:00"} {"global_step": 1543, "acc_step": 0, "speed/wps": 12842.809158480995, "speed/FLOPS": 201713965703092.3, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20923574268817902, "optim/lr": 0.0009252, "optim/total_tokens": 404488192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478197, "loss/out": 4.010222434997559, "created_at": "2025-01-14T15:10:00.176998+00:00"} {"global_step": 1544, "acc_step": 0, "speed/wps": 12846.407962128798, "speed/FLOPS": 201770489859655.5, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2024526596069336, "optim/lr": 0.0009258, "optim/total_tokens": 404750336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494106, "loss/out": 3.890511989593506, "created_at": "2025-01-14T15:10:05.290399+00:00"} {"global_step": 1545, "acc_step": 0, "speed/wps": 12844.615307923626, "speed/FLOPS": 201742333761998.25, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20853854715824127, "optim/lr": 0.0009264000000000001, "optim/total_tokens": 405012480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 282237, "loss/out": 3.8860931396484375, "created_at": "2025-01-14T15:10:10.396213+00:00"} {"global_step": 1546, "acc_step": 0, "speed/wps": 12847.78506119233, "speed/FLOPS": 201792119092780.66, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21721866726875305, "optim/lr": 0.000927, "optim/total_tokens": 405274624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476387, "loss/out": 3.8669915199279785, "created_at": "2025-01-14T15:10:15.505828+00:00"} {"global_step": 1547, "acc_step": 0, "speed/wps": 12847.466885822912, "speed/FLOPS": 201787121711385.2, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22485195100307465, "optim/lr": 0.0009276, "optim/total_tokens": 405536768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481951, "loss/out": 3.8681626319885254, "created_at": "2025-01-14T15:10:20.612693+00:00"} {"global_step": 1548, "acc_step": 0, "speed/wps": 12839.907758008181, "speed/FLOPS": 201668395221724.72, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18009741604328156, "optim/lr": 0.0009282, "optim/total_tokens": 405798912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486752, "loss/out": 3.95530366897583, "created_at": "2025-01-14T15:10:25.720471+00:00"} {"global_step": 1549, "acc_step": 0, "speed/wps": 12843.81155670266, "speed/FLOPS": 201729709744603.03, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18216688930988312, "optim/lr": 0.0009288, "optim/total_tokens": 406061056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478528, "loss/out": 3.939824104309082, "created_at": "2025-01-14T15:10:30.828445+00:00"} {"global_step": 1550, "acc_step": 0, "speed/wps": 12845.54553717359, "speed/FLOPS": 201756944290637.88, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.17388667166233063, "optim/lr": 0.0009294, "optim/total_tokens": 406323200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502556, "loss/out": 3.9894707202911377, "created_at": "2025-01-14T15:10:35.935446+00:00"} {"global_step": 1551, "acc_step": 0, "speed/wps": 12845.536757854201, "speed/FLOPS": 201756806399362.8, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1920613795518875, "optim/lr": 0.00093, "optim/total_tokens": 406585344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468737, "loss/out": 3.9383504390716553, "created_at": "2025-01-14T15:10:41.040914+00:00"} {"global_step": 1552, "acc_step": 0, "speed/wps": 12840.826238639549, "speed/FLOPS": 201682821222165.0, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21972770988941193, "optim/lr": 0.0009306, "optim/total_tokens": 406847488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501747, "loss/out": 3.9290518760681152, "created_at": "2025-01-14T15:10:46.150300+00:00"} {"global_step": 1553, "acc_step": 0, "speed/wps": 12843.038949796477, "speed/FLOPS": 201717574891468.3, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.27468547224998474, "optim/lr": 0.0009312000000000001, "optim/total_tokens": 407109632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497525, "loss/out": 4.066803932189941, "created_at": "2025-01-14T15:10:51.259152+00:00"} {"global_step": 1554, "acc_step": 0, "speed/wps": 12844.75637730066, "speed/FLOPS": 201744549450414.2, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.28579360246658325, "optim/lr": 0.0009318, "optim/total_tokens": 407371776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328886, "loss/out": 3.8635380268096924, "created_at": "2025-01-14T15:10:56.366647+00:00"} {"global_step": 1555, "acc_step": 0, "speed/wps": 12844.40241082145, "speed/FLOPS": 201738989920457.3, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2635902464389801, "optim/lr": 0.0009324000000000001, "optim/total_tokens": 407633920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486468, "loss/out": 3.853168487548828, "created_at": "2025-01-14T15:11:01.476855+00:00"} {"global_step": 1556, "acc_step": 0, "speed/wps": 12844.141488679237, "speed/FLOPS": 201734891779669.25, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22884084284305573, "optim/lr": 0.000933, "optim/total_tokens": 407896064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415893, "loss/out": 3.925593614578247, "created_at": "2025-01-14T15:11:06.585272+00:00"} {"global_step": 1557, "acc_step": 0, "speed/wps": 12842.311840656243, "speed/FLOPS": 201706154643268.25, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2057962268590927, "optim/lr": 0.0009335999999999999, "optim/total_tokens": 408158208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489648, "loss/out": 3.820669174194336, "created_at": "2025-01-14T15:11:11.693038+00:00"} {"global_step": 1558, "acc_step": 0, "speed/wps": 12842.446102337119, "speed/FLOPS": 201708263407461.3, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.001, "optim/grad_norm": 0.20874294638633728, "optim/lr": 0.0009342, "optim/total_tokens": 408420352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494342, "loss/out": 3.8763935565948486, "created_at": "2025-01-14T15:11:16.802016+00:00"} {"global_step": 1559, "acc_step": 0, "speed/wps": 12843.008388924465, "speed/FLOPS": 201717094890978.3, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21965298056602478, "optim/lr": 0.0009348, "optim/total_tokens": 408682496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315260, "loss/out": 3.9686856269836426, "created_at": "2025-01-14T15:11:21.910879+00:00"} {"global_step": 1560, "acc_step": 0, "speed/wps": 12844.464783881951, "speed/FLOPS": 201739969575082.03, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21022634208202362, "optim/lr": 0.0009354000000000001, "optim/total_tokens": 408944640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469529, "loss/out": 4.089876651763916, "created_at": "2025-01-14T15:11:27.018475+00:00"} {"global_step": 1561, "acc_step": 0, "speed/wps": 12842.652781984414, "speed/FLOPS": 201711509595329.66, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.001, "optim/grad_norm": 0.23345611989498138, "optim/lr": 0.000936, "optim/total_tokens": 409206784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451797, "loss/out": 3.8984181880950928, "created_at": "2025-01-14T15:11:32.124991+00:00"} {"global_step": 1562, "acc_step": 0, "speed/wps": 12841.699827176342, "speed/FLOPS": 201696542130571.78, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.24270015954971313, "optim/lr": 0.0009366, "optim/total_tokens": 409468928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502342, "loss/out": 3.9078173637390137, "created_at": "2025-01-14T15:11:37.231916+00:00"} {"global_step": 1563, "acc_step": 0, "speed/wps": 12839.25723612296, "speed/FLOPS": 201658177881606.56, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.24809134006500244, "optim/lr": 0.0009372, "optim/total_tokens": 409731072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413539, "loss/out": 3.907876968383789, "created_at": "2025-01-14T15:11:42.341042+00:00"} {"global_step": 1564, "acc_step": 0, "speed/wps": 12841.667355702271, "speed/FLOPS": 201696032121451.0, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2241031676530838, "optim/lr": 0.0009378, "optim/total_tokens": 409993216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484091, "loss/out": 3.9057579040527344, "created_at": "2025-01-14T15:11:47.448494+00:00"} {"global_step": 1565, "acc_step": 0, "speed/wps": 12844.272700990623, "speed/FLOPS": 201736952649324.16, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21718250215053558, "optim/lr": 0.0009384, "optim/total_tokens": 410255360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484091, "loss/out": 3.9068217277526855, "created_at": "2025-01-14T15:11:52.556064+00:00"} {"global_step": 1566, "acc_step": 0, "speed/wps": 12840.270443377105, "speed/FLOPS": 201674091693825.6, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20533300936222076, "optim/lr": 0.0009390000000000001, "optim/total_tokens": 410517504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480094, "loss/out": 3.8008906841278076, "created_at": "2025-01-14T15:11:57.666435+00:00"} {"global_step": 1567, "acc_step": 0, "speed/wps": 12839.206961565824, "speed/FLOPS": 201657388250602.25, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20058588683605194, "optim/lr": 0.0009396, "optim/total_tokens": 410779648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501747, "loss/out": 3.810062885284424, "created_at": "2025-01-14T15:12:02.779623+00:00"} {"global_step": 1568, "acc_step": 0, "speed/wps": 12845.198111333952, "speed/FLOPS": 201751487490413.22, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2276664674282074, "optim/lr": 0.0009402000000000001, "optim/total_tokens": 411041792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321933, "loss/out": 3.937380790710449, "created_at": "2025-01-14T15:12:07.885437+00:00"} {"global_step": 1569, "acc_step": 0, "speed/wps": 12829.864951332836, "speed/FLOPS": 201510659142625.88, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21261276304721832, "optim/lr": 0.0009408, "optim/total_tokens": 411303936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502273, "loss/out": 3.906914710998535, "created_at": "2025-01-14T15:12:12.998696+00:00"} {"global_step": 1570, "acc_step": 0, "speed/wps": 12844.531696566683, "speed/FLOPS": 201741020530743.4, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24787184596061707, "optim/lr": 0.0009414000000000001, "optim/total_tokens": 411566080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487272, "loss/out": 3.9453024864196777, "created_at": "2025-01-14T15:12:18.105792+00:00"} {"global_step": 1571, "acc_step": 0, "speed/wps": 12841.186888005874, "speed/FLOPS": 201688485716047.28, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.255044549703598, "optim/lr": 0.000942, "optim/total_tokens": 411828224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485941, "loss/out": 3.8847312927246094, "created_at": "2025-01-14T15:12:23.214567+00:00"} {"global_step": 1572, "acc_step": 0, "speed/wps": 12844.904511922337, "speed/FLOPS": 201746876108205.94, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24119991064071655, "optim/lr": 0.0009425999999999999, "optim/total_tokens": 412090368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504647, "loss/out": 3.8456485271453857, "created_at": "2025-01-14T15:12:28.324484+00:00"} {"global_step": 1573, "acc_step": 0, "speed/wps": 12842.483973146533, "speed/FLOPS": 201708858220563.1, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31005120277404785, "optim/lr": 0.0009432, "optim/total_tokens": 412352512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 285653, "loss/out": 3.8599648475646973, "created_at": "2025-01-14T15:12:33.440029+00:00"} {"global_step": 1574, "acc_step": 0, "speed/wps": 12841.355334673408, "speed/FLOPS": 201691131402430.88, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3737517297267914, "optim/lr": 0.0009438, "optim/total_tokens": 412614656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488845, "loss/out": 3.918471574783325, "created_at": "2025-01-14T15:12:38.551879+00:00"} {"global_step": 1575, "acc_step": 0, "speed/wps": 12839.038039501493, "speed/FLOPS": 201654735097458.7, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3179127275943756, "optim/lr": 0.0009444000000000001, "optim/total_tokens": 412876800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497305, "loss/out": 3.9424571990966797, "created_at": "2025-01-14T15:12:43.661625+00:00"} {"global_step": 1576, "acc_step": 0, "speed/wps": 12840.302101798006, "speed/FLOPS": 201674588932829.06, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.24266447126865387, "optim/lr": 0.000945, "optim/total_tokens": 413138944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484320, "loss/out": 3.8969221115112305, "created_at": "2025-01-14T15:12:48.772069+00:00"} {"global_step": 1577, "acc_step": 0, "speed/wps": 12843.265845028056, "speed/FLOPS": 201721138592866.97, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24681085348129272, "optim/lr": 0.0009456, "optim/total_tokens": 413401088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485366, "loss/out": 3.8698089122772217, "created_at": "2025-01-14T15:12:53.880839+00:00"} {"global_step": 1578, "acc_step": 0, "speed/wps": 12841.391959397055, "speed/FLOPS": 201691706644043.4, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24363508820533752, "optim/lr": 0.0009462, "optim/total_tokens": 413663232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 3.9735851287841797, "created_at": "2025-01-14T15:12:58.987854+00:00"} {"global_step": 1579, "acc_step": 0, "speed/wps": 12842.395751241085, "speed/FLOPS": 201707472574308.44, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.262290358543396, "optim/lr": 0.0009468, "optim/total_tokens": 413925376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499373, "loss/out": 3.8737409114837646, "created_at": "2025-01-14T15:13:04.095902+00:00"} {"global_step": 1580, "acc_step": 0, "speed/wps": 12840.979950916377, "speed/FLOPS": 201685235484695.38, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.219131737947464, "optim/lr": 0.0009474, "optim/total_tokens": 414187520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292143, "loss/out": 3.8968920707702637, "created_at": "2025-01-14T15:13:09.203991+00:00"} {"global_step": 1581, "acc_step": 0, "speed/wps": 12839.945812991724, "speed/FLOPS": 201668992927534.28, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19686569273471832, "optim/lr": 0.0009480000000000001, "optim/total_tokens": 414449664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479054, "loss/out": 3.9555845260620117, "created_at": "2025-01-14T15:13:14.312753+00:00"} {"global_step": 1582, "acc_step": 0, "speed/wps": 12841.410488601334, "speed/FLOPS": 201691997670658.4, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2045433074235916, "optim/lr": 0.0009486, "optim/total_tokens": 414711808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479338, "loss/out": 3.877927303314209, "created_at": "2025-01-14T15:13:19.420945+00:00"} {"global_step": 1583, "acc_step": 0, "speed/wps": 12847.947797786022, "speed/FLOPS": 201794675094608.12, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2083146721124649, "optim/lr": 0.0009492000000000001, "optim/total_tokens": 414973952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494402, "loss/out": 3.961421012878418, "created_at": "2025-01-14T15:13:24.525712+00:00"} {"global_step": 1584, "acc_step": 0, "speed/wps": 12843.900454466522, "speed/FLOPS": 201731106006142.84, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22542349994182587, "optim/lr": 0.0009498, "optim/total_tokens": 415236096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502273, "loss/out": 3.923327922821045, "created_at": "2025-01-14T15:13:29.637420+00:00"} {"global_step": 1585, "acc_step": 0, "speed/wps": 12837.777654851243, "speed/FLOPS": 201634939024576.53, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22414788603782654, "optim/lr": 0.0009504000000000001, "optim/total_tokens": 415498240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295368, "loss/out": 3.854975700378418, "created_at": "2025-01-14T15:13:34.749535+00:00"} {"global_step": 1586, "acc_step": 0, "speed/wps": 12837.926811321604, "speed/FLOPS": 201637281731906.0, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21915294229984283, "optim/lr": 0.000951, "optim/total_tokens": 415760384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487272, "loss/out": 3.8770968914031982, "created_at": "2025-01-14T15:13:39.861325+00:00"} {"global_step": 1587, "acc_step": 0, "speed/wps": 12840.748522771884, "speed/FLOPS": 201681600587670.53, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21324169635772705, "optim/lr": 0.0009515999999999999, "optim/total_tokens": 416022528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465030, "loss/out": 3.8888087272644043, "created_at": "2025-01-14T15:13:44.972545+00:00"} {"global_step": 1588, "acc_step": 0, "speed/wps": 12839.562023951663, "speed/FLOPS": 201662964993278.72, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1917690485715866, "optim/lr": 0.0009522, "optim/total_tokens": 416284672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486185, "loss/out": 3.871400833129883, "created_at": "2025-01-14T15:13:50.082980+00:00"} {"global_step": 1589, "acc_step": 0, "speed/wps": 12844.31207130239, "speed/FLOPS": 201737571014170.97, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20265983045101166, "optim/lr": 0.0009528, "optim/total_tokens": 416546816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494625, "loss/out": 3.829176187515259, "created_at": "2025-01-14T15:13:55.189392+00:00"} {"global_step": 1590, "acc_step": 0, "speed/wps": 12843.904664839172, "speed/FLOPS": 201731172135830.97, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20528587698936462, "optim/lr": 0.0009534000000000001, "optim/total_tokens": 416808960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477242, "loss/out": 3.8944785594940186, "created_at": "2025-01-14T15:14:00.297696+00:00"} {"global_step": 1591, "acc_step": 0, "speed/wps": 12848.360380208433, "speed/FLOPS": 201801155268499.75, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20278474688529968, "optim/lr": 0.000954, "optim/total_tokens": 417071104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496480, "loss/out": 3.826988458633423, "created_at": "2025-01-14T15:14:05.403047+00:00"} {"global_step": 1592, "acc_step": 0, "speed/wps": 12842.82383369549, "speed/FLOPS": 201714196197501.78, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2272496074438095, "optim/lr": 0.0009546, "optim/total_tokens": 417333248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321163, "loss/out": 3.9589004516601562, "created_at": "2025-01-14T15:14:10.514107+00:00"} {"global_step": 1593, "acc_step": 0, "speed/wps": 12842.007151937954, "speed/FLOPS": 201701369088261.38, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20698973536491394, "optim/lr": 0.0009552, "optim/total_tokens": 417595392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476914, "loss/out": 3.805661201477051, "created_at": "2025-01-14T15:14:15.624859+00:00"} {"global_step": 1594, "acc_step": 0, "speed/wps": 12846.965499097241, "speed/FLOPS": 201779246743880.97, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.40264958143234253, "optim/lr": 0.0009558, "optim/total_tokens": 417857536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415159, "loss/out": 3.873859405517578, "created_at": "2025-01-14T15:14:20.731001+00:00"} {"global_step": 1595, "acc_step": 0, "speed/wps": 12843.540461402987, "speed/FLOPS": 201725451820397.7, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.35743868350982666, "optim/lr": 0.0009563999999999999, "optim/total_tokens": 418119680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495151, "loss/out": 3.903144598007202, "created_at": "2025-01-14T15:14:25.840134+00:00"} {"global_step": 1596, "acc_step": 0, "speed/wps": 12841.720773426523, "speed/FLOPS": 201696871120214.03, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2590872049331665, "optim/lr": 0.0009570000000000001, "optim/total_tokens": 418381824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500702, "loss/out": 3.980590581893921, "created_at": "2025-01-14T15:14:30.948618+00:00"} {"global_step": 1597, "acc_step": 0, "speed/wps": 12845.712079195951, "speed/FLOPS": 201759560061951.78, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.31077179312705994, "optim/lr": 0.0009576, "optim/total_tokens": 418643968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325437, "loss/out": 3.8346943855285645, "created_at": "2025-01-14T15:14:36.056175+00:00"} {"global_step": 1598, "acc_step": 0, "speed/wps": 12839.715607566357, "speed/FLOPS": 201665377234993.47, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3234524130821228, "optim/lr": 0.0009582000000000001, "optim/total_tokens": 418906112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486694, "loss/out": 3.862248182296753, "created_at": "2025-01-14T15:14:41.164414+00:00"} {"global_step": 1599, "acc_step": 0, "speed/wps": 12842.795697440873, "speed/FLOPS": 201713754278959.78, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3546642065048218, "optim/lr": 0.0009588, "optim/total_tokens": 419168256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 384156, "loss/out": 3.8338818550109863, "created_at": "2025-01-14T15:14:46.272192+00:00"} {"global_step": 1600, "acc_step": 0, "speed/wps": 12843.96625005975, "speed/FLOPS": 201732139416347.1, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31128260493278503, "optim/lr": 0.0009593999999999999, "optim/total_tokens": 419430400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505975, "loss/out": 4.030799388885498, "created_at": "2025-01-14T15:14:51.379584+00:00"} {"global_step": 1601, "acc_step": 0, "speed/wps": 12845.51788845488, "speed/FLOPS": 201756510029517.22, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.35302987694740295, "optim/lr": 0.00096, "optim/total_tokens": 419692544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500492, "loss/out": 3.932427406311035, "created_at": "2025-01-14T15:14:56.488381+00:00"} {"global_step": 1602, "acc_step": 0, "speed/wps": 12843.25202438816, "speed/FLOPS": 201720921520725.88, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.30836647748947144, "optim/lr": 0.0009605999999999999, "optim/total_tokens": 419954688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476387, "loss/out": 3.8627686500549316, "created_at": "2025-01-14T15:15:01.595462+00:00"} {"global_step": 1603, "acc_step": 0, "speed/wps": 12839.411220070016, "speed/FLOPS": 201660596411091.2, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2537052035331726, "optim/lr": 0.0009612, "optim/total_tokens": 420216832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493072, "loss/out": 4.001795768737793, "created_at": "2025-01-14T15:15:06.705826+00:00"} {"global_step": 1604, "acc_step": 0, "speed/wps": 12837.187133711934, "speed/FLOPS": 201625664078622.6, "speed/curr_iter_time": 1.2886, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.22892019152641296, "optim/lr": 0.0009618, "optim/total_tokens": 420478976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470304, "loss/out": 3.9400787353515625, "created_at": "2025-01-14T15:15:11.815203+00:00"} {"global_step": 1605, "acc_step": 0, "speed/wps": 12842.443631999964, "speed/FLOPS": 201708224607421.7, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23257531225681305, "optim/lr": 0.0009624, "optim/total_tokens": 420741120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474821, "loss/out": 3.9218504428863525, "created_at": "2025-01-14T15:15:16.929803+00:00"} {"global_step": 1606, "acc_step": 0, "speed/wps": 12840.47758018936, "speed/FLOPS": 201677345062097.84, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.23771502077579498, "optim/lr": 0.000963, "optim/total_tokens": 421003264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352511, "loss/out": 3.855745792388916, "created_at": "2025-01-14T15:15:22.038033+00:00"} {"global_step": 1607, "acc_step": 0, "speed/wps": 12842.588333484146, "speed/FLOPS": 201710497343071.44, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2610703706741333, "optim/lr": 0.0009636, "optim/total_tokens": 421265408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489648, "loss/out": 4.020522594451904, "created_at": "2025-01-14T15:15:27.147667+00:00"} {"global_step": 1608, "acc_step": 0, "speed/wps": 12841.892287774324, "speed/FLOPS": 201699564988731.66, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.001, "optim/grad_norm": 0.3237769901752472, "optim/lr": 0.0009642000000000001, "optim/total_tokens": 421527552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493533, "loss/out": 3.8447890281677246, "created_at": "2025-01-14T15:15:32.255543+00:00"} {"global_step": 1609, "acc_step": 0, "speed/wps": 12838.081086578939, "speed/FLOPS": 201639704836818.88, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.28648078441619873, "optim/lr": 0.0009648, "optim/total_tokens": 421789696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328886, "loss/out": 3.975512981414795, "created_at": "2025-01-14T15:15:37.369043+00:00"} {"global_step": 1610, "acc_step": 0, "speed/wps": 12844.657076497006, "speed/FLOPS": 201742989794839.2, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.27834731340408325, "optim/lr": 0.0009653999999999999, "optim/total_tokens": 422051840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474537, "loss/out": 4.009904861450195, "created_at": "2025-01-14T15:15:42.477564+00:00"} {"global_step": 1611, "acc_step": 0, "speed/wps": 12848.500900790568, "speed/FLOPS": 201803362337337.7, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25761497020721436, "optim/lr": 0.0009660000000000001, "optim/total_tokens": 422313984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366944, "loss/out": 3.9343056678771973, "created_at": "2025-01-14T15:15:47.586398+00:00"} {"global_step": 1612, "acc_step": 0, "speed/wps": 12842.944492372551, "speed/FLOPS": 201716091307827.44, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22592313587665558, "optim/lr": 0.0009666, "optim/total_tokens": 422576128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505166, "loss/out": 3.900183916091919, "created_at": "2025-01-14T15:15:52.694987+00:00"} {"global_step": 1613, "acc_step": 0, "speed/wps": 12838.886866104764, "speed/FLOPS": 201652360711529.06, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1951683908700943, "optim/lr": 0.0009672000000000001, "optim/total_tokens": 422838272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499090, "loss/out": 3.7853431701660156, "created_at": "2025-01-14T15:15:57.803623+00:00"} {"global_step": 1614, "acc_step": 0, "speed/wps": 12841.644816058815, "speed/FLOPS": 201695678105377.8, "speed/curr_iter_time": 1.2874, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24924784898757935, "optim/lr": 0.0009678, "optim/total_tokens": 423100416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326486, "loss/out": 3.891226291656494, "created_at": "2025-01-14T15:16:02.911718+00:00"} {"global_step": 1615, "acc_step": 0, "speed/wps": 12834.328581264414, "speed/FLOPS": 201580766584371.6, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2226339429616928, "optim/lr": 0.0009683999999999999, "optim/total_tokens": 423362560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470624, "loss/out": 3.8535420894622803, "created_at": "2025-01-14T15:16:08.024934+00:00"} {"global_step": 1616, "acc_step": 0, "speed/wps": 12845.455711514955, "speed/FLOPS": 201755533455235.6, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.25767579674720764, "optim/lr": 0.000969, "optim/total_tokens": 423624704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329179, "loss/out": 3.8326921463012695, "created_at": "2025-01-14T15:16:13.134214+00:00"} {"global_step": 1617, "acc_step": 0, "speed/wps": 12839.18480771443, "speed/FLOPS": 201657040293923.53, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.24651238322257996, "optim/lr": 0.0009695999999999999, "optim/total_tokens": 423886848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489311, "loss/out": 3.708500385284424, "created_at": "2025-01-14T15:16:18.244262+00:00"} {"global_step": 1618, "acc_step": 0, "speed/wps": 12847.239260443304, "speed/FLOPS": 201783546542008.4, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23341411352157593, "optim/lr": 0.0009702000000000001, "optim/total_tokens": 424148992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303859, "loss/out": 3.9103450775146484, "created_at": "2025-01-14T15:16:23.350503+00:00"} {"global_step": 1619, "acc_step": 0, "speed/wps": 12843.255910854108, "speed/FLOPS": 201720982563014.12, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21156994998455048, "optim/lr": 0.0009708000000000001, "optim/total_tokens": 424411136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500492, "loss/out": 3.7456140518188477, "created_at": "2025-01-14T15:16:28.457331+00:00"} {"global_step": 1620, "acc_step": 0, "speed/wps": 12845.277545884537, "speed/FLOPS": 201752735119184.66, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2681083381175995, "optim/lr": 0.0009714, "optim/total_tokens": 424673280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435636, "loss/out": 4.015433311462402, "created_at": "2025-01-14T15:16:33.566251+00:00"} {"global_step": 1621, "acc_step": 0, "speed/wps": 12838.063880941376, "speed/FLOPS": 201639434598637.9, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.28349974751472473, "optim/lr": 0.0009720000000000001, "optim/total_tokens": 424935424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479338, "loss/out": 3.899411678314209, "created_at": "2025-01-14T15:16:38.677423+00:00"} {"global_step": 1622, "acc_step": 0, "speed/wps": 12842.316000769173, "speed/FLOPS": 201706219983558.44, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21551258862018585, "optim/lr": 0.0009726, "optim/total_tokens": 425197568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488561, "loss/out": 3.88706636428833, "created_at": "2025-01-14T15:16:43.786661+00:00"} {"global_step": 1623, "acc_step": 0, "speed/wps": 12841.869822335295, "speed/FLOPS": 201699212138140.94, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19635435938835144, "optim/lr": 0.0009732000000000001, "optim/total_tokens": 425459712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297648, "loss/out": 3.8779523372650146, "created_at": "2025-01-14T15:16:48.894055+00:00"} {"global_step": 1624, "acc_step": 0, "speed/wps": 12839.838168788217, "speed/FLOPS": 201667302227399.56, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19675210118293762, "optim/lr": 0.0009738, "optim/total_tokens": 425721856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499373, "loss/out": 3.9241693019866943, "created_at": "2025-01-14T15:16:54.009696+00:00"} {"global_step": 1625, "acc_step": 0, "speed/wps": 12846.22883436546, "speed/FLOPS": 201767676411984.9, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19420763850212097, "optim/lr": 0.0009743999999999999, "optim/total_tokens": 425984000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468991, "loss/out": 3.8985681533813477, "created_at": "2025-01-14T15:16:59.117563+00:00"} {"global_step": 1626, "acc_step": 0, "speed/wps": 12843.344721226898, "speed/FLOPS": 201722377451957.03, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18748189508914948, "optim/lr": 0.0009750000000000001, "optim/total_tokens": 426246144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484563, "loss/out": 3.9277000427246094, "created_at": "2025-01-14T15:17:04.225974+00:00"} {"global_step": 1627, "acc_step": 0, "speed/wps": 12842.7535612628, "speed/FLOPS": 201713092472381.97, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19727440178394318, "optim/lr": 0.0009756, "optim/total_tokens": 426508288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503601, "loss/out": 3.8031420707702637, "created_at": "2025-01-14T15:17:09.334729+00:00"} {"global_step": 1628, "acc_step": 0, "speed/wps": 12836.417864154948, "speed/FLOPS": 201613581643142.97, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.20047767460346222, "optim/lr": 0.0009762000000000001, "optim/total_tokens": 426770432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300645, "loss/out": 4.031486511230469, "created_at": "2025-01-14T15:17:14.445495+00:00"} {"global_step": 1629, "acc_step": 0, "speed/wps": 12842.192281334263, "speed/FLOPS": 201704276799824.84, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17520321905612946, "optim/lr": 0.0009768, "optim/total_tokens": 427032576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475304, "loss/out": 3.740541934967041, "created_at": "2025-01-14T15:17:19.556831+00:00"} {"global_step": 1630, "acc_step": 0, "speed/wps": 12845.709935897303, "speed/FLOPS": 201759526398501.28, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.190492182970047, "optim/lr": 0.0009774, "optim/total_tokens": 427294720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327543, "loss/out": 3.9278810024261475, "created_at": "2025-01-14T15:17:24.664475+00:00"} {"global_step": 1631, "acc_step": 0, "speed/wps": 12841.02131624277, "speed/FLOPS": 201685885183988.28, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.170087069272995, "optim/lr": 0.0009780000000000001, "optim/total_tokens": 427556864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462890, "loss/out": 3.739804744720459, "created_at": "2025-01-14T15:17:29.772557+00:00"} {"global_step": 1632, "acc_step": 0, "speed/wps": 12843.40388194198, "speed/FLOPS": 201723306652282.75, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18456456065177917, "optim/lr": 0.0009786, "optim/total_tokens": 427819008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354908, "loss/out": 4.05372428894043, "created_at": "2025-01-14T15:17:34.879728+00:00"} {"global_step": 1633, "acc_step": 0, "speed/wps": 12845.263343401546, "speed/FLOPS": 201752512049673.03, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18517610430717468, "optim/lr": 0.0009792000000000002, "optim/total_tokens": 428081152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501464, "loss/out": 3.8806586265563965, "created_at": "2025-01-14T15:17:39.986193+00:00"} {"global_step": 1634, "acc_step": 0, "speed/wps": 12847.77600201653, "speed/FLOPS": 201791976805976.12, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21072638034820557, "optim/lr": 0.0009798, "optim/total_tokens": 428343296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481431, "loss/out": 3.7923216819763184, "created_at": "2025-01-14T15:17:45.093734+00:00"} {"global_step": 1635, "acc_step": 0, "speed/wps": 12840.921620939862, "speed/FLOPS": 201684319332260.4, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2873748242855072, "optim/lr": 0.0009804, "optim/total_tokens": 428605440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486411, "loss/out": 3.902249813079834, "created_at": "2025-01-14T15:17:50.202938+00:00"} {"global_step": 1636, "acc_step": 0, "speed/wps": 12843.175265995658, "speed/FLOPS": 201719715924694.53, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22035761177539825, "optim/lr": 0.000981, "optim/total_tokens": 428867584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491968, "loss/out": 3.9461746215820312, "created_at": "2025-01-14T15:17:55.317841+00:00"} {"global_step": 1637, "acc_step": 0, "speed/wps": 12830.47617178094, "speed/FLOPS": 201520259199668.25, "speed/curr_iter_time": 1.2883, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23688508570194244, "optim/lr": 0.0009816, "optim/total_tokens": 429129728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358652, "loss/out": 3.7409508228302, "created_at": "2025-01-14T15:18:00.431688+00:00"} {"global_step": 1638, "acc_step": 0, "speed/wps": 12844.917574099198, "speed/FLOPS": 201747081267642.84, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2117433398962021, "optim/lr": 0.0009822000000000001, "optim/total_tokens": 429391872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500418, "loss/out": 3.8265433311462402, "created_at": "2025-01-14T15:18:05.542828+00:00"} {"global_step": 1639, "acc_step": 0, "speed/wps": 12844.551775197133, "speed/FLOPS": 201741335893225.56, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19842678308486938, "optim/lr": 0.0009828, "optim/total_tokens": 429654016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495448, "loss/out": 3.9109749794006348, "created_at": "2025-01-14T15:18:10.649916+00:00"} {"global_step": 1640, "acc_step": 0, "speed/wps": 12837.07804636382, "speed/FLOPS": 201623950711918.75, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9780697226524353, "optim/lr": 0.0009834, "optim/total_tokens": 429916160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327018, "loss/out": 3.8633081912994385, "created_at": "2025-01-14T15:18:15.759475+00:00"} {"global_step": 1641, "acc_step": 0, "speed/wps": 12835.804574339461, "speed/FLOPS": 201603949083842.25, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20044063031673431, "optim/lr": 0.000984, "optim/total_tokens": 430178304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476387, "loss/out": 3.882071018218994, "created_at": "2025-01-14T15:18:20.871223+00:00"} {"global_step": 1642, "acc_step": 0, "speed/wps": 12844.163474431667, "speed/FLOPS": 201735237096122.75, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1979055106639862, "optim/lr": 0.0009846, "optim/total_tokens": 430440448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298243, "loss/out": 3.9012889862060547, "created_at": "2025-01-14T15:18:25.977660+00:00"} {"global_step": 1643, "acc_step": 0, "speed/wps": 12847.57895051125, "speed/FLOPS": 201788881841308.56, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21059101819992065, "optim/lr": 0.0009852, "optim/total_tokens": 430702592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498854, "loss/out": 3.8592915534973145, "created_at": "2025-01-14T15:18:31.082614+00:00"} {"global_step": 1644, "acc_step": 0, "speed/wps": 12845.909003087023, "speed/FLOPS": 201762653022262.84, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.18737107515335083, "optim/lr": 0.0009858, "optim/total_tokens": 430964736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317660, "loss/out": 3.7918801307678223, "created_at": "2025-01-14T15:18:36.189759+00:00"} {"global_step": 1645, "acc_step": 0, "speed/wps": 12841.792501610224, "speed/FLOPS": 201697997709903.9, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18664246797561646, "optim/lr": 0.0009864, "optim/total_tokens": 431226880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475631, "loss/out": 3.880770206451416, "created_at": "2025-01-14T15:18:41.299728+00:00"} {"global_step": 1646, "acc_step": 0, "speed/wps": 12845.378523466126, "speed/FLOPS": 201754321110861.25, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.20490305125713348, "optim/lr": 0.000987, "optim/total_tokens": 431489024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427085, "loss/out": 3.929309606552124, "created_at": "2025-01-14T15:18:46.408162+00:00"} {"global_step": 1647, "acc_step": 0, "speed/wps": 12844.120623618048, "speed/FLOPS": 201734564065211.66, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20078136026859283, "optim/lr": 0.0009876, "optim/total_tokens": 431751168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500182, "loss/out": 3.797274589538574, "created_at": "2025-01-14T15:18:51.519098+00:00"} {"global_step": 1648, "acc_step": 0, "speed/wps": 12845.410382944789, "speed/FLOPS": 201754821507753.16, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22110360860824585, "optim/lr": 0.0009882, "optim/total_tokens": 432013312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491741, "loss/out": 3.9046149253845215, "created_at": "2025-01-14T15:18:56.626729+00:00"} {"global_step": 1649, "acc_step": 0, "speed/wps": 12848.809107190855, "speed/FLOPS": 201808203142374.16, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2012689858675003, "optim/lr": 0.0009888, "optim/total_tokens": 432275456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310176, "loss/out": 3.9834418296813965, "created_at": "2025-01-14T15:19:01.760118+00:00"} {"global_step": 1650, "acc_step": 0, "speed/wps": 12849.413295446215, "speed/FLOPS": 201817692749166.25, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21344366669654846, "optim/lr": 0.0009893999999999999, "optim/total_tokens": 432537600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484846, "loss/out": 3.796096086502075, "created_at": "2025-01-14T15:19:06.864692+00:00"} {"global_step": 1651, "acc_step": 0, "speed/wps": 12846.471778231604, "speed/FLOPS": 201771492179242.56, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.46742939949035645, "optim/lr": 0.00099, "optim/total_tokens": 432799744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488549, "loss/out": 3.7991204261779785, "created_at": "2025-01-14T15:19:11.974790+00:00"} {"global_step": 1652, "acc_step": 0, "speed/wps": 12844.336383346532, "speed/FLOPS": 201737952868232.44, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6559444665908813, "optim/lr": 0.0009906, "optim/total_tokens": 433061888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295368, "loss/out": 4.0367913246154785, "created_at": "2025-01-14T15:19:17.085371+00:00"} {"global_step": 1653, "acc_step": 0, "speed/wps": 12851.019470806365, "speed/FLOPS": 201842919940313.66, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6765699982643127, "optim/lr": 0.0009912, "optim/total_tokens": 433324032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493072, "loss/out": 4.103709697723389, "created_at": "2025-01-14T15:19:22.191638+00:00"} {"global_step": 1654, "acc_step": 0, "speed/wps": 12845.36565966312, "speed/FLOPS": 201754119067158.34, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5384112596511841, "optim/lr": 0.0009918, "optim/total_tokens": 433586176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 373164, "loss/out": 4.044886589050293, "created_at": "2025-01-14T15:19:27.301454+00:00"} {"global_step": 1655, "acc_step": 0, "speed/wps": 12848.643671949041, "speed/FLOPS": 201805604754569.62, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.918710708618164, "optim/lr": 0.0009924, "optim/total_tokens": 433848320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502273, "loss/out": 4.029296398162842, "created_at": "2025-01-14T15:19:32.407202+00:00"} {"global_step": 1656, "acc_step": 0, "speed/wps": 12850.825553732095, "speed/FLOPS": 201839874206192.25, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.8481118679046631, "optim/lr": 0.0009930000000000002, "optim/total_tokens": 434110464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317214, "loss/out": 4.2126641273498535, "created_at": "2025-01-14T15:19:37.510357+00:00"} {"global_step": 1657, "acc_step": 0, "speed/wps": 12849.290473933495, "speed/FLOPS": 201815763668534.56, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9828270673751831, "optim/lr": 0.0009936, "optim/total_tokens": 434372608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462890, "loss/out": 4.036751747131348, "created_at": "2025-01-14T15:19:42.617377+00:00"} {"global_step": 1658, "acc_step": 0, "speed/wps": 12847.229092606149, "speed/FLOPS": 201783386842153.78, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4280809164047241, "optim/lr": 0.0009942, "optim/total_tokens": 434634752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370246, "loss/out": 4.2581658363342285, "created_at": "2025-01-14T15:19:47.723410+00:00"} {"global_step": 1659, "acc_step": 0, "speed/wps": 12848.08164952506, "speed/FLOPS": 201796777420101.28, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0046597719192505, "optim/lr": 0.0009948000000000001, "optim/total_tokens": 434896896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492777, "loss/out": 4.302987575531006, "created_at": "2025-01-14T15:19:52.827977+00:00"} {"global_step": 1660, "acc_step": 0, "speed/wps": 12841.163422099897, "speed/FLOPS": 201688117151748.38, "speed/curr_iter_time": 1.2874, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9187906980514526, "optim/lr": 0.0009954, "optim/total_tokens": 435159040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481715, "loss/out": 4.130674362182617, "created_at": "2025-01-14T15:19:57.937544+00:00"} {"global_step": 1661, "acc_step": 0, "speed/wps": 12843.931249327114, "speed/FLOPS": 201731589681745.0, "speed/curr_iter_time": 1.2874, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0032780170440674, "optim/lr": 0.0009960000000000001, "optim/total_tokens": 435421184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486752, "loss/out": 4.2357025146484375, "created_at": "2025-01-14T15:20:03.044931+00:00"} {"global_step": 1662, "acc_step": 0, "speed/wps": 12844.839487483518, "speed/FLOPS": 201745854810041.34, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3666949272155762, "optim/lr": 0.0009966, "optim/total_tokens": 435683328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479006, "loss/out": 4.228402137756348, "created_at": "2025-01-14T15:20:08.153248+00:00"} {"global_step": 1663, "acc_step": 0, "speed/wps": 12838.33477320111, "speed/FLOPS": 201643689333822.0, "speed/curr_iter_time": 1.2881, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7951761484146118, "optim/lr": 0.0009972, "optim/total_tokens": 435945472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308138, "loss/out": 4.213890075683594, "created_at": "2025-01-14T15:20:13.263373+00:00"} {"global_step": 1664, "acc_step": 0, "speed/wps": 12839.516308251783, "speed/FLOPS": 201662246965392.12, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5499114990234375, "optim/lr": 0.0009978, "optim/total_tokens": 436207616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502030, "loss/out": 4.1639404296875, "created_at": "2025-01-14T15:20:18.373182+00:00"} {"global_step": 1665, "acc_step": 0, "speed/wps": 12841.939131256368, "speed/FLOPS": 201700300729986.84, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7096060514450073, "optim/lr": 0.0009984, "optim/total_tokens": 436469760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484611, "loss/out": 4.240663528442383, "created_at": "2025-01-14T15:20:23.485190+00:00"} {"global_step": 1666, "acc_step": 0, "speed/wps": 12837.427051841214, "speed/FLOPS": 201629432322525.0, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.39958178997039795, "optim/lr": 0.000999, "optim/total_tokens": 436731904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469257, "loss/out": 4.143128871917725, "created_at": "2025-01-14T15:20:28.599425+00:00"} {"global_step": 1667, "acc_step": 0, "speed/wps": 12844.003403089015, "speed/FLOPS": 201732722955725.3, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4007452428340912, "optim/lr": 0.0009996, "optim/total_tokens": 436994048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476952, "loss/out": 4.081543445587158, "created_at": "2025-01-14T15:20:33.710351+00:00"} {"global_step": 1668, "acc_step": 0, "speed/wps": 12839.93127321087, "speed/FLOPS": 201668764560296.22, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.44543394446372986, "optim/lr": 0.0010002, "optim/total_tokens": 437256192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322598, "loss/out": 4.056058883666992, "created_at": "2025-01-14T15:20:38.818325+00:00"} {"global_step": 1669, "acc_step": 0, "speed/wps": 12840.49540383636, "speed/FLOPS": 201677625006966.22, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.32359424233436584, "optim/lr": 0.0010008, "optim/total_tokens": 437518336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473815, "loss/out": 3.9848508834838867, "created_at": "2025-01-14T15:20:43.935357+00:00"} {"global_step": 1670, "acc_step": 0, "speed/wps": 12841.85651325196, "speed/FLOPS": 201699003100698.56, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.28745660185813904, "optim/lr": 0.0010014, "optim/total_tokens": 437780480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355439, "loss/out": 4.016024112701416, "created_at": "2025-01-14T15:20:49.045298+00:00"} {"global_step": 1671, "acc_step": 0, "speed/wps": 12842.037281505418, "speed/FLOPS": 201701842314520.75, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24800339341163635, "optim/lr": 0.001002, "optim/total_tokens": 438042624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502030, "loss/out": 4.024054527282715, "created_at": "2025-01-14T15:20:54.158473+00:00"} {"global_step": 1672, "acc_step": 0, "speed/wps": 12846.972260957418, "speed/FLOPS": 201779352948186.8, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.33592718839645386, "optim/lr": 0.0010026, "optim/total_tokens": 438304768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471924, "loss/out": 4.0062713623046875, "created_at": "2025-01-14T15:20:59.263677+00:00"} {"global_step": 1673, "acc_step": 0, "speed/wps": 12840.671649027176, "speed/FLOPS": 201680393179874.28, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5557680726051331, "optim/lr": 0.0010031999999999999, "optim/total_tokens": 438566912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486988, "loss/out": 3.9363045692443848, "created_at": "2025-01-14T15:21:04.373029+00:00"} {"global_step": 1674, "acc_step": 0, "speed/wps": 12833.80240940066, "speed/FLOPS": 201572502332215.44, "speed/curr_iter_time": 1.2881, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.4659840762615204, "optim/lr": 0.0010038, "optim/total_tokens": 438829056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493533, "loss/out": 4.07765007019043, "created_at": "2025-01-14T15:21:09.484344+00:00"} {"global_step": 1675, "acc_step": 0, "speed/wps": 12824.432308566289, "speed/FLOPS": 201425332022743.78, "speed/curr_iter_time": 1.2883, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.45635873079299927, "optim/lr": 0.0010044, "optim/total_tokens": 439091200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359462, "loss/out": 3.911040782928467, "created_at": "2025-01-14T15:21:14.599919+00:00"} {"global_step": 1676, "acc_step": 0, "speed/wps": 12832.545268567128, "speed/FLOPS": 201552757207932.28, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3468971848487854, "optim/lr": 0.001005, "optim/total_tokens": 439353344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494106, "loss/out": 4.073237419128418, "created_at": "2025-01-14T15:21:19.715226+00:00"} {"global_step": 1677, "acc_step": 0, "speed/wps": 12833.257220260883, "speed/FLOPS": 201563939387606.84, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2931746542453766, "optim/lr": 0.0010056, "optim/total_tokens": 439615488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501747, "loss/out": 4.032676696777344, "created_at": "2025-01-14T15:21:24.826659+00:00"} {"global_step": 1678, "acc_step": 0, "speed/wps": 12831.89783240122, "speed/FLOPS": 201542588333276.2, "speed/curr_iter_time": 1.2875, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2577109634876251, "optim/lr": 0.0010061999999999998, "optim/total_tokens": 439877632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298243, "loss/out": 4.01204776763916, "created_at": "2025-01-14T15:21:29.939917+00:00"} {"global_step": 1679, "acc_step": 0, "speed/wps": 12838.062773092892, "speed/FLOPS": 201639417198355.12, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22992423176765442, "optim/lr": 0.0010068, "optim/total_tokens": 440139776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490695, "loss/out": 3.9957685470581055, "created_at": "2025-01-14T15:21:35.053323+00:00"} {"global_step": 1680, "acc_step": 0, "speed/wps": 12837.996229927148, "speed/FLOPS": 201638372046497.25, "speed/curr_iter_time": 1.2877, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.26208409667015076, "optim/lr": 0.0010073999999999999, "optim/total_tokens": 440401920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485892, "loss/out": 3.853365421295166, "created_at": "2025-01-14T15:21:40.163545+00:00"} {"global_step": 1681, "acc_step": 0, "speed/wps": 12835.957061270814, "speed/FLOPS": 201606344100638.12, "speed/curr_iter_time": 1.2879, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2211582064628601, "optim/lr": 0.001008, "optim/total_tokens": 440664064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495387, "loss/out": 3.9589884281158447, "created_at": "2025-01-14T15:21:45.273524+00:00"} {"global_step": 1682, "acc_step": 0, "speed/wps": 12849.736689499252, "speed/FLOPS": 201822772097159.22, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1988964080810547, "optim/lr": 0.0010086000000000001, "optim/total_tokens": 440926208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436666, "loss/out": 3.9799835681915283, "created_at": "2025-01-14T15:21:50.378625+00:00"} {"global_step": 1683, "acc_step": 0, "speed/wps": 12843.85261441836, "speed/FLOPS": 201730354612447.3, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2226259410381317, "optim/lr": 0.0010092, "optim/total_tokens": 441188352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485422, "loss/out": 3.9072203636169434, "created_at": "2025-01-14T15:21:55.487115+00:00"} {"global_step": 1684, "acc_step": 0, "speed/wps": 12840.788682218328, "speed/FLOPS": 201682231346961.78, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19572433829307556, "optim/lr": 0.0010098000000000002, "optim/total_tokens": 441450496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474011, "loss/out": 3.958486557006836, "created_at": "2025-01-14T15:22:00.594405+00:00"} {"global_step": 1685, "acc_step": 0, "speed/wps": 12841.856717141565, "speed/FLOPS": 201699006303065.0, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20107798278331757, "optim/lr": 0.0010104, "optim/total_tokens": 441712640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488785, "loss/out": 4.000936508178711, "created_at": "2025-01-14T15:22:05.704843+00:00"} {"global_step": 1686, "acc_step": 0, "speed/wps": 12841.26344922538, "speed/FLOPS": 201689688215205.78, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22163955867290497, "optim/lr": 0.0010110000000000002, "optim/total_tokens": 441974784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494342, "loss/out": 3.834402084350586, "created_at": "2025-01-14T15:22:10.814324+00:00"} {"global_step": 1687, "acc_step": 0, "speed/wps": 12842.257787557857, "speed/FLOPS": 201705305665081.38, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19844256341457367, "optim/lr": 0.0010116, "optim/total_tokens": 442236928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295309, "loss/out": 3.9660043716430664, "created_at": "2025-01-14T15:22:15.921988+00:00"} {"global_step": 1688, "acc_step": 0, "speed/wps": 12846.860715801271, "speed/FLOPS": 201777600978230.4, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19314776360988617, "optim/lr": 0.0010122, "optim/total_tokens": 442499072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478008, "loss/out": 3.8678271770477295, "created_at": "2025-01-14T15:22:21.029473+00:00"} {"global_step": 1689, "acc_step": 0, "speed/wps": 12838.744620893329, "speed/FLOPS": 201650126555018.06, "speed/curr_iter_time": 1.2879, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.19537070393562317, "optim/lr": 0.0010128000000000001, "optim/total_tokens": 442761216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429739, "loss/out": 3.8693456649780273, "created_at": "2025-01-14T15:22:26.140206+00:00"} {"global_step": 1690, "acc_step": 0, "speed/wps": 12841.275119080085, "speed/FLOPS": 201689871506309.97, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1901617795228958, "optim/lr": 0.0010134, "optim/total_tokens": 443023360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478244, "loss/out": 3.8683927059173584, "created_at": "2025-01-14T15:22:31.249207+00:00"} {"global_step": 1691, "acc_step": 0, "speed/wps": 12850.365276401188, "speed/FLOPS": 201832644918223.03, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19272717833518982, "optim/lr": 0.0010140000000000001, "optim/total_tokens": 443285504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487456, "loss/out": 3.9260170459747314, "created_at": "2025-01-14T15:22:36.353034+00:00"} {"global_step": 1692, "acc_step": 0, "speed/wps": 12837.289190208772, "speed/FLOPS": 201627267016145.75, "speed/curr_iter_time": 1.2875, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22311843931674957, "optim/lr": 0.0010146, "optim/total_tokens": 443547648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312861, "loss/out": 3.938831329345703, "created_at": "2025-01-14T15:22:41.467140+00:00"} {"global_step": 1693, "acc_step": 0, "speed/wps": 12844.160951856975, "speed/FLOPS": 201735197475620.8, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18758095800876617, "optim/lr": 0.0010152, "optim/total_tokens": 443809792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464740, "loss/out": 3.8490476608276367, "created_at": "2025-01-14T15:22:46.573569+00:00"} {"global_step": 1694, "acc_step": 0, "speed/wps": 12843.720090883062, "speed/FLOPS": 201728273148219.06, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20345760881900787, "optim/lr": 0.0010158, "optim/total_tokens": 444071936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332921, "loss/out": 3.9693620204925537, "created_at": "2025-01-14T15:22:51.681685+00:00"} {"global_step": 1695, "acc_step": 0, "speed/wps": 12843.942520808134, "speed/FLOPS": 201731766715843.06, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.25324657559394836, "optim/lr": 0.0010164, "optim/total_tokens": 444334080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467137, "loss/out": 3.976841926574707, "created_at": "2025-01-14T15:22:56.795179+00:00"} {"global_step": 1696, "acc_step": 0, "speed/wps": 12827.164557721595, "speed/FLOPS": 201468245750231.66, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2093808650970459, "optim/lr": 0.0010170000000000001, "optim/total_tokens": 444596224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 403092, "loss/out": 3.86043119430542, "created_at": "2025-01-14T15:23:01.911032+00:00"} {"global_step": 1697, "acc_step": 0, "speed/wps": 12835.105787223609, "speed/FLOPS": 201592973671953.66, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22003965079784393, "optim/lr": 0.0010176, "optim/total_tokens": 444858368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502556, "loss/out": 3.843890428543091, "created_at": "2025-01-14T15:23:07.024677+00:00"} {"global_step": 1698, "acc_step": 0, "speed/wps": 12841.371925794969, "speed/FLOPS": 201691391988793.6, "speed/curr_iter_time": 1.2878, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2205013483762741, "optim/lr": 0.0010182, "optim/total_tokens": 445120512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487515, "loss/out": 3.832341432571411, "created_at": "2025-01-14T15:23:12.134362+00:00"} {"global_step": 1699, "acc_step": 0, "speed/wps": 12844.156490465684, "speed/FLOPS": 201735127403340.0, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19544118642807007, "optim/lr": 0.0010188, "optim/total_tokens": 445382656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332921, "loss/out": 3.897110939025879, "created_at": "2025-01-14T15:23:17.244524+00:00"} {"global_step": 1700, "acc_step": 0, "speed/wps": 12842.836270503105, "speed/FLOPS": 201714391534655.34, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2198788970708847, "optim/lr": 0.0010194, "optim/total_tokens": 445644800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465550, "loss/out": 3.905930757522583, "created_at": "2025-01-14T15:23:22.352859+00:00"} {"global_step": 1701, "acc_step": 0, "speed/wps": 12840.783920900858, "speed/FLOPS": 201682156563928.2, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20875027775764465, "optim/lr": 0.00102, "optim/total_tokens": 445906944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462154, "loss/out": 3.7215216159820557, "created_at": "2025-01-14T15:23:27.464647+00:00"} {"global_step": 1702, "acc_step": 0, "speed/wps": 12844.427131150049, "speed/FLOPS": 201739378187187.22, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18610839545726776, "optim/lr": 0.0010206, "optim/total_tokens": 446169088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 3.8798794746398926, "created_at": "2025-01-14T15:23:32.571023+00:00"} {"global_step": 1703, "acc_step": 0, "speed/wps": 12842.03314500929, "speed/FLOPS": 201701777345164.94, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19338025152683258, "optim/lr": 0.0010211999999999999, "optim/total_tokens": 446431232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500182, "loss/out": 3.956063985824585, "created_at": "2025-01-14T15:23:37.679980+00:00"} {"global_step": 1704, "acc_step": 0, "speed/wps": 12836.61503005788, "speed/FLOPS": 201616678404582.84, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18667244911193848, "optim/lr": 0.0010218, "optim/total_tokens": 446693376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307776, "loss/out": 3.9284515380859375, "created_at": "2025-01-14T15:23:42.791114+00:00"} {"global_step": 1705, "acc_step": 0, "speed/wps": 12844.609333973058, "speed/FLOPS": 201742239932895.5, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.001, "optim/grad_norm": 0.20278455317020416, "optim/lr": 0.0010224, "optim/total_tokens": 446955520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470304, "loss/out": 3.940256118774414, "created_at": "2025-01-14T15:23:47.901921+00:00"} {"global_step": 1706, "acc_step": 0, "speed/wps": 12841.331640845181, "speed/FLOPS": 201690759258299.84, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2598437964916229, "optim/lr": 0.001023, "optim/total_tokens": 447217664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469009, "loss/out": 3.763430595397949, "created_at": "2025-01-14T15:23:53.014265+00:00"} {"global_step": 1707, "acc_step": 0, "speed/wps": 12841.315103725019, "speed/FLOPS": 201690499520103.3, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20434071123600006, "optim/lr": 0.0010236, "optim/total_tokens": 447479808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478764, "loss/out": 3.809180736541748, "created_at": "2025-01-14T15:23:58.127600+00:00"} {"global_step": 1708, "acc_step": 0, "speed/wps": 12846.17661501672, "speed/FLOPS": 201766856235356.06, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.22325587272644043, "optim/lr": 0.0010242, "optim/total_tokens": 447741952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501228, "loss/out": 3.8659164905548096, "created_at": "2025-01-14T15:24:03.235601+00:00"} {"global_step": 1709, "acc_step": 0, "speed/wps": 12835.456893240316, "speed/FLOPS": 201598488274416.28, "speed/curr_iter_time": 1.2881, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2770741879940033, "optim/lr": 0.0010248, "optim/total_tokens": 448004096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476151, "loss/out": 3.7293007373809814, "created_at": "2025-01-14T15:24:08.348239+00:00"} {"global_step": 1710, "acc_step": 0, "speed/wps": 12841.707518552728, "speed/FLOPS": 201696662934206.97, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.26176342368125916, "optim/lr": 0.0010254, "optim/total_tokens": 448266240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503885, "loss/out": 3.882485866546631, "created_at": "2025-01-14T15:24:13.458090+00:00"} {"global_step": 1711, "acc_step": 0, "speed/wps": 12840.484869438778, "speed/FLOPS": 201677459549776.44, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.001, "optim/grad_norm": 0.22103512287139893, "optim/lr": 0.001026, "optim/total_tokens": 448528384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 284116, "loss/out": 3.842132568359375, "created_at": "2025-01-14T15:24:18.566194+00:00"} {"global_step": 1712, "acc_step": 0, "speed/wps": 12850.41561022162, "speed/FLOPS": 201833435480038.84, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25337785482406616, "optim/lr": 0.0010266000000000001, "optim/total_tokens": 448790528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490695, "loss/out": 3.778503179550171, "created_at": "2025-01-14T15:24:23.673141+00:00"} {"global_step": 1713, "acc_step": 0, "speed/wps": 12839.934675038696, "speed/FLOPS": 201668817990676.8, "speed/curr_iter_time": 1.2879, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21529853343963623, "optim/lr": 0.0010272, "optim/total_tokens": 449052672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 412976, "loss/out": 3.9406309127807617, "created_at": "2025-01-14T15:24:28.780853+00:00"} {"global_step": 1714, "acc_step": 0, "speed/wps": 12845.505734874605, "speed/FLOPS": 201756319140838.22, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21185964345932007, "optim/lr": 0.0010278000000000002, "optim/total_tokens": 449314816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488785, "loss/out": 3.93672513961792, "created_at": "2025-01-14T15:24:33.887158+00:00"} {"global_step": 1715, "acc_step": 0, "speed/wps": 12844.001436186129, "speed/FLOPS": 201732692062812.7, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1904071569442749, "optim/lr": 0.0010284, "optim/total_tokens": 449576960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480904, "loss/out": 3.8488707542419434, "created_at": "2025-01-14T15:24:38.995038+00:00"} {"global_step": 1716, "acc_step": 0, "speed/wps": 12843.425649509216, "speed/FLOPS": 201723648541837.97, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.001, "optim/grad_norm": 0.21184390783309937, "optim/lr": 0.0010290000000000002, "optim/total_tokens": 449839104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479054, "loss/out": 3.732684850692749, "created_at": "2025-01-14T15:24:44.104129+00:00"} {"global_step": 1717, "acc_step": 0, "speed/wps": 12843.968702228076, "speed/FLOPS": 201732177931020.28, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21517615020275116, "optim/lr": 0.0010296, "optim/total_tokens": 450101248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487515, "loss/out": 3.8531289100646973, "created_at": "2025-01-14T15:24:49.215838+00:00"} {"global_step": 1718, "acc_step": 0, "speed/wps": 12842.035439416728, "speed/FLOPS": 201701813381986.56, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2625441551208496, "optim/lr": 0.0010302, "optim/total_tokens": 450363392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356361, "loss/out": 3.795579195022583, "created_at": "2025-01-14T15:24:54.323154+00:00"} {"global_step": 1719, "acc_step": 0, "speed/wps": 12841.779492659058, "speed/FLOPS": 201697793386449.66, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.23067253828048706, "optim/lr": 0.0010308000000000001, "optim/total_tokens": 450625536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496999, "loss/out": 3.790465831756592, "created_at": "2025-01-14T15:24:59.430050+00:00"} {"global_step": 1720, "acc_step": 0, "speed/wps": 12839.061303209566, "speed/FLOPS": 201655100485961.7, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23888756334781647, "optim/lr": 0.0010314, "optim/total_tokens": 450887680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 3.8380255699157715, "created_at": "2025-01-14T15:25:04.539365+00:00"} {"global_step": 1721, "acc_step": 0, "speed/wps": 12846.009709187774, "speed/FLOPS": 201764234749959.8, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2503044009208679, "optim/lr": 0.001032, "optim/total_tokens": 451149824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502273, "loss/out": 3.844245672225952, "created_at": "2025-01-14T15:25:09.648680+00:00"} {"global_step": 1722, "acc_step": 0, "speed/wps": 12843.564421243032, "speed/FLOPS": 201725828142609.88, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.23585525155067444, "optim/lr": 0.0010326, "optim/total_tokens": 451411968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359462, "loss/out": 3.802194595336914, "created_at": "2025-01-14T15:25:14.758327+00:00"} {"global_step": 1723, "acc_step": 0, "speed/wps": 12843.797612986365, "speed/FLOPS": 201729490739377.97, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23246413469314575, "optim/lr": 0.0010332, "optim/total_tokens": 451674112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494625, "loss/out": 3.7521181106567383, "created_at": "2025-01-14T15:25:19.864651+00:00"} {"global_step": 1724, "acc_step": 0, "speed/wps": 12849.457213968919, "speed/FLOPS": 201818382549915.72, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21453997492790222, "optim/lr": 0.0010338, "optim/total_tokens": 451936256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476431, "loss/out": 3.9465508460998535, "created_at": "2025-01-14T15:25:24.973112+00:00"} {"global_step": 1725, "acc_step": 0, "speed/wps": 12841.260119336996, "speed/FLOPS": 201689635914732.94, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.216119185090065, "optim/lr": 0.0010344, "optim/total_tokens": 452198400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469784, "loss/out": 4.005363464355469, "created_at": "2025-01-14T15:25:30.080654+00:00"} {"global_step": 1726, "acc_step": 0, "speed/wps": 12847.688718380154, "speed/FLOPS": 201790605896526.53, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2092050015926361, "optim/lr": 0.001035, "optim/total_tokens": 452460544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501228, "loss/out": 3.88922119140625, "created_at": "2025-01-14T15:25:35.187621+00:00"} {"global_step": 1727, "acc_step": 0, "speed/wps": 12845.693136587817, "speed/FLOPS": 201759262542261.16, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.20802801847457886, "optim/lr": 0.0010356, "optim/total_tokens": 452722688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 283870, "loss/out": 3.8228535652160645, "created_at": "2025-01-14T15:25:40.296694+00:00"} {"global_step": 1728, "acc_step": 0, "speed/wps": 12840.756972191703, "speed/FLOPS": 201681733297419.47, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18382878601551056, "optim/lr": 0.0010362, "optim/total_tokens": 452984832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488266, "loss/out": 3.909886121749878, "created_at": "2025-01-14T15:25:45.415417+00:00"} {"global_step": 1729, "acc_step": 0, "speed/wps": 12849.713394127928, "speed/FLOPS": 201822406211341.38, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21067684888839722, "optim/lr": 0.0010368, "optim/total_tokens": 453246976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493362, "loss/out": 3.9277570247650146, "created_at": "2025-01-14T15:25:50.549405+00:00"} {"global_step": 1730, "acc_step": 0, "speed/wps": 12850.726888920857, "speed/FLOPS": 201838324539752.12, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1892915815114975, "optim/lr": 0.0010374, "optim/total_tokens": 453509120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362250, "loss/out": 3.9413747787475586, "created_at": "2025-01-14T15:25:55.655572+00:00"} {"global_step": 1731, "acc_step": 0, "speed/wps": 12842.629106806466, "speed/FLOPS": 201711137744126.88, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2219778597354889, "optim/lr": 0.0010379999999999999, "optim/total_tokens": 453771264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504930, "loss/out": 3.9293808937072754, "created_at": "2025-01-14T15:26:00.765393+00:00"} {"global_step": 1732, "acc_step": 0, "speed/wps": 12839.398466607741, "speed/FLOPS": 201660396100441.06, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18235145509243011, "optim/lr": 0.0010386, "optim/total_tokens": 454033408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494342, "loss/out": 3.8285441398620605, "created_at": "2025-01-14T15:26:05.874464+00:00"} {"global_step": 1733, "acc_step": 0, "speed/wps": 12850.036296433334, "speed/FLOPS": 201827477835762.03, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23633989691734314, "optim/lr": 0.0010391999999999999, "optim/total_tokens": 454295552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474258, "loss/out": 3.8269784450531006, "created_at": "2025-01-14T15:26:10.982452+00:00"} {"global_step": 1734, "acc_step": 0, "speed/wps": 12847.025600861529, "speed/FLOPS": 201780190724677.6, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20990827679634094, "optim/lr": 0.0010398, "optim/total_tokens": 454557696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489365, "loss/out": 3.841160774230957, "created_at": "2025-01-14T15:26:16.091153+00:00"} {"global_step": 1735, "acc_step": 0, "speed/wps": 12844.007388472159, "speed/FLOPS": 201732785551644.06, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.18087902665138245, "optim/lr": 0.0010404, "optim/total_tokens": 454819840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469227, "loss/out": 3.907616138458252, "created_at": "2025-01-14T15:26:21.199590+00:00"} {"global_step": 1736, "acc_step": 0, "speed/wps": 12842.25677976474, "speed/FLOPS": 201705289836305.62, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22069156169891357, "optim/lr": 0.001041, "optim/total_tokens": 455081984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493816, "loss/out": 3.734502077102661, "created_at": "2025-01-14T15:26:26.309777+00:00"} {"global_step": 1737, "acc_step": 0, "speed/wps": 12844.656370778186, "speed/FLOPS": 201742978710555.38, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2359342873096466, "optim/lr": 0.0010416, "optim/total_tokens": 455344128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317135, "loss/out": 3.819876194000244, "created_at": "2025-01-14T15:26:31.418038+00:00"} {"global_step": 1738, "acc_step": 0, "speed/wps": 12847.22568925756, "speed/FLOPS": 201783333387887.53, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.26066160202026367, "optim/lr": 0.0010422, "optim/total_tokens": 455606272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470839, "loss/out": 3.9136128425598145, "created_at": "2025-01-14T15:26:36.526297+00:00"} {"global_step": 1739, "acc_step": 0, "speed/wps": 12846.0091792009, "speed/FLOPS": 201764226425787.7, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.214981809258461, "optim/lr": 0.0010428, "optim/total_tokens": 455868416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368996, "loss/out": 3.928330898284912, "created_at": "2025-01-14T15:26:41.633107+00:00"} {"global_step": 1740, "acc_step": 0, "speed/wps": 12845.906756541026, "speed/FLOPS": 201762617737170.88, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21667586266994476, "optim/lr": 0.0010434, "optim/total_tokens": 456130560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487740, "loss/out": 3.782519817352295, "created_at": "2025-01-14T15:26:46.740682+00:00"} {"global_step": 1741, "acc_step": 0, "speed/wps": 12835.629221305615, "speed/FLOPS": 201601194923485.38, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23902642726898193, "optim/lr": 0.001044, "optim/total_tokens": 456392704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498044, "loss/out": 3.8466341495513916, "created_at": "2025-01-14T15:26:51.851165+00:00"} {"global_step": 1742, "acc_step": 0, "speed/wps": 12845.100072205902, "speed/FLOPS": 201749947651187.3, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22129599750041962, "optim/lr": 0.0010446000000000001, "optim/total_tokens": 456654848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327304, "loss/out": 3.6805102825164795, "created_at": "2025-01-14T15:26:56.963550+00:00"} {"global_step": 1743, "acc_step": 0, "speed/wps": 12844.128997767048, "speed/FLOPS": 201734695592729.0, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23356091976165771, "optim/lr": 0.0010452, "optim/total_tokens": 456916992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462890, "loss/out": 3.815162420272827, "created_at": "2025-01-14T15:27:02.069700+00:00"} {"global_step": 1744, "acc_step": 0, "speed/wps": 12849.783665313553, "speed/FLOPS": 201823509916874.34, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20941582322120667, "optim/lr": 0.0010458000000000002, "optim/total_tokens": 457179136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361049, "loss/out": 3.8561389446258545, "created_at": "2025-01-14T15:27:07.177776+00:00"} {"global_step": 1745, "acc_step": 0, "speed/wps": 12837.871054946208, "speed/FLOPS": 201636406001413.12, "speed/curr_iter_time": 1.288, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.20662595331668854, "optim/lr": 0.0010464, "optim/total_tokens": 457441280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500702, "loss/out": 3.7669782638549805, "created_at": "2025-01-14T15:27:12.286467+00:00"} {"global_step": 1746, "acc_step": 0, "speed/wps": 12841.05611365626, "speed/FLOPS": 201686431725183.78, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.001, "optim/grad_norm": 0.18624529242515564, "optim/lr": 0.001047, "optim/total_tokens": 457703424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494692, "loss/out": 3.880791664123535, "created_at": "2025-01-14T15:27:17.395178+00:00"} {"global_step": 1747, "acc_step": 0, "speed/wps": 12845.233932614143, "speed/FLOPS": 201752050112842.22, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21654583513736725, "optim/lr": 0.0010476, "optim/total_tokens": 457965568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467117, "loss/out": 3.846543312072754, "created_at": "2025-01-14T15:27:22.500772+00:00"} {"global_step": 1748, "acc_step": 0, "speed/wps": 12843.639721500344, "speed/FLOPS": 201727010836624.34, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19064295291900635, "optim/lr": 0.0010482, "optim/total_tokens": 458227712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499682, "loss/out": 3.718435287475586, "created_at": "2025-01-14T15:27:27.607559+00:00"} {"global_step": 1749, "acc_step": 0, "speed/wps": 12839.867512701225, "speed/FLOPS": 201667763113876.1, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2106536328792572, "optim/lr": 0.0010488000000000001, "optim/total_tokens": 458489856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489891, "loss/out": 3.851940870285034, "created_at": "2025-01-14T15:27:32.715267+00:00"} {"global_step": 1750, "acc_step": 0, "speed/wps": 12839.809167059793, "speed/FLOPS": 201666846715402.12, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23530058562755585, "optim/lr": 0.0010494, "optim/total_tokens": 458752000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496999, "loss/out": 3.7837347984313965, "created_at": "2025-01-14T15:27:37.822940+00:00"} {"global_step": 1751, "acc_step": 0, "speed/wps": 12841.927054754642, "speed/FLOPS": 201700111051932.38, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18559181690216064, "optim/lr": 0.00105, "optim/total_tokens": 459014144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468211, "loss/out": 3.8759267330169678, "created_at": "2025-01-14T15:27:42.930314+00:00"} {"global_step": 1752, "acc_step": 0, "speed/wps": 12848.309630838661, "speed/FLOPS": 201800358179910.53, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19437485933303833, "optim/lr": 0.0010506, "optim/total_tokens": 459276288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490695, "loss/out": 3.7806479930877686, "created_at": "2025-01-14T15:27:48.036779+00:00"} {"global_step": 1753, "acc_step": 0, "speed/wps": 12843.084718712986, "speed/FLOPS": 201718293755195.16, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2138480246067047, "optim/lr": 0.0010512, "optim/total_tokens": 459538432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316078, "loss/out": 3.7626516819000244, "created_at": "2025-01-14T15:27:53.145855+00:00"} {"global_step": 1754, "acc_step": 0, "speed/wps": 12836.239501915074, "speed/FLOPS": 201610780219070.34, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.20099236071109772, "optim/lr": 0.0010518, "optim/total_tokens": 459800576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489830, "loss/out": 3.9097213745117188, "created_at": "2025-01-14T15:27:58.255166+00:00"} {"global_step": 1755, "acc_step": 0, "speed/wps": 12845.866696396053, "speed/FLOPS": 201761988537545.66, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19450248777866364, "optim/lr": 0.0010524, "optim/total_tokens": 460062720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471634, "loss/out": 3.8540892601013184, "created_at": "2025-01-14T15:28:03.360822+00:00"} {"global_step": 1756, "acc_step": 0, "speed/wps": 12841.71394678814, "speed/FLOPS": 201696763898477.44, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2353348284959793, "optim/lr": 0.001053, "optim/total_tokens": 460324864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480621, "loss/out": 3.8631362915039062, "created_at": "2025-01-14T15:28:08.471348+00:00"} {"global_step": 1757, "acc_step": 0, "speed/wps": 12844.714887089098, "speed/FLOPS": 201743897789628.1, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1845128834247589, "optim/lr": 0.0010536, "optim/total_tokens": 460587008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495954, "loss/out": 3.887662410736084, "created_at": "2025-01-14T15:28:13.577763+00:00"} {"global_step": 1758, "acc_step": 0, "speed/wps": 12838.504627748307, "speed/FLOPS": 201646357132890.34, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22423473000526428, "optim/lr": 0.0010542, "optim/total_tokens": 460849152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293678, "loss/out": 3.8464043140411377, "created_at": "2025-01-14T15:28:18.685853+00:00"} {"global_step": 1759, "acc_step": 0, "speed/wps": 12842.582105942129, "speed/FLOPS": 201710399530966.53, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.001, "optim/grad_norm": 0.19121024012565613, "optim/lr": 0.0010548, "optim/total_tokens": 461111296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478811, "loss/out": 3.699212074279785, "created_at": "2025-01-14T15:28:23.796456+00:00"} {"global_step": 1760, "acc_step": 0, "speed/wps": 12843.391629220403, "speed/FLOPS": 201723114206453.38, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.20243249833583832, "optim/lr": 0.0010554, "optim/total_tokens": 461373440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356884, "loss/out": 3.8138809204101562, "created_at": "2025-01-14T15:28:28.904303+00:00"} {"global_step": 1761, "acc_step": 0, "speed/wps": 12845.4872196439, "speed/FLOPS": 201756028333697.38, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2277844101190567, "optim/lr": 0.0010559999999999999, "optim/total_tokens": 461635584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479574, "loss/out": 3.8428139686584473, "created_at": "2025-01-14T15:28:34.013781+00:00"} {"global_step": 1762, "acc_step": 0, "speed/wps": 12837.429972243834, "speed/FLOPS": 201629478191461.2, "speed/curr_iter_time": 1.2881, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.23620782792568207, "optim/lr": 0.0010566, "optim/total_tokens": 461897728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354145, "loss/out": 3.8366098403930664, "created_at": "2025-01-14T15:28:39.122519+00:00"} {"global_step": 1763, "acc_step": 0, "speed/wps": 12838.064183081904, "speed/FLOPS": 201639439344170.1, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.23502498865127563, "optim/lr": 0.0010572, "optim/total_tokens": 462159872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485608, "loss/out": 3.840045213699341, "created_at": "2025-01-14T15:28:44.231281+00:00"} {"global_step": 1764, "acc_step": 0, "speed/wps": 12842.812894668821, "speed/FLOPS": 201714024385056.16, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.001, "optim/grad_norm": 0.3066922724246979, "optim/lr": 0.0010578, "optim/total_tokens": 462422016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482522, "loss/out": 3.8661980628967285, "created_at": "2025-01-14T15:28:49.339105+00:00"} {"global_step": 1765, "acc_step": 0, "speed/wps": 12841.521540747812, "speed/FLOPS": 201693741897220.47, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.27867045998573303, "optim/lr": 0.0010584, "optim/total_tokens": 462684160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490411, "loss/out": 3.7885451316833496, "created_at": "2025-01-14T15:28:54.451966+00:00"} {"global_step": 1766, "acc_step": 0, "speed/wps": 12846.95184640021, "speed/FLOPS": 201779032309513.3, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.24085120856761932, "optim/lr": 0.001059, "optim/total_tokens": 462946304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493816, "loss/out": 3.84297251701355, "created_at": "2025-01-14T15:28:59.557561+00:00"} {"global_step": 1767, "acc_step": 0, "speed/wps": 12844.045899122451, "speed/FLOPS": 201733390414326.1, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.241943821310997, "optim/lr": 0.0010596000000000002, "optim/total_tokens": 463208448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313754, "loss/out": 3.833159923553467, "created_at": "2025-01-14T15:29:04.663812+00:00"} {"global_step": 1768, "acc_step": 0, "speed/wps": 12847.211862439492, "speed/FLOPS": 201783116218709.72, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20712749660015106, "optim/lr": 0.0010602, "optim/total_tokens": 463470592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472720, "loss/out": 3.8337242603302, "created_at": "2025-01-14T15:29:09.769864+00:00"} {"global_step": 1769, "acc_step": 0, "speed/wps": 12841.983654910735, "speed/FLOPS": 201701000035160.6, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.19345170259475708, "optim/lr": 0.0010608000000000002, "optim/total_tokens": 463732736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 384078, "loss/out": 3.8233821392059326, "created_at": "2025-01-14T15:29:14.877347+00:00"} {"global_step": 1770, "acc_step": 0, "speed/wps": 12844.116640509017, "speed/FLOPS": 201734501505010.97, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20022635161876678, "optim/lr": 0.0010614, "optim/total_tokens": 463994880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497235, "loss/out": 3.8207883834838867, "created_at": "2025-01-14T15:29:19.985041+00:00"} {"global_step": 1771, "acc_step": 0, "speed/wps": 12842.285068343252, "speed/FLOPS": 201705734147303.72, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1739778369665146, "optim/lr": 0.001062, "optim/total_tokens": 464257024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499899, "loss/out": 3.7457761764526367, "created_at": "2025-01-14T15:29:25.101900+00:00"} {"global_step": 1772, "acc_step": 0, "speed/wps": 12846.860563351216, "speed/FLOPS": 201777598583792.8, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20522063970565796, "optim/lr": 0.0010626000000000001, "optim/total_tokens": 464519168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306261, "loss/out": 3.708961248397827, "created_at": "2025-01-14T15:29:30.207706+00:00"} {"global_step": 1773, "acc_step": 0, "speed/wps": 12820.134279321197, "speed/FLOPS": 201357825567338.53, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21295663714408875, "optim/lr": 0.0010632, "optim/total_tokens": 464781312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498115, "loss/out": 3.7353267669677734, "created_at": "2025-01-14T15:29:35.323524+00:00"} {"global_step": 1774, "acc_step": 0, "speed/wps": 12843.548961381015, "speed/FLOPS": 201725585324232.84, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2647903561592102, "optim/lr": 0.0010638000000000002, "optim/total_tokens": 465043456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336220, "loss/out": 3.887357234954834, "created_at": "2025-01-14T15:29:40.432634+00:00"} {"global_step": 1775, "acc_step": 0, "speed/wps": 12834.152692731814, "speed/FLOPS": 201578004013270.94, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.24890589714050293, "optim/lr": 0.0010644, "optim/total_tokens": 465305600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470304, "loss/out": 3.75111722946167, "created_at": "2025-01-14T15:29:45.544382+00:00"} {"global_step": 1776, "acc_step": 0, "speed/wps": 12838.72604784839, "speed/FLOPS": 201649834839825.25, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2605037987232208, "optim/lr": 0.001065, "optim/total_tokens": 465567744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423127, "loss/out": 3.848062515258789, "created_at": "2025-01-14T15:29:50.653878+00:00"} {"global_step": 1777, "acc_step": 0, "speed/wps": 12843.92040687496, "speed/FLOPS": 201731419386134.84, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2684195041656494, "optim/lr": 0.0010656, "optim/total_tokens": 465829888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501822, "loss/out": 3.852776050567627, "created_at": "2025-01-14T15:29:55.762146+00:00"} {"global_step": 1778, "acc_step": 0, "speed/wps": 12843.152425710543, "speed/FLOPS": 201719357186630.94, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.2237435281276703, "optim/lr": 0.0010662, "optim/total_tokens": 466092032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501532, "loss/out": 3.885582447052002, "created_at": "2025-01-14T15:30:00.870072+00:00"} {"global_step": 1779, "acc_step": 0, "speed/wps": 12844.392427999488, "speed/FLOPS": 201738833126520.53, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.26545995473861694, "optim/lr": 0.0010668000000000001, "optim/total_tokens": 466354176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302630, "loss/out": 3.8390941619873047, "created_at": "2025-01-14T15:30:05.978155+00:00"} {"global_step": 1780, "acc_step": 0, "speed/wps": 12847.615690496674, "speed/FLOPS": 201789458893265.66, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2489885836839676, "optim/lr": 0.0010674, "optim/total_tokens": 466616320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485658, "loss/out": 3.7641563415527344, "created_at": "2025-01-14T15:30:11.082790+00:00"} {"global_step": 1781, "acc_step": 0, "speed/wps": 12841.792602382242, "speed/FLOPS": 201697999292666.9, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22777843475341797, "optim/lr": 0.001068, "optim/total_tokens": 466878464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 281178, "loss/out": 3.7425498962402344, "created_at": "2025-01-14T15:30:16.191535+00:00"} {"global_step": 1782, "acc_step": 0, "speed/wps": 12841.164832770073, "speed/FLOPS": 201688139308261.97, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2567620575428009, "optim/lr": 0.0010686, "optim/total_tokens": 467140608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491505, "loss/out": 3.9669430255889893, "created_at": "2025-01-14T15:30:21.298776+00:00"} {"global_step": 1783, "acc_step": 0, "speed/wps": 12848.434957853788, "speed/FLOPS": 201802326612900.47, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20625504851341248, "optim/lr": 0.0010692, "optim/total_tokens": 467402752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430268, "loss/out": 3.859541416168213, "created_at": "2025-01-14T15:30:26.405181+00:00"} {"global_step": 1784, "acc_step": 0, "speed/wps": 12847.14893261551, "speed/FLOPS": 201782127819349.44, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21928231418132782, "optim/lr": 0.0010697999999999999, "optim/total_tokens": 467664896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496480, "loss/out": 3.8939006328582764, "created_at": "2025-01-14T15:30:31.510176+00:00"} {"global_step": 1785, "acc_step": 0, "speed/wps": 12843.036179200612, "speed/FLOPS": 201717531375453.03, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2040128856897354, "optim/lr": 0.0010704, "optim/total_tokens": 467927040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494625, "loss/out": 3.8291280269622803, "created_at": "2025-01-14T15:30:36.626029+00:00"} {"global_step": 1786, "acc_step": 0, "speed/wps": 12842.297977538801, "speed/FLOPS": 201705936903958.9, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2165548950433731, "optim/lr": 0.001071, "optim/total_tokens": 468189184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305664, "loss/out": 3.8131024837493896, "created_at": "2025-01-14T15:30:41.739819+00:00"} {"global_step": 1787, "acc_step": 0, "speed/wps": 12844.549280604413, "speed/FLOPS": 201741296712218.97, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22410760819911957, "optim/lr": 0.0010716, "optim/total_tokens": 468451328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461844, "loss/out": 3.798882484436035, "created_at": "2025-01-14T15:30:46.857704+00:00"} {"global_step": 1788, "acc_step": 0, "speed/wps": 12843.618991688854, "speed/FLOPS": 201726685246449.2, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22397109866142273, "optim/lr": 0.0010722, "optim/total_tokens": 468713472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288055, "loss/out": 3.8026742935180664, "created_at": "2025-01-14T15:30:51.963833+00:00"} {"global_step": 1789, "acc_step": 0, "speed/wps": 12844.201807842443, "speed/FLOPS": 201735839175014.6, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.27039772272109985, "optim/lr": 0.0010727999999999998, "optim/total_tokens": 468975616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470624, "loss/out": 3.816265344619751, "created_at": "2025-01-14T15:30:57.072515+00:00"} {"global_step": 1790, "acc_step": 0, "speed/wps": 12843.284306974023, "speed/FLOPS": 201721428563097.75, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.2301265299320221, "optim/lr": 0.0010734, "optim/total_tokens": 469237760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432953, "loss/out": 3.9161624908447266, "created_at": "2025-01-14T15:31:02.181142+00:00"} {"global_step": 1791, "acc_step": 0, "speed/wps": 12842.014815559085, "speed/FLOPS": 201701489455962.16, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.2188151776790619, "optim/lr": 0.001074, "optim/total_tokens": 469499904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496716, "loss/out": 3.777144432067871, "created_at": "2025-01-14T15:31:07.293284+00:00"} {"global_step": 1792, "acc_step": 0, "speed/wps": 12843.453436902884, "speed/FLOPS": 201724084981039.4, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.25705811381340027, "optim/lr": 0.0010746, "optim/total_tokens": 469762048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495151, "loss/out": 3.8672595024108887, "created_at": "2025-01-14T15:31:12.402588+00:00"} {"global_step": 1793, "acc_step": 0, "speed/wps": 12839.170578877372, "speed/FLOPS": 201656816810485.06, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3215225338935852, "optim/lr": 0.0010752000000000001, "optim/total_tokens": 470024192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294251, "loss/out": 3.894646167755127, "created_at": "2025-01-14T15:31:17.513777+00:00"} {"global_step": 1794, "acc_step": 0, "speed/wps": 12842.11029520204, "speed/FLOPS": 201702989094957.53, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2384018898010254, "optim/lr": 0.0010758, "optim/total_tokens": 470286336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485138, "loss/out": 3.9490578174591064, "created_at": "2025-01-14T15:31:22.620440+00:00"} {"global_step": 1795, "acc_step": 0, "speed/wps": 12845.025205009375, "speed/FLOPS": 201748771758987.53, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.19948317110538483, "optim/lr": 0.0010764000000000001, "optim/total_tokens": 470548480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401462, "loss/out": 3.673271656036377, "created_at": "2025-01-14T15:31:27.726024+00:00"} {"global_step": 1796, "acc_step": 0, "speed/wps": 12847.106922668787, "speed/FLOPS": 201781467995408.44, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.24819451570510864, "optim/lr": 0.001077, "optim/total_tokens": 470810624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502273, "loss/out": 3.9907896518707275, "created_at": "2025-01-14T15:31:32.838929+00:00"} {"global_step": 1797, "acc_step": 0, "speed/wps": 12842.076663894355, "speed/FLOPS": 201702460869056.06, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2539609372615814, "optim/lr": 0.0010776000000000002, "optim/total_tokens": 471072768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488549, "loss/out": 4.0002055168151855, "created_at": "2025-01-14T15:31:37.952488+00:00"} {"global_step": 1798, "acc_step": 0, "speed/wps": 12844.620735595465, "speed/FLOPS": 201742419011042.88, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1785292625427246, "optim/lr": 0.0010782, "optim/total_tokens": 471334912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297424, "loss/out": 3.95302677154541, "created_at": "2025-01-14T15:31:43.058985+00:00"} {"global_step": 1799, "acc_step": 0, "speed/wps": 12838.680136654993, "speed/FLOPS": 201649113741444.38, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18461036682128906, "optim/lr": 0.0010788, "optim/total_tokens": 471597056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467927, "loss/out": 3.988671064376831, "created_at": "2025-01-14T15:31:48.167593+00:00"} {"global_step": 1800, "acc_step": 0, "speed/wps": 12840.202192738765, "speed/FLOPS": 201673019723764.25, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1870948225259781, "optim/lr": 0.0010794, "optim/total_tokens": 471859200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348250, "loss/out": 3.8357298374176025, "created_at": "2025-01-14T15:31:53.277419+00:00"} {"global_step": 1801, "acc_step": 0, "speed/wps": 12849.268691174373, "speed/FLOPS": 201815421540369.9, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2194053679704666, "optim/lr": 0.00108, "optim/total_tokens": 472121344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479331, "loss/out": 3.958240509033203, "created_at": "2025-01-14T15:31:58.393484+00:00"} {"global_step": 1802, "acc_step": 0, "speed/wps": 12845.186514070234, "speed/FLOPS": 201751305339450.22, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0066004991531372, "optim/lr": 0.0010806000000000001, "optim/total_tokens": 472383488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361964, "loss/out": 3.8030967712402344, "created_at": "2025-01-14T15:32:03.499079+00:00"} {"global_step": 1803, "acc_step": 0, "speed/wps": 12842.358059076345, "speed/FLOPS": 201706880567066.72, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.178394615650177, "optim/lr": 0.0010812, "optim/total_tokens": 472645632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498044, "loss/out": 3.815777540206909, "created_at": "2025-01-14T15:32:08.608908+00:00"} {"global_step": 1804, "acc_step": 0, "speed/wps": 12845.835237995723, "speed/FLOPS": 201761494440140.12, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1859610676765442, "optim/lr": 0.0010818, "optim/total_tokens": 472907776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485138, "loss/out": 3.8333487510681152, "created_at": "2025-01-14T15:32:13.714657+00:00"} {"global_step": 1805, "acc_step": 0, "speed/wps": 12844.597102396941, "speed/FLOPS": 201742047819184.66, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.181529700756073, "optim/lr": 0.0010824, "optim/total_tokens": 473169920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481715, "loss/out": 3.67720890045166, "created_at": "2025-01-14T15:32:18.821670+00:00"} {"global_step": 1806, "acc_step": 0, "speed/wps": 12843.244286624302, "speed/FLOPS": 201720799988511.66, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21664083003997803, "optim/lr": 0.001083, "optim/total_tokens": 473432064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491685, "loss/out": 3.9484593868255615, "created_at": "2025-01-14T15:32:23.930828+00:00"} {"global_step": 1807, "acc_step": 0, "speed/wps": 12842.367260612875, "speed/FLOPS": 201707025089842.16, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.012212872505188, "optim/lr": 0.0010836, "optim/total_tokens": 473694208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303334, "loss/out": 3.8555140495300293, "created_at": "2025-01-14T15:32:29.039070+00:00"} {"global_step": 1808, "acc_step": 0, "speed/wps": 12846.795795859482, "speed/FLOPS": 201776581321335.25, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.001, "optim/grad_norm": 1.2982310056686401, "optim/lr": 0.0010842, "optim/total_tokens": 473956352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490175, "loss/out": 4.399818420410156, "created_at": "2025-01-14T15:32:34.143955+00:00"} {"global_step": 1809, "acc_step": 0, "speed/wps": 12841.828402327557, "speed/FLOPS": 201698561580002.38, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0367534160614014, "optim/lr": 0.0010848, "optim/total_tokens": 474218496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471924, "loss/out": 4.176558971405029, "created_at": "2025-01-14T15:32:39.281544+00:00"} {"global_step": 1810, "acc_step": 0, "speed/wps": 12841.414249752781, "speed/FLOPS": 201692056744710.25, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5442543029785156, "optim/lr": 0.0010854, "optim/total_tokens": 474480640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358516, "loss/out": 3.8895716667175293, "created_at": "2025-01-14T15:32:44.390824+00:00"} {"global_step": 1811, "acc_step": 0, "speed/wps": 12842.320969470968, "speed/FLOPS": 201706298023847.8, "speed/curr_iter_time": 1.2874, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1687614917755127, "optim/lr": 0.001086, "optim/total_tokens": 474742784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487220, "loss/out": 3.7901623249053955, "created_at": "2025-01-14T15:32:49.501941+00:00"} {"global_step": 1812, "acc_step": 0, "speed/wps": 12844.417042787461, "speed/FLOPS": 201739219735589.94, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7020095586776733, "optim/lr": 0.0010866, "optim/total_tokens": 475004928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463984, "loss/out": 4.004159927368164, "created_at": "2025-01-14T15:32:54.608172+00:00"} {"global_step": 1813, "acc_step": 0, "speed/wps": 12844.56275000988, "speed/FLOPS": 201741508267740.72, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6126168966293335, "optim/lr": 0.0010872, "optim/total_tokens": 475267072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 3.978147268295288, "created_at": "2025-01-14T15:32:59.716080+00:00"} {"global_step": 1814, "acc_step": 0, "speed/wps": 12843.369789101043, "speed/FLOPS": 201722771177367.38, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.2254984378814697, "optim/lr": 0.0010877999999999999, "optim/total_tokens": 475529216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 4.52518367767334, "created_at": "2025-01-14T15:33:04.826634+00:00"} {"global_step": 1815, "acc_step": 0, "speed/wps": 12839.411318461998, "speed/FLOPS": 201660597956472.47, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3992544412612915, "optim/lr": 0.0010884, "optim/total_tokens": 475791360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484902, "loss/out": 4.3232221603393555, "created_at": "2025-01-14T15:33:09.940419+00:00"} {"global_step": 1816, "acc_step": 0, "speed/wps": 12836.574147391291, "speed/FLOPS": 201616036286125.44, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.2027747631073, "optim/lr": 0.001089, "optim/total_tokens": 476053504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 4.485569000244141, "created_at": "2025-01-14T15:33:15.052271+00:00"} {"global_step": 1817, "acc_step": 0, "speed/wps": 12853.875283236597, "speed/FLOPS": 201887774398827.97, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.7917861938476562, "optim/lr": 0.0010896, "optim/total_tokens": 476315648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317059, "loss/out": 5.566603660583496, "created_at": "2025-01-14T15:33:20.156746+00:00"} {"global_step": 1818, "acc_step": 0, "speed/wps": 12841.647363423579, "speed/FLOPS": 201695718115241.88, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.629517912864685, "optim/lr": 0.0010902, "optim/total_tokens": 476577792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 4.82336950302124, "created_at": "2025-01-14T15:33:25.267006+00:00"} {"global_step": 1819, "acc_step": 0, "speed/wps": 12847.27756530492, "speed/FLOPS": 201784148172496.47, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1759217977523804, "optim/lr": 0.0010908, "optim/total_tokens": 476839936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432365, "loss/out": 4.631283760070801, "created_at": "2025-01-14T15:33:30.371795+00:00"} {"global_step": 1820, "acc_step": 0, "speed/wps": 12846.480751139354, "speed/FLOPS": 201771633111088.84, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2832121849060059, "optim/lr": 0.0010914, "optim/total_tokens": 477102080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 4.515436172485352, "created_at": "2025-01-14T15:33:35.477057+00:00"} {"global_step": 1821, "acc_step": 0, "speed/wps": 12840.924828805328, "speed/FLOPS": 201684369716195.62, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.7890850901603699, "optim/lr": 0.001092, "optim/total_tokens": 477364224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 4.326633453369141, "created_at": "2025-01-14T15:33:40.588346+00:00"} {"global_step": 1822, "acc_step": 0, "speed/wps": 12840.972944636424, "speed/FLOPS": 201685125441441.06, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7847440242767334, "optim/lr": 0.0010926, "optim/total_tokens": 477626368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318319, "loss/out": 4.3584465980529785, "created_at": "2025-01-14T15:33:45.697235+00:00"} {"global_step": 1823, "acc_step": 0, "speed/wps": 12837.96090571954, "speed/FLOPS": 201637817231276.2, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0204304456710815, "optim/lr": 0.0010932000000000001, "optim/total_tokens": 477888512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 4.343440055847168, "created_at": "2025-01-14T15:33:50.808439+00:00"} {"global_step": 1824, "acc_step": 0, "speed/wps": 12841.695543257109, "speed/FLOPS": 201696474845733.47, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7161767482757568, "optim/lr": 0.0010938, "optim/total_tokens": 478150656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 4.297247886657715, "created_at": "2025-01-14T15:33:55.915556+00:00"} {"global_step": 1825, "acc_step": 0, "speed/wps": 12848.729382619278, "speed/FLOPS": 201806950958424.5, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6863335967063904, "optim/lr": 0.0010944000000000001, "optim/total_tokens": 478412800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504506, "loss/out": 4.260922431945801, "created_at": "2025-01-14T15:34:01.023647+00:00"} {"global_step": 1826, "acc_step": 0, "speed/wps": 12851.039191872642, "speed/FLOPS": 201843229686759.12, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6189444661140442, "optim/lr": 0.001095, "optim/total_tokens": 478674944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 4.181692600250244, "created_at": "2025-01-14T15:34:06.132295+00:00"} {"global_step": 1827, "acc_step": 0, "speed/wps": 12851.032348282628, "speed/FLOPS": 201843122198773.88, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9659280776977539, "optim/lr": 0.0010956000000000002, "optim/total_tokens": 478937088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 4.098256587982178, "created_at": "2025-01-14T15:34:11.240888+00:00"} {"global_step": 1828, "acc_step": 0, "speed/wps": 12845.218775899626, "speed/FLOPS": 201751812055813.22, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3888074457645416, "optim/lr": 0.0010962, "optim/total_tokens": 479199232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461580, "loss/out": 3.9811551570892334, "created_at": "2025-01-14T15:34:16.349478+00:00"} {"global_step": 1829, "acc_step": 0, "speed/wps": 12851.014903732408, "speed/FLOPS": 201842848208141.34, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2970942258834839, "optim/lr": 0.0010968, "optim/total_tokens": 479461376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 4.0724077224731445, "created_at": "2025-01-14T15:34:21.453575+00:00"} {"global_step": 1830, "acc_step": 0, "speed/wps": 12846.041644495166, "speed/FLOPS": 201764736337846.12, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31947511434555054, "optim/lr": 0.0010974000000000001, "optim/total_tokens": 479723520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 4.111994743347168, "created_at": "2025-01-14T15:34:26.561624+00:00"} {"global_step": 1831, "acc_step": 0, "speed/wps": 12842.430874856524, "speed/FLOPS": 201708024238953.8, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.30452606081962585, "optim/lr": 0.001098, "optim/total_tokens": 479985664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313309, "loss/out": 4.017817974090576, "created_at": "2025-01-14T15:34:31.678700+00:00"} {"global_step": 1832, "acc_step": 0, "speed/wps": 12845.373479707994, "speed/FLOPS": 201754241891708.94, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.31184807419776917, "optim/lr": 0.0010986000000000001, "optim/total_tokens": 480247808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.9800262451171875, "created_at": "2025-01-14T15:34:36.786872+00:00"} {"global_step": 1833, "acc_step": 0, "speed/wps": 12846.533606182536, "speed/FLOPS": 201772463272172.47, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.32559144496917725, "optim/lr": 0.0010992, "optim/total_tokens": 480509952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 4.0275726318359375, "created_at": "2025-01-14T15:34:41.894576+00:00"} {"global_step": 1834, "acc_step": 0, "speed/wps": 12850.543604316046, "speed/FLOPS": 201835445803174.12, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2534179389476776, "optim/lr": 0.0010998, "optim/total_tokens": 480772096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 4.00686502456665, "created_at": "2025-01-14T15:34:47.004710+00:00"} {"global_step": 1835, "acc_step": 0, "speed/wps": 12840.231632034289, "speed/FLOPS": 201673482108354.28, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25356200337409973, "optim/lr": 0.0011004, "optim/total_tokens": 481034240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 4.001791477203369, "created_at": "2025-01-14T15:34:52.114721+00:00"} {"global_step": 1836, "acc_step": 0, "speed/wps": 12846.708298043963, "speed/FLOPS": 201775207047907.94, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.20860612392425537, "optim/lr": 0.001101, "optim/total_tokens": 481296384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317214, "loss/out": 4.070089340209961, "created_at": "2025-01-14T15:34:57.219770+00:00"} {"global_step": 1837, "acc_step": 0, "speed/wps": 12850.458453767838, "speed/FLOPS": 201834108396805.4, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.21584221720695496, "optim/lr": 0.0011016, "optim/total_tokens": 481558528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 3.9405407905578613, "created_at": "2025-01-14T15:35:02.325855+00:00"} {"global_step": 1838, "acc_step": 0, "speed/wps": 12843.43679586367, "speed/FLOPS": 201723823610651.34, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19884054362773895, "optim/lr": 0.0011022, "optim/total_tokens": 481820672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379737, "loss/out": 3.936687469482422, "created_at": "2025-01-14T15:35:07.436427+00:00"} {"global_step": 1839, "acc_step": 0, "speed/wps": 12846.426482359107, "speed/FLOPS": 201770780745321.97, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.23032420873641968, "optim/lr": 0.0011028, "optim/total_tokens": 482082816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 4.082603454589844, "created_at": "2025-01-14T15:35:12.544123+00:00"} {"global_step": 1840, "acc_step": 0, "speed/wps": 12844.409198117675, "speed/FLOPS": 201739096524271.22, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.21364812552928925, "optim/lr": 0.0011034, "optim/total_tokens": 482344960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375761, "loss/out": 3.9290056228637695, "created_at": "2025-01-14T15:35:17.653500+00:00"} {"global_step": 1841, "acc_step": 0, "speed/wps": 12848.578572111292, "speed/FLOPS": 201804582272161.22, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20626860857009888, "optim/lr": 0.001104, "optim/total_tokens": 482607104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457652, "loss/out": 3.880434989929199, "created_at": "2025-01-14T15:35:22.759071+00:00"} {"global_step": 1842, "acc_step": 0, "speed/wps": 12848.460603943082, "speed/FLOPS": 201802729419974.25, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.20991694927215576, "optim/lr": 0.0011046, "optim/total_tokens": 482869248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 378157, "loss/out": 3.991806745529175, "created_at": "2025-01-14T15:35:27.864266+00:00"} {"global_step": 1843, "acc_step": 0, "speed/wps": 12853.83745556739, "speed/FLOPS": 201887180263302.4, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21017718315124512, "optim/lr": 0.0011052, "optim/total_tokens": 483131392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 3.940702438354492, "created_at": "2025-01-14T15:35:32.971532+00:00"} {"global_step": 1844, "acc_step": 0, "speed/wps": 12847.357942692946, "speed/FLOPS": 201785410609825.53, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2228500247001648, "optim/lr": 0.0011057999999999999, "optim/total_tokens": 483393536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.7639074325561523, "created_at": "2025-01-14T15:35:38.079181+00:00"} {"global_step": 1845, "acc_step": 0, "speed/wps": 12846.378203699114, "speed/FLOPS": 201770022462624.94, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3153783082962036, "optim/lr": 0.0011064, "optim/total_tokens": 483655680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.9008169174194336, "created_at": "2025-01-14T15:35:43.187499+00:00"} {"global_step": 1846, "acc_step": 0, "speed/wps": 12843.945333992837, "speed/FLOPS": 201731810900774.53, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.22590947151184082, "optim/lr": 0.001107, "optim/total_tokens": 483917824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 3.9201178550720215, "created_at": "2025-01-14T15:35:48.296656+00:00"} {"global_step": 1847, "acc_step": 0, "speed/wps": 12850.042136986776, "speed/FLOPS": 201827569569679.62, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25627654790878296, "optim/lr": 0.0011076, "optim/total_tokens": 484179968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309312, "loss/out": 3.7175557613372803, "created_at": "2025-01-14T15:35:53.404246+00:00"} {"global_step": 1848, "acc_step": 0, "speed/wps": 12846.100841784304, "speed/FLOPS": 201765666112617.62, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2184661477804184, "optim/lr": 0.0011082, "optim/total_tokens": 484442112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.870051860809326, "created_at": "2025-01-14T15:35:58.514275+00:00"} {"global_step": 1849, "acc_step": 0, "speed/wps": 12848.999695202849, "speed/FLOPS": 201811196588997.9, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.21716967225074768, "optim/lr": 0.0011088, "optim/total_tokens": 484704256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 3.9507193565368652, "created_at": "2025-01-14T15:36:03.618385+00:00"} {"global_step": 1850, "acc_step": 0, "speed/wps": 12846.353325767086, "speed/FLOPS": 201769631720516.56, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.19551986455917358, "optim/lr": 0.0011094, "optim/total_tokens": 484966400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.9036405086517334, "created_at": "2025-01-14T15:36:08.723518+00:00"} {"global_step": 1851, "acc_step": 0, "speed/wps": 12843.488451440493, "speed/FLOPS": 201724634932467.56, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.23335222899913788, "optim/lr": 0.00111, "optim/total_tokens": 485228544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.848710298538208, "created_at": "2025-01-14T15:36:13.832771+00:00"} {"global_step": 1852, "acc_step": 0, "speed/wps": 12842.072686740208, "speed/FLOPS": 201702398402385.03, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.1823820024728775, "optim/lr": 0.0011106, "optim/total_tokens": 485490688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287143, "loss/out": 3.905857563018799, "created_at": "2025-01-14T15:36:18.948039+00:00"} {"global_step": 1853, "acc_step": 0, "speed/wps": 12847.81742286821, "speed/FLOPS": 201792627377369.97, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20530131459236145, "optim/lr": 0.0011112000000000001, "optim/total_tokens": 485752832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480954, "loss/out": 3.846769094467163, "created_at": "2025-01-14T15:36:24.052687+00:00"} {"global_step": 1854, "acc_step": 0, "speed/wps": 12844.398490838783, "speed/FLOPS": 201738928351742.88, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2171049267053604, "optim/lr": 0.0011118, "optim/total_tokens": 486014976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 3.880192279815674, "created_at": "2025-01-14T15:36:29.159423+00:00"} {"global_step": 1855, "acc_step": 0, "speed/wps": 12844.366662168575, "speed/FLOPS": 201738428438740.1, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20683042705059052, "optim/lr": 0.0011124000000000002, "optim/total_tokens": 486277120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 3.8367717266082764, "created_at": "2025-01-14T15:36:34.265287+00:00"} {"global_step": 1856, "acc_step": 0, "speed/wps": 12845.166900063912, "speed/FLOPS": 201750997274528.47, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20269784331321716, "optim/lr": 0.001113, "optim/total_tokens": 486539264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321540, "loss/out": 3.880537986755371, "created_at": "2025-01-14T15:36:39.372651+00:00"} {"global_step": 1857, "acc_step": 0, "speed/wps": 12847.066083381469, "speed/FLOPS": 201780826558282.06, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22167980670928955, "optim/lr": 0.0011136, "optim/total_tokens": 486801408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 3.97480845451355, "created_at": "2025-01-14T15:36:44.479737+00:00"} {"global_step": 1858, "acc_step": 0, "speed/wps": 12842.364523119224, "speed/FLOPS": 201706982093742.66, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23178806900978088, "optim/lr": 0.0011142, "optim/total_tokens": 487063552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357199, "loss/out": 3.8272151947021484, "created_at": "2025-01-14T15:36:49.591589+00:00"} {"global_step": 1859, "acc_step": 0, "speed/wps": 12832.345117685412, "speed/FLOPS": 201549613563298.8, "speed/curr_iter_time": 1.2888, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.18487587571144104, "optim/lr": 0.0011148, "optim/total_tokens": 487325696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 3.897919178009033, "created_at": "2025-01-14T15:36:54.703226+00:00"} {"global_step": 1860, "acc_step": 0, "speed/wps": 12832.126713933021, "speed/FLOPS": 201546183232249.28, "speed/curr_iter_time": 1.2878, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19943000376224518, "optim/lr": 0.0011154000000000001, "optim/total_tokens": 487587840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 3.8619699478149414, "created_at": "2025-01-14T15:36:59.817058+00:00"} {"global_step": 1861, "acc_step": 0, "speed/wps": 12841.550315867664, "speed/FLOPS": 201694193850018.9, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1983441412448883, "optim/lr": 0.001116, "optim/total_tokens": 487849984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319225, "loss/out": 3.920670986175537, "created_at": "2025-01-14T15:37:04.926320+00:00"} {"global_step": 1862, "acc_step": 0, "speed/wps": 12844.940971663884, "speed/FLOPS": 201747448758547.88, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.23023013770580292, "optim/lr": 0.0011166, "optim/total_tokens": 488112128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.774137496948242, "created_at": "2025-01-14T15:37:10.040229+00:00"} {"global_step": 1863, "acc_step": 0, "speed/wps": 12844.50807541844, "speed/FLOPS": 201740649528151.38, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2065262645483017, "optim/lr": 0.0011172, "optim/total_tokens": 488374272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477763, "loss/out": 3.854743719100952, "created_at": "2025-01-14T15:37:15.149017+00:00"} {"global_step": 1864, "acc_step": 0, "speed/wps": 12849.627172893133, "speed/FLOPS": 201821051988368.53, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22203122079372406, "optim/lr": 0.0011178, "optim/total_tokens": 488636416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 3.904278516769409, "created_at": "2025-01-14T15:37:20.253283+00:00"} {"global_step": 1865, "acc_step": 0, "speed/wps": 12843.69828949852, "speed/FLOPS": 201727930727516.56, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23278403282165527, "optim/lr": 0.0011184, "optim/total_tokens": 488898560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504796, "loss/out": 3.8895485401153564, "created_at": "2025-01-14T15:37:25.360554+00:00"} {"global_step": 1866, "acc_step": 0, "speed/wps": 12844.714570567832, "speed/FLOPS": 201743892818226.66, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22552521526813507, "optim/lr": 0.001119, "optim/total_tokens": 489160704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.817779302597046, "created_at": "2025-01-14T15:37:30.467549+00:00"} {"global_step": 1867, "acc_step": 0, "speed/wps": 12847.059546586497, "speed/FLOPS": 201780723888934.53, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1841098666191101, "optim/lr": 0.0011196, "optim/total_tokens": 489422848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.8754830360412598, "created_at": "2025-01-14T15:37:35.576390+00:00"} {"global_step": 1868, "acc_step": 0, "speed/wps": 12841.891289412857, "speed/FLOPS": 201699549308092.9, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21736295521259308, "optim/lr": 0.0011202, "optim/total_tokens": 489684992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 3.733224630355835, "created_at": "2025-01-14T15:37:40.685447+00:00"} {"global_step": 1869, "acc_step": 0, "speed/wps": 12847.710378497904, "speed/FLOPS": 201790946098439.94, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.20767194032669067, "optim/lr": 0.0011208, "optim/total_tokens": 489947136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.7871012687683105, "created_at": "2025-01-14T15:37:45.791945+00:00"} {"global_step": 1870, "acc_step": 0, "speed/wps": 12849.85137491374, "speed/FLOPS": 201824573389188.03, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21392828226089478, "optim/lr": 0.0011214, "optim/total_tokens": 490209280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318558, "loss/out": 3.7698094844818115, "created_at": "2025-01-14T15:37:50.895857+00:00"} {"global_step": 1871, "acc_step": 0, "speed/wps": 12846.747767720763, "speed/FLOPS": 201775826973419.1, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19973139464855194, "optim/lr": 0.001122, "optim/total_tokens": 490471424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 3.8839800357818604, "created_at": "2025-01-14T15:37:56.002914+00:00"} {"global_step": 1872, "acc_step": 0, "speed/wps": 12850.07911867782, "speed/FLOPS": 201828150417955.62, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23139089345932007, "optim/lr": 0.0011225999999999999, "optim/total_tokens": 490733568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 346546, "loss/out": 3.9267563819885254, "created_at": "2025-01-14T15:38:01.109432+00:00"} {"global_step": 1873, "acc_step": 0, "speed/wps": 12848.438399378483, "speed/FLOPS": 201802380666774.97, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19848507642745972, "optim/lr": 0.0011232, "optim/total_tokens": 490995712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.8216090202331543, "created_at": "2025-01-14T15:38:06.213665+00:00"} {"global_step": 1874, "acc_step": 0, "speed/wps": 12841.637244274947, "speed/FLOPS": 201695559180107.44, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.21673579514026642, "optim/lr": 0.0011237999999999999, "optim/total_tokens": 491257856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 3.7703006267547607, "created_at": "2025-01-14T15:38:11.320792+00:00"} {"global_step": 1875, "acc_step": 0, "speed/wps": 12844.304515128442, "speed/FLOPS": 201737452334075.8, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.001, "optim/grad_norm": 0.23618721961975098, "optim/lr": 0.0011244, "optim/total_tokens": 491520000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.9237427711486816, "created_at": "2025-01-14T15:38:16.426850+00:00"} {"global_step": 1876, "acc_step": 0, "speed/wps": 12846.620001673782, "speed/FLOPS": 201773820232082.3, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23652467131614685, "optim/lr": 0.0011250000000000001, "optim/total_tokens": 491782144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.7784411907196045, "created_at": "2025-01-14T15:38:21.533208+00:00"} {"global_step": 1877, "acc_step": 0, "speed/wps": 12844.533626119404, "speed/FLOPS": 201741050837020.38, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22209659218788147, "optim/lr": 0.0011256, "optim/total_tokens": 492044288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326058, "loss/out": 3.9338607788085938, "created_at": "2025-01-14T15:38:26.651865+00:00"} {"global_step": 1878, "acc_step": 0, "speed/wps": 12848.600526136155, "speed/FLOPS": 201804927090289.72, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.24158218502998352, "optim/lr": 0.0011262000000000002, "optim/total_tokens": 492306432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461368, "loss/out": 3.835062265396118, "created_at": "2025-01-14T15:38:31.757626+00:00"} {"global_step": 1879, "acc_step": 0, "speed/wps": 12855.827559864972, "speed/FLOPS": 201918437585985.8, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19625435769557953, "optim/lr": 0.0011268, "optim/total_tokens": 492568576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352687, "loss/out": 3.9395036697387695, "created_at": "2025-01-14T15:38:36.864239+00:00"} {"global_step": 1880, "acc_step": 0, "speed/wps": 12851.556307670146, "speed/FLOPS": 201851351700950.47, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.20807814598083496, "optim/lr": 0.0011274000000000002, "optim/total_tokens": 492830720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471454, "loss/out": 3.8605101108551025, "created_at": "2025-01-14T15:38:41.974129+00:00"} {"global_step": 1881, "acc_step": 0, "speed/wps": 12849.738920959984, "speed/FLOPS": 201822807145316.34, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.19316530227661133, "optim/lr": 0.001128, "optim/total_tokens": 493092864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 344385, "loss/out": 3.8699214458465576, "created_at": "2025-01-14T15:38:47.083069+00:00"} {"global_step": 1882, "acc_step": 0, "speed/wps": 12844.985792810492, "speed/FLOPS": 201748152736246.06, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21138149499893188, "optim/lr": 0.0011286, "optim/total_tokens": 493355008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.9444079399108887, "created_at": "2025-01-14T15:38:52.190566+00:00"} {"global_step": 1883, "acc_step": 0, "speed/wps": 12843.529000749928, "speed/FLOPS": 201725271815092.8, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24515584111213684, "optim/lr": 0.0011292000000000001, "optim/total_tokens": 493617152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.6813480854034424, "created_at": "2025-01-14T15:38:57.296882+00:00"} {"global_step": 1884, "acc_step": 0, "speed/wps": 12850.990043041955, "speed/FLOPS": 201842457736835.6, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21952874958515167, "optim/lr": 0.0011298, "optim/total_tokens": 493879296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 3.972698211669922, "created_at": "2025-01-14T15:39:02.403198+00:00"} {"global_step": 1885, "acc_step": 0, "speed/wps": 12839.087169572222, "speed/FLOPS": 201655506752731.28, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22824645042419434, "optim/lr": 0.0011304000000000002, "optim/total_tokens": 494141440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.892073154449463, "created_at": "2025-01-14T15:39:07.513578+00:00"} {"global_step": 1886, "acc_step": 0, "speed/wps": 12851.490835641238, "speed/FLOPS": 201850323372768.5, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20544545352458954, "optim/lr": 0.001131, "optim/total_tokens": 494403584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323655, "loss/out": 3.84391713142395, "created_at": "2025-01-14T15:39:12.622172+00:00"} {"global_step": 1887, "acc_step": 0, "speed/wps": 12847.387668132662, "speed/FLOPS": 201785877488703.28, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2007777988910675, "optim/lr": 0.0011316, "optim/total_tokens": 494665728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.7697389125823975, "created_at": "2025-01-14T15:39:17.726964+00:00"} {"global_step": 1888, "acc_step": 0, "speed/wps": 12847.2719782366, "speed/FLOPS": 201784060419911.06, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.20802654325962067, "optim/lr": 0.0011322, "optim/total_tokens": 494927872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390497, "loss/out": 3.8926095962524414, "created_at": "2025-01-14T15:39:22.834163+00:00"} {"global_step": 1889, "acc_step": 0, "speed/wps": 12841.815203476623, "speed/FLOPS": 201698354273911.3, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1871110051870346, "optim/lr": 0.0011328, "optim/total_tokens": 495190016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 3.8304033279418945, "created_at": "2025-01-14T15:39:27.945857+00:00"} {"global_step": 1890, "acc_step": 0, "speed/wps": 12844.156975756126, "speed/FLOPS": 201735135025493.25, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20422540605068207, "optim/lr": 0.0011334000000000001, "optim/total_tokens": 495452160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 3.7846217155456543, "created_at": "2025-01-14T15:39:33.052528+00:00"} {"global_step": 1891, "acc_step": 0, "speed/wps": 12835.433761994365, "speed/FLOPS": 201598124966412.94, "speed/curr_iter_time": 1.289, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.20068904757499695, "optim/lr": 0.001134, "optim/total_tokens": 495714304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320967, "loss/out": 3.819021463394165, "created_at": "2025-01-14T15:39:38.163642+00:00"} {"global_step": 1892, "acc_step": 0, "speed/wps": 12840.355400386195, "speed/FLOPS": 201675426060396.5, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.2032955288887024, "optim/lr": 0.0011346, "optim/total_tokens": 495976448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468018, "loss/out": 3.8389949798583984, "created_at": "2025-01-14T15:39:43.277279+00:00"} {"global_step": 1893, "acc_step": 0, "speed/wps": 12832.380109122512, "speed/FLOPS": 201550163151901.72, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2111211121082306, "optim/lr": 0.0011352, "optim/total_tokens": 496238592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295180, "loss/out": 3.8492379188537598, "created_at": "2025-01-14T15:39:48.388131+00:00"} {"global_step": 1894, "acc_step": 0, "speed/wps": 12834.676109942395, "speed/FLOPS": 201586224999813.72, "speed/curr_iter_time": 1.288, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.21136687695980072, "optim/lr": 0.0011358, "optim/total_tokens": 496500736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.812239646911621, "created_at": "2025-01-14T15:39:53.499183+00:00"} {"global_step": 1895, "acc_step": 0, "speed/wps": 12838.032294452634, "speed/FLOPS": 201638938489427.84, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.218189999461174, "optim/lr": 0.0011364, "optim/total_tokens": 496762880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 400642, "loss/out": 3.860419988632202, "created_at": "2025-01-14T15:39:58.610890+00:00"} {"global_step": 1896, "acc_step": 0, "speed/wps": 12840.788440871935, "speed/FLOPS": 201682227556285.03, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23117655515670776, "optim/lr": 0.001137, "optim/total_tokens": 497025024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.850928783416748, "created_at": "2025-01-14T15:40:03.718214+00:00"} {"global_step": 1897, "acc_step": 0, "speed/wps": 12846.775316193176, "speed/FLOPS": 201776259660033.94, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20004071295261383, "optim/lr": 0.0011376, "optim/total_tokens": 497287168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.7034554481506348, "created_at": "2025-01-14T15:40:08.826503+00:00"} {"global_step": 1898, "acc_step": 0, "speed/wps": 12838.75254063031, "speed/FLOPS": 201650250945370.22, "speed/curr_iter_time": 1.2877, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.22160714864730835, "optim/lr": 0.0011382, "optim/total_tokens": 497549312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316688, "loss/out": 3.769023895263672, "created_at": "2025-01-14T15:40:13.939075+00:00"} {"global_step": 1899, "acc_step": 0, "speed/wps": 12847.412672041542, "speed/FLOPS": 201786270209451.22, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.19510847330093384, "optim/lr": 0.0011388, "optim/total_tokens": 497811456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473602, "loss/out": 3.8772060871124268, "created_at": "2025-01-14T15:40:19.047973+00:00"} {"global_step": 1900, "acc_step": 0, "speed/wps": 12847.769856233905, "speed/FLOPS": 201791880278014.8, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.20031927525997162, "optim/lr": 0.0011394, "optim/total_tokens": 498073600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 3.750425100326538, "created_at": "2025-01-14T15:40:24.155649+00:00"} {"global_step": 1901, "acc_step": 0, "speed/wps": 12842.26692567057, "speed/FLOPS": 201705449191698.62, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21323229372501373, "optim/lr": 0.00114, "optim/total_tokens": 498335744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480954, "loss/out": 3.8580029010772705, "created_at": "2025-01-14T15:40:29.265644+00:00"} {"global_step": 1902, "acc_step": 0, "speed/wps": 12840.343226113679, "speed/FLOPS": 201675234846717.44, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.32676705718040466, "optim/lr": 0.0011406, "optim/total_tokens": 498597888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502417, "loss/out": 3.8364715576171875, "created_at": "2025-01-14T15:40:34.378145+00:00"} {"global_step": 1903, "acc_step": 0, "speed/wps": 12837.618611521973, "speed/FLOPS": 201632441030541.9, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0016, "optim/grad_norm": 0.39956483244895935, "optim/lr": 0.0011412, "optim/total_tokens": 498860032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329371, "loss/out": 4.023638725280762, "created_at": "2025-01-14T15:40:39.491221+00:00"} {"global_step": 1904, "acc_step": 0, "speed/wps": 12834.594418727349, "speed/FLOPS": 201584941927025.94, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4306720197200775, "optim/lr": 0.0011418, "optim/total_tokens": 499122176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 3.8901007175445557, "created_at": "2025-01-14T15:40:44.603393+00:00"} {"global_step": 1905, "acc_step": 0, "speed/wps": 12836.85060162271, "speed/FLOPS": 201620378379717.78, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.4319017827510834, "optim/lr": 0.0011424, "optim/total_tokens": 499384320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 3.951047658920288, "created_at": "2025-01-14T15:40:49.712288+00:00"} {"global_step": 1906, "acc_step": 0, "speed/wps": 12833.41469274337, "speed/FLOPS": 201566412709334.2, "speed/curr_iter_time": 1.288, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.344620019197464, "optim/lr": 0.0011430000000000001, "optim/total_tokens": 499646464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502653, "loss/out": 3.8171749114990234, "created_at": "2025-01-14T15:40:54.822948+00:00"} {"global_step": 1907, "acc_step": 0, "speed/wps": 12836.332659524873, "speed/FLOPS": 201612243387344.22, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.30630409717559814, "optim/lr": 0.0011436, "optim/total_tokens": 499908608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 3.875062942504883, "created_at": "2025-01-14T15:40:59.936037+00:00"} {"global_step": 1908, "acc_step": 0, "speed/wps": 12853.314934799586, "speed/FLOPS": 201878973356625.25, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3182244598865509, "optim/lr": 0.0011442000000000002, "optim/total_tokens": 500170752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.937197685241699, "created_at": "2025-01-14T15:41:05.039089+00:00"} {"global_step": 1909, "acc_step": 0, "speed/wps": 12846.771204788422, "speed/FLOPS": 201776195084772.6, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.36562079191207886, "optim/lr": 0.0011448, "optim/total_tokens": 500432896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 3.8753039836883545, "created_at": "2025-01-14T15:41:10.144235+00:00"} {"global_step": 1910, "acc_step": 0, "speed/wps": 12839.748434645891, "speed/FLOPS": 201665892829386.7, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.335469514131546, "optim/lr": 0.0011454, "optim/total_tokens": 500695040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319663, "loss/out": 3.9804136753082275, "created_at": "2025-01-14T15:41:15.253758+00:00"} {"global_step": 1911, "acc_step": 0, "speed/wps": 12840.344130514193, "speed/FLOPS": 201675249051570.28, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.001, "optim/grad_norm": 0.40010857582092285, "optim/lr": 0.001146, "optim/total_tokens": 500957184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.919581890106201, "created_at": "2025-01-14T15:41:20.363515+00:00"} {"global_step": 1912, "acc_step": 0, "speed/wps": 12846.863673333031, "speed/FLOPS": 201777647430330.78, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2715589702129364, "optim/lr": 0.0011466, "optim/total_tokens": 501219328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351058, "loss/out": 3.88643479347229, "created_at": "2025-01-14T15:41:25.471043+00:00"} {"global_step": 1913, "acc_step": 0, "speed/wps": 12839.059295654353, "speed/FLOPS": 201655068954548.38, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.25847968459129333, "optim/lr": 0.0011472000000000001, "optim/total_tokens": 501481472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.9409003257751465, "created_at": "2025-01-14T15:41:30.579753+00:00"} {"global_step": 1914, "acc_step": 0, "speed/wps": 12847.746338113926, "speed/FLOPS": 201791510893623.25, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25551339983940125, "optim/lr": 0.0011478, "optim/total_tokens": 501743616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.8788623809814453, "created_at": "2025-01-14T15:41:35.690213+00:00"} {"global_step": 1915, "acc_step": 0, "speed/wps": 12845.697737399363, "speed/FLOPS": 201759334804328.66, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.29111409187316895, "optim/lr": 0.0011484, "optim/total_tokens": 502005760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480954, "loss/out": 3.876739978790283, "created_at": "2025-01-14T15:41:40.797601+00:00"} {"global_step": 1916, "acc_step": 0, "speed/wps": 12840.402408383747, "speed/FLOPS": 201676164385593.8, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23203682899475098, "optim/lr": 0.001149, "optim/total_tokens": 502267904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.9421145915985107, "created_at": "2025-01-14T15:41:45.907395+00:00"} {"global_step": 1917, "acc_step": 0, "speed/wps": 12844.102700819052, "speed/FLOPS": 201734282563024.94, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21221069991588593, "optim/lr": 0.0011496, "optim/total_tokens": 502530048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 339571, "loss/out": 3.89638614654541, "created_at": "2025-01-14T15:41:51.017737+00:00"} {"global_step": 1918, "acc_step": 0, "speed/wps": 12844.813593248518, "speed/FLOPS": 201745448105498.25, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22192241251468658, "optim/lr": 0.0011502, "optim/total_tokens": 502792192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 3.939995288848877, "created_at": "2025-01-14T15:41:56.123667+00:00"} {"global_step": 1919, "acc_step": 0, "speed/wps": 12846.70936751445, "speed/FLOPS": 201775223845411.53, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1997840404510498, "optim/lr": 0.0011508, "optim/total_tokens": 503054336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361228, "loss/out": 3.8072566986083984, "created_at": "2025-01-14T15:42:01.230651+00:00"} {"global_step": 1920, "acc_step": 0, "speed/wps": 12845.236518919712, "speed/FLOPS": 201752090734325.12, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.21167002618312836, "optim/lr": 0.0011514, "optim/total_tokens": 503316480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.776870012283325, "created_at": "2025-01-14T15:42:06.345722+00:00"} {"global_step": 1921, "acc_step": 0, "speed/wps": 12847.824591437779, "speed/FLOPS": 201792739969605.7, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19893582165241241, "optim/lr": 0.001152, "optim/total_tokens": 503578624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376441, "loss/out": 3.7663283348083496, "created_at": "2025-01-14T15:42:11.451657+00:00"} {"global_step": 1922, "acc_step": 0, "speed/wps": 12839.720862418004, "speed/FLOPS": 201665459769659.53, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.19760148227214813, "optim/lr": 0.0011526, "optim/total_tokens": 503840768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 505317, "loss/out": 3.910529136657715, "created_at": "2025-01-14T15:42:16.561183+00:00"} {"global_step": 1923, "acc_step": 0, "speed/wps": 12844.417331160143, "speed/FLOPS": 201739224264879.16, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18915702402591705, "optim/lr": 0.0011532, "optim/total_tokens": 504102912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.9388844966888428, "created_at": "2025-01-14T15:42:21.668172+00:00"} {"global_step": 1924, "acc_step": 0, "speed/wps": 12842.310340672522, "speed/FLOPS": 201706131083962.75, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18477846682071686, "optim/lr": 0.0011538, "optim/total_tokens": 504365056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 3.815946102142334, "created_at": "2025-01-14T15:42:26.777840+00:00"} {"global_step": 1925, "acc_step": 0, "speed/wps": 12846.319848121302, "speed/FLOPS": 201769105908086.44, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21611346304416656, "optim/lr": 0.0011543999999999999, "optim/total_tokens": 504627200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.8270978927612305, "created_at": "2025-01-14T15:42:31.883108+00:00"} {"global_step": 1926, "acc_step": 0, "speed/wps": 12846.453916882238, "speed/FLOPS": 201771211642206.7, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20282231271266937, "optim/lr": 0.001155, "optim/total_tokens": 504889344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291719, "loss/out": 3.8558712005615234, "created_at": "2025-01-14T15:42:36.996629+00:00"} {"global_step": 1927, "acc_step": 0, "speed/wps": 12846.830976382334, "speed/FLOPS": 201777133879789.78, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20022758841514587, "optim/lr": 0.0011556, "optim/total_tokens": 505151488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.8307738304138184, "created_at": "2025-01-14T15:42:42.103172+00:00"} {"global_step": 1928, "acc_step": 0, "speed/wps": 12843.330033101483, "speed/FLOPS": 201722146754763.8, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.18891793489456177, "optim/lr": 0.0011562, "optim/total_tokens": 505413632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 3.9113457202911377, "created_at": "2025-01-14T15:42:47.213109+00:00"} {"global_step": 1929, "acc_step": 0, "speed/wps": 12846.225368227662, "speed/FLOPS": 201767621971527.84, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21950902044773102, "optim/lr": 0.0011568, "optim/total_tokens": 505675776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.7951247692108154, "created_at": "2025-01-14T15:42:52.318544+00:00"} {"global_step": 1930, "acc_step": 0, "speed/wps": 12842.201548256404, "speed/FLOPS": 201704422349571.16, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.27450212836265564, "optim/lr": 0.0011574, "optim/total_tokens": 505937920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.942572593688965, "created_at": "2025-01-14T15:42:57.430412+00:00"} {"global_step": 1931, "acc_step": 0, "speed/wps": 12851.007816090663, "speed/FLOPS": 201842736886988.3, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.20504255592823029, "optim/lr": 0.001158, "optim/total_tokens": 506200064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328221, "loss/out": 3.777937412261963, "created_at": "2025-01-14T15:43:02.538522+00:00"} {"global_step": 1932, "acc_step": 0, "speed/wps": 12846.751501509863, "speed/FLOPS": 201775885617707.7, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2135191410779953, "optim/lr": 0.0011586, "optim/total_tokens": 506462208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 3.838256359100342, "created_at": "2025-01-14T15:43:07.646147+00:00"} {"global_step": 1933, "acc_step": 0, "speed/wps": 12844.016503287137, "speed/FLOPS": 201732928712338.38, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20237599313259125, "optim/lr": 0.0011592, "optim/total_tokens": 506724352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.7725558280944824, "created_at": "2025-01-14T15:43:12.754810+00:00"} {"global_step": 1934, "acc_step": 0, "speed/wps": 12842.297077550471, "speed/FLOPS": 201705922768405.47, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1914406269788742, "optim/lr": 0.0011598000000000001, "optim/total_tokens": 506986496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.7656455039978027, "created_at": "2025-01-14T15:43:17.865802+00:00"} {"global_step": 1935, "acc_step": 0, "speed/wps": 12844.098288700143, "speed/FLOPS": 201734213264634.56, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2213115394115448, "optim/lr": 0.0011604, "optim/total_tokens": 507248640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294941, "loss/out": 3.775547504425049, "created_at": "2025-01-14T15:43:22.972212+00:00"} {"global_step": 1936, "acc_step": 0, "speed/wps": 12842.89216611736, "speed/FLOPS": 201715269452085.88, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1885443776845932, "optim/lr": 0.0011610000000000001, "optim/total_tokens": 507510784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 3.7372255325317383, "created_at": "2025-01-14T15:43:28.080509+00:00"} {"global_step": 1937, "acc_step": 0, "speed/wps": 12846.51935870717, "speed/FLOPS": 201772239495993.75, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1749531477689743, "optim/lr": 0.0011616, "optim/total_tokens": 507772928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488482, "loss/out": 3.7185349464416504, "created_at": "2025-01-14T15:43:33.189030+00:00"} {"global_step": 1938, "acc_step": 0, "speed/wps": 12840.255863030463, "speed/FLOPS": 201673862689446.06, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19225041568279266, "optim/lr": 0.0011622000000000002, "optim/total_tokens": 508035072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.912968873977661, "created_at": "2025-01-14T15:43:38.296579+00:00"} {"global_step": 1939, "acc_step": 0, "speed/wps": 12842.336409970703, "speed/FLOPS": 201706540538113.62, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19484730064868927, "optim/lr": 0.0011628, "optim/total_tokens": 508297216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 3.8712682723999023, "created_at": "2025-01-14T15:43:43.410336+00:00"} {"global_step": 1940, "acc_step": 0, "speed/wps": 12847.019530850985, "speed/FLOPS": 201780095386820.9, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19197849929332733, "optim/lr": 0.0011634, "optim/total_tokens": 508559360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324713, "loss/out": 3.730457305908203, "created_at": "2025-01-14T15:43:48.515265+00:00"} {"global_step": 1941, "acc_step": 0, "speed/wps": 12842.492710792892, "speed/FLOPS": 201708995457305.97, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19812530279159546, "optim/lr": 0.001164, "optim/total_tokens": 508821504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.719602346420288, "created_at": "2025-01-14T15:43:53.624920+00:00"} {"global_step": 1942, "acc_step": 0, "speed/wps": 12843.04589504331, "speed/FLOPS": 201717683976113.56, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18169966340065002, "optim/lr": 0.0011646, "optim/total_tokens": 509083648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417484, "loss/out": 3.7088801860809326, "created_at": "2025-01-14T15:43:58.734318+00:00"} {"global_step": 1943, "acc_step": 0, "speed/wps": 12841.385419001434, "speed/FLOPS": 201691603918142.72, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.184669628739357, "optim/lr": 0.0011652000000000001, "optim/total_tokens": 509345792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 3.9261488914489746, "created_at": "2025-01-14T15:44:03.845888+00:00"} {"global_step": 1944, "acc_step": 0, "speed/wps": 12844.5003337965, "speed/FLOPS": 201740527935340.72, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.21750925481319427, "optim/lr": 0.0011658, "optim/total_tokens": 509607936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.7912158966064453, "created_at": "2025-01-14T15:44:08.955446+00:00"} {"global_step": 1945, "acc_step": 0, "speed/wps": 12840.830686001707, "speed/FLOPS": 201682891074099.0, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.22965329885482788, "optim/lr": 0.0011664, "optim/total_tokens": 509870080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.7036778926849365, "created_at": "2025-01-14T15:44:14.065421+00:00"} {"global_step": 1946, "acc_step": 0, "speed/wps": 12843.754073059723, "speed/FLOPS": 201728806884999.9, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22126510739326477, "optim/lr": 0.001167, "optim/total_tokens": 510132224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 3.857529401779175, "created_at": "2025-01-14T15:44:19.171819+00:00"} {"global_step": 1947, "acc_step": 0, "speed/wps": 12852.078157318796, "speed/FLOPS": 201859548066779.1, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23049777746200562, "optim/lr": 0.0011676, "optim/total_tokens": 510394368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 4.009036064147949, "created_at": "2025-01-14T15:44:24.278716+00:00"} {"global_step": 1948, "acc_step": 0, "speed/wps": 12842.355021587813, "speed/FLOPS": 201706832859135.38, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1865355670452118, "optim/lr": 0.0011682, "optim/total_tokens": 510656512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.6895854473114014, "created_at": "2025-01-14T15:44:29.387950+00:00"} {"global_step": 1949, "acc_step": 0, "speed/wps": 12853.94508339588, "speed/FLOPS": 201888870706244.97, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22621838748455048, "optim/lr": 0.0011688, "optim/total_tokens": 510918656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329565, "loss/out": 3.820291519165039, "created_at": "2025-01-14T15:44:34.491987+00:00"} {"global_step": 1950, "acc_step": 0, "speed/wps": 12845.056113121094, "speed/FLOPS": 201749257213354.28, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24225901067256927, "optim/lr": 0.0011694, "optim/total_tokens": 511180800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 3.7774658203125, "created_at": "2025-01-14T15:44:39.600853+00:00"} {"global_step": 1951, "acc_step": 0, "speed/wps": 12841.004894741014, "speed/FLOPS": 201685627261737.72, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1909259706735611, "optim/lr": 0.00117, "optim/total_tokens": 511442944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 393639, "loss/out": 3.792264938354492, "created_at": "2025-01-14T15:44:44.709598+00:00"} {"global_step": 1952, "acc_step": 0, "speed/wps": 12842.891577789287, "speed/FLOPS": 201715260211585.03, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18590368330478668, "optim/lr": 0.0011706, "optim/total_tokens": 511705088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.7586114406585693, "created_at": "2025-01-14T15:44:49.824373+00:00"} {"global_step": 1953, "acc_step": 0, "speed/wps": 12848.255234002532, "speed/FLOPS": 201799503802850.03, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18724863231182098, "optim/lr": 0.0011712, "optim/total_tokens": 511967232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 3.8194103240966797, "created_at": "2025-01-14T15:44:54.929009+00:00"} {"global_step": 1954, "acc_step": 0, "speed/wps": 12847.808363646775, "speed/FLOPS": 201792485089848.66, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16968216001987457, "optim/lr": 0.0011718, "optim/total_tokens": 512229376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322837, "loss/out": 3.837161064147949, "created_at": "2025-01-14T15:45:00.035338+00:00"} {"global_step": 1955, "acc_step": 0, "speed/wps": 12843.272790520292, "speed/FLOPS": 201721247681366.66, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20313698053359985, "optim/lr": 0.0011723999999999999, "optim/total_tokens": 512491520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 3.7673585414886475, "created_at": "2025-01-14T15:45:05.144430+00:00"} {"global_step": 1956, "acc_step": 0, "speed/wps": 12842.317479660675, "speed/FLOPS": 201706243211581.66, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16884729266166687, "optim/lr": 0.001173, "optim/total_tokens": 512753664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361998, "loss/out": 3.7184619903564453, "created_at": "2025-01-14T15:45:10.260372+00:00"} {"global_step": 1957, "acc_step": 0, "speed/wps": 12850.10458829872, "speed/FLOPS": 201828550453350.78, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17970198392868042, "optim/lr": 0.0011736, "optim/total_tokens": 513015808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.8507676124572754, "created_at": "2025-01-14T15:45:15.366939+00:00"} {"global_step": 1958, "acc_step": 0, "speed/wps": 12846.339878403147, "speed/FLOPS": 201769420511187.28, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1925962269306183, "optim/lr": 0.0011742, "optim/total_tokens": 513277952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 3.8841376304626465, "created_at": "2025-01-14T15:45:20.474588+00:00"} {"global_step": 1959, "acc_step": 0, "speed/wps": 12848.944623871655, "speed/FLOPS": 201810331618065.56, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22201064229011536, "optim/lr": 0.0011748, "optim/total_tokens": 513540096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 3.9458632469177246, "created_at": "2025-01-14T15:45:25.581929+00:00"} {"global_step": 1960, "acc_step": 0, "speed/wps": 12849.437082457947, "speed/FLOPS": 201818066356872.3, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1829395443201065, "optim/lr": 0.0011754, "optim/total_tokens": 513802240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481847, "loss/out": 3.7286267280578613, "created_at": "2025-01-14T15:45:30.689861+00:00"} {"global_step": 1961, "acc_step": 0, "speed/wps": 12849.635655119175, "speed/FLOPS": 201821185213384.3, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23375041782855988, "optim/lr": 0.001176, "optim/total_tokens": 514064384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 506365, "loss/out": 3.8276097774505615, "created_at": "2025-01-14T15:45:35.797536+00:00"} {"global_step": 1962, "acc_step": 0, "speed/wps": 12848.588462908676, "speed/FLOPS": 201804737620725.53, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18262603878974915, "optim/lr": 0.0011766, "optim/total_tokens": 514326528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500558, "loss/out": 3.8204219341278076, "created_at": "2025-01-14T15:45:40.904942+00:00"} {"global_step": 1963, "acc_step": 0, "speed/wps": 12847.402278753902, "speed/FLOPS": 201786106968586.97, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25486868619918823, "optim/lr": 0.0011772000000000002, "optim/total_tokens": 514588672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 3.8447937965393066, "created_at": "2025-01-14T15:45:46.012223+00:00"} {"global_step": 1964, "acc_step": 0, "speed/wps": 12846.582737348617, "speed/FLOPS": 201773234944649.06, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2335716187953949, "optim/lr": 0.0011778000000000001, "optim/total_tokens": 514850816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.921377182006836, "created_at": "2025-01-14T15:45:51.117953+00:00"} {"global_step": 1965, "acc_step": 0, "speed/wps": 12847.164992210472, "speed/FLOPS": 201782380057356.5, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18871115148067474, "optim/lr": 0.0011784, "optim/total_tokens": 515112960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291719, "loss/out": 3.9009056091308594, "created_at": "2025-01-14T15:45:56.225698+00:00"} {"global_step": 1966, "acc_step": 0, "speed/wps": 12848.555752485761, "speed/FLOPS": 201804223858584.8, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2685626149177551, "optim/lr": 0.0011790000000000001, "optim/total_tokens": 515375104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 3.887714385986328, "created_at": "2025-01-14T15:46:01.334718+00:00"} {"global_step": 1967, "acc_step": 0, "speed/wps": 12846.444378810687, "speed/FLOPS": 201771061833686.3, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2306850254535675, "optim/lr": 0.0011796, "optim/total_tokens": 515637248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.7245988845825195, "created_at": "2025-01-14T15:46:06.446105+00:00"} {"global_step": 1968, "acc_step": 0, "speed/wps": 12832.39458734467, "speed/FLOPS": 201550390552275.88, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20439493656158447, "optim/lr": 0.0011802000000000002, "optim/total_tokens": 515899392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.781733751296997, "created_at": "2025-01-14T15:46:11.560041+00:00"} {"global_step": 1969, "acc_step": 0, "speed/wps": 12850.365536881038, "speed/FLOPS": 201832649009417.0, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19753791391849518, "optim/lr": 0.0011808, "optim/total_tokens": 516161536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 3.776503562927246, "created_at": "2025-01-14T15:46:16.666437+00:00"} {"global_step": 1970, "acc_step": 0, "speed/wps": 12842.362458281033, "speed/FLOPS": 201706949662621.5, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2138899862766266, "optim/lr": 0.0011814, "optim/total_tokens": 516423680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325381, "loss/out": 3.797623872756958, "created_at": "2025-01-14T15:46:21.774724+00:00"} {"global_step": 1971, "acc_step": 0, "speed/wps": 12851.1655367434, "speed/FLOPS": 201845214106570.4, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1879298985004425, "optim/lr": 0.001182, "optim/total_tokens": 516685824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472260, "loss/out": 3.809854745864868, "created_at": "2025-01-14T15:46:26.878736+00:00"} {"global_step": 1972, "acc_step": 0, "speed/wps": 12850.928517177726, "speed/FLOPS": 201841491388592.5, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17225335538387299, "optim/lr": 0.0011826, "optim/total_tokens": 516947968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319663, "loss/out": 3.6618971824645996, "created_at": "2025-01-14T15:46:31.982070+00:00"} {"global_step": 1973, "acc_step": 0, "speed/wps": 12847.412446865193, "speed/FLOPS": 201786266672747.25, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19889289140701294, "optim/lr": 0.0011832, "optim/total_tokens": 517210112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502937, "loss/out": 3.7714619636535645, "created_at": "2025-01-14T15:46:37.087431+00:00"} {"global_step": 1974, "acc_step": 0, "speed/wps": 12841.55036508005, "speed/FLOPS": 201694194622967.03, "speed/curr_iter_time": 1.2875, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.18925005197525024, "optim/lr": 0.0011838, "optim/total_tokens": 517472256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300225, "loss/out": 3.7654757499694824, "created_at": "2025-01-14T15:46:42.194425+00:00"} {"global_step": 1975, "acc_step": 0, "speed/wps": 12847.126134486272, "speed/FLOPS": 201781769743401.84, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23165443539619446, "optim/lr": 0.0011844, "optim/total_tokens": 517734400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462391, "loss/out": 3.904632329940796, "created_at": "2025-01-14T15:46:47.301702+00:00"} {"global_step": 1976, "acc_step": 0, "speed/wps": 12849.357702037838, "speed/FLOPS": 201816819578294.34, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.43555372953414917, "optim/lr": 0.001185, "optim/total_tokens": 517996544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362914, "loss/out": 3.822035074234009, "created_at": "2025-01-14T15:46:52.408479+00:00"} {"global_step": 1977, "acc_step": 0, "speed/wps": 12841.442342460188, "speed/FLOPS": 201692497979283.5, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.35682418942451477, "optim/lr": 0.0011856, "optim/total_tokens": 518258688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.8669071197509766, "created_at": "2025-01-14T15:46:57.519079+00:00"} {"global_step": 1978, "acc_step": 0, "speed/wps": 12846.102481015025, "speed/FLOPS": 201765691858988.62, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.34604257345199585, "optim/lr": 0.0011861999999999999, "optim/total_tokens": 518520832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 3.7893753051757812, "created_at": "2025-01-14T15:47:02.627435+00:00"} {"global_step": 1979, "acc_step": 0, "speed/wps": 12846.800648414812, "speed/FLOPS": 201776657537384.84, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.33537808060646057, "optim/lr": 0.0011868, "optim/total_tokens": 518782976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 3.8473973274230957, "created_at": "2025-01-14T15:47:07.737759+00:00"} {"global_step": 1980, "acc_step": 0, "speed/wps": 12852.370270226987, "speed/FLOPS": 201864136101407.44, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3055933713912964, "optim/lr": 0.0011874, "optim/total_tokens": 519045120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481190, "loss/out": 3.850795269012451, "created_at": "2025-01-14T15:47:12.847440+00:00"} {"global_step": 1981, "acc_step": 0, "speed/wps": 12851.349770858546, "speed/FLOPS": 201848107756512.97, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.001, "optim/grad_norm": 0.37516745924949646, "optim/lr": 0.001188, "optim/total_tokens": 519307264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361859, "loss/out": 3.8432438373565674, "created_at": "2025-01-14T15:47:17.951472+00:00"} {"global_step": 1982, "acc_step": 0, "speed/wps": 12849.573206018778, "speed/FLOPS": 201820204364448.53, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.28977981209754944, "optim/lr": 0.0011886, "optim/total_tokens": 519569408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.8860013484954834, "created_at": "2025-01-14T15:47:23.069980+00:00"} {"global_step": 1983, "acc_step": 0, "speed/wps": 12848.350855727205, "speed/FLOPS": 201801005673434.06, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23962201178073883, "optim/lr": 0.0011891999999999998, "optim/total_tokens": 519831552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.8549227714538574, "created_at": "2025-01-14T15:47:28.175103+00:00"} {"global_step": 1984, "acc_step": 0, "speed/wps": 12851.535608626069, "speed/FLOPS": 201851026594019.78, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23751704394817352, "optim/lr": 0.0011898, "optim/total_tokens": 520093696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463750, "loss/out": 3.820505142211914, "created_at": "2025-01-14T15:47:33.281249+00:00"} {"global_step": 1985, "acc_step": 0, "speed/wps": 12848.931398657842, "speed/FLOPS": 201810123897909.28, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.22218751907348633, "optim/lr": 0.0011903999999999999, "optim/total_tokens": 520355840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.810454845428467, "created_at": "2025-01-14T15:47:38.385650+00:00"} {"global_step": 1986, "acc_step": 0, "speed/wps": 12845.105509666415, "speed/FLOPS": 201750033053976.53, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19959993660449982, "optim/lr": 0.001191, "optim/total_tokens": 520617984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 3.7504382133483887, "created_at": "2025-01-14T15:47:43.494748+00:00"} {"global_step": 1987, "acc_step": 0, "speed/wps": 12844.012569473942, "speed/FLOPS": 201732866926396.53, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20769675076007843, "optim/lr": 0.0011916000000000001, "optim/total_tokens": 520880128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 3.748715400695801, "created_at": "2025-01-14T15:47:48.601340+00:00"} {"global_step": 1988, "acc_step": 0, "speed/wps": 12852.194769441554, "speed/FLOPS": 201861379620408.88, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20478880405426025, "optim/lr": 0.0011922, "optim/total_tokens": 521142272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 3.842836856842041, "created_at": "2025-01-14T15:47:53.709117+00:00"} {"global_step": 1989, "acc_step": 0, "speed/wps": 12842.064266056028, "speed/FLOPS": 201702266143968.72, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.001, "optim/grad_norm": 0.23065467178821564, "optim/lr": 0.0011928000000000002, "optim/total_tokens": 521404416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 3.9300408363342285, "created_at": "2025-01-14T15:47:58.818609+00:00"} {"global_step": 1990, "acc_step": 0, "speed/wps": 12848.402942613218, "speed/FLOPS": 201801823769553.78, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.20188115537166595, "optim/lr": 0.0011934, "optim/total_tokens": 521666560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306907, "loss/out": 3.83174991607666, "created_at": "2025-01-14T15:48:03.924940+00:00"} {"global_step": 1991, "acc_step": 0, "speed/wps": 12848.880543893789, "speed/FLOPS": 201809325153958.06, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.24679872393608093, "optim/lr": 0.0011940000000000002, "optim/total_tokens": 521928704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.7300686836242676, "created_at": "2025-01-14T15:48:09.031438+00:00"} {"global_step": 1992, "acc_step": 0, "speed/wps": 12851.397265428843, "speed/FLOPS": 201848853724003.88, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22317616641521454, "optim/lr": 0.0011946, "optim/total_tokens": 522190848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465084, "loss/out": 3.9032132625579834, "created_at": "2025-01-14T15:48:14.134585+00:00"} {"global_step": 1993, "acc_step": 0, "speed/wps": 12846.83130473479, "speed/FLOPS": 201777139037016.28, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21337968111038208, "optim/lr": 0.0011952, "optim/total_tokens": 522452992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.8885440826416016, "created_at": "2025-01-14T15:48:19.240773+00:00"} {"global_step": 1994, "acc_step": 0, "speed/wps": 12850.596807261132, "speed/FLOPS": 201836281428535.2, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21528318524360657, "optim/lr": 0.0011958000000000001, "optim/total_tokens": 522715136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.9300529956817627, "created_at": "2025-01-14T15:48:24.345144+00:00"} {"global_step": 1995, "acc_step": 0, "speed/wps": 12845.033542822466, "speed/FLOPS": 201748902715799.5, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.30473220348358154, "optim/lr": 0.0011964, "optim/total_tokens": 522977280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310665, "loss/out": 3.736482620239258, "created_at": "2025-01-14T15:48:29.463064+00:00"} {"global_step": 1996, "acc_step": 0, "speed/wps": 12847.05629109152, "speed/FLOPS": 201780672756912.38, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24113014340400696, "optim/lr": 0.0011970000000000001, "optim/total_tokens": 523239424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 3.772977352142334, "created_at": "2025-01-14T15:48:34.569114+00:00"} {"global_step": 1997, "acc_step": 0, "speed/wps": 12846.594897657536, "speed/FLOPS": 201773425939010.66, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2087160050868988, "optim/lr": 0.0011976, "optim/total_tokens": 523501568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303511, "loss/out": 3.848592758178711, "created_at": "2025-01-14T15:48:39.678289+00:00"} {"global_step": 1998, "acc_step": 0, "speed/wps": 12845.668275378779, "speed/FLOPS": 201758872062810.3, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1859544813632965, "optim/lr": 0.0011982, "optim/total_tokens": 523763712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477479, "loss/out": 3.883016586303711, "created_at": "2025-01-14T15:48:44.786994+00:00"} {"global_step": 1999, "acc_step": 0, "speed/wps": 12846.199709954542, "speed/FLOPS": 201767218973090.3, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19692164659500122, "optim/lr": 0.0011988, "optim/total_tokens": 524025856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304748, "loss/out": 3.8127598762512207, "created_at": "2025-01-14T15:48:49.895144+00:00"} {"global_step": 2000, "acc_step": 0, "speed/wps": 12854.776117600772, "speed/FLOPS": 201901923240393.47, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17880135774612427, "optim/lr": 0.0011994, "optim/total_tokens": 524288000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470119, "loss/out": 3.720947265625, "created_at": "2025-01-14T15:48:55.002611+00:00"} {"global_step": 2001, "acc_step": 0, "speed/wps": 12210.643855616352, "speed/FLOPS": 191784940935445.16, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.2708, "optim/grad_norm": 0.17522595822811127, "optim/lr": 0.0012000000000000001, "optim/total_tokens": 524550144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504796, "loss/out": 3.7643165588378906, "created_at": "2025-01-14T15:49:00.376516+00:00"} {"global_step": 2002, "acc_step": 0, "speed/wps": 12867.740798466895, "speed/FLOPS": 202105551368734.34, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16370199620723724, "optim/lr": 0.0012006, "optim/total_tokens": 524812288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 3.764566659927368, "created_at": "2025-01-14T15:49:05.474377+00:00"} {"global_step": 2003, "acc_step": 0, "speed/wps": 12860.872338661486, "speed/FLOPS": 201997672769237.56, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17710267007350922, "optim/lr": 0.0012012, "optim/total_tokens": 525074432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.7005553245544434, "created_at": "2025-01-14T15:49:10.573748+00:00"} {"global_step": 2004, "acc_step": 0, "speed/wps": 12860.498095202422, "speed/FLOPS": 201991794761448.56, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16817635297775269, "optim/lr": 0.0012018, "optim/total_tokens": 525336576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 3.878217935562134, "created_at": "2025-01-14T15:49:15.673171+00:00"} {"global_step": 2005, "acc_step": 0, "speed/wps": 12860.043928571196, "speed/FLOPS": 201984661450414.8, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16867652535438538, "optim/lr": 0.0012024, "optim/total_tokens": 525598720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.6612086296081543, "created_at": "2025-01-14T15:49:20.778874+00:00"} {"global_step": 2006, "acc_step": 0, "speed/wps": 12861.93456522886, "speed/FLOPS": 202014356497130.44, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1814562976360321, "optim/lr": 0.001203, "optim/total_tokens": 525860864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.744321823120117, "created_at": "2025-01-14T15:49:25.877766+00:00"} {"global_step": 2007, "acc_step": 0, "speed/wps": 12856.718853828706, "speed/FLOPS": 201932436582452.5, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17606814205646515, "optim/lr": 0.0012036, "optim/total_tokens": 526123008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474645, "loss/out": 3.8806073665618896, "created_at": "2025-01-14T15:49:30.978812+00:00"} {"global_step": 2008, "acc_step": 0, "speed/wps": 12866.998467223131, "speed/FLOPS": 202093892036481.03, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.18495826423168182, "optim/lr": 0.0012041999999999999, "optim/total_tokens": 526385152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.8440380096435547, "created_at": "2025-01-14T15:49:36.078543+00:00"} {"global_step": 2009, "acc_step": 0, "speed/wps": 12865.360460037564, "speed/FLOPS": 202068164882773.94, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19246281683444977, "optim/lr": 0.0012048, "optim/total_tokens": 526647296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 3.838087320327759, "created_at": "2025-01-14T15:49:41.179679+00:00"} {"global_step": 2010, "acc_step": 0, "speed/wps": 12867.050580716603, "speed/FLOPS": 202094710550507.44, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21683187782764435, "optim/lr": 0.0012054, "optim/total_tokens": 526909440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 3.739234685897827, "created_at": "2025-01-14T15:49:46.279074+00:00"} {"global_step": 2011, "acc_step": 0, "speed/wps": 12863.664763396557, "speed/FLOPS": 202041531636895.56, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3011719882488251, "optim/lr": 0.001206, "optim/total_tokens": 527171584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 3.8194875717163086, "created_at": "2025-01-14T15:49:51.379088+00:00"} {"global_step": 2012, "acc_step": 0, "speed/wps": 12855.105290993051, "speed/FLOPS": 201907093360850.84, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2625911235809326, "optim/lr": 0.0012066, "optim/total_tokens": 527433728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435044, "loss/out": 3.775740146636963, "created_at": "2025-01-14T15:49:56.481264+00:00"} {"global_step": 2013, "acc_step": 0, "speed/wps": 12863.54080864619, "speed/FLOPS": 202039584757209.38, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23915037512779236, "optim/lr": 0.0012072, "optim/total_tokens": 527695872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 3.7127771377563477, "created_at": "2025-01-14T15:50:01.581887+00:00"} {"global_step": 2014, "acc_step": 0, "speed/wps": 12859.719331630306, "speed/FLOPS": 201979563209415.2, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2351548820734024, "optim/lr": 0.0012078, "optim/total_tokens": 527958016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377340, "loss/out": 3.897399425506592, "created_at": "2025-01-14T15:50:06.682350+00:00"} {"global_step": 2015, "acc_step": 0, "speed/wps": 12861.490534558585, "speed/FLOPS": 202007382385291.7, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1856161504983902, "optim/lr": 0.0012084, "optim/total_tokens": 528220160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462963, "loss/out": 3.75290584564209, "created_at": "2025-01-14T15:50:11.781450+00:00"} {"global_step": 2016, "acc_step": 0, "speed/wps": 12849.78496290211, "speed/FLOPS": 201823530297285.7, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1989014446735382, "optim/lr": 0.001209, "optim/total_tokens": 528482304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428385, "loss/out": 3.8602285385131836, "created_at": "2025-01-14T15:50:16.885429+00:00"} {"global_step": 2017, "acc_step": 0, "speed/wps": 12867.47942609646, "speed/FLOPS": 202101446156493.4, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16448678076267242, "optim/lr": 0.0012096000000000001, "optim/total_tokens": 528744448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 3.8083183765411377, "created_at": "2025-01-14T15:50:21.982321+00:00"} {"global_step": 2018, "acc_step": 0, "speed/wps": 12870.101689743575, "speed/FLOPS": 202142632410439.2, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1840585172176361, "optim/lr": 0.0012102, "optim/total_tokens": 529006592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 373603, "loss/out": 3.814124584197998, "created_at": "2025-01-14T15:50:27.082264+00:00"} {"global_step": 2019, "acc_step": 0, "speed/wps": 12856.956065894085, "speed/FLOPS": 201936162323902.8, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16404566168785095, "optim/lr": 0.0012108000000000002, "optim/total_tokens": 529268736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 3.8729164600372314, "created_at": "2025-01-14T15:50:32.184827+00:00"} {"global_step": 2020, "acc_step": 0, "speed/wps": 12860.340222953633, "speed/FLOPS": 201989315160842.12, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16665561497211456, "optim/lr": 0.0012114, "optim/total_tokens": 529530880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474937, "loss/out": 3.791255474090576, "created_at": "2025-01-14T15:50:37.287928+00:00"} {"global_step": 2021, "acc_step": 0, "speed/wps": 12859.787059025644, "speed/FLOPS": 201980626961226.2, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.18399465084075928, "optim/lr": 0.0012120000000000002, "optim/total_tokens": 529793024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504796, "loss/out": 3.8008172512054443, "created_at": "2025-01-14T15:50:42.387762+00:00"} {"global_step": 2022, "acc_step": 0, "speed/wps": 12861.164148641465, "speed/FLOPS": 202002256045962.06, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18803627789020538, "optim/lr": 0.0012126, "optim/total_tokens": 530055168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 3.7967638969421387, "created_at": "2025-01-14T15:50:47.491952+00:00"} {"global_step": 2023, "acc_step": 0, "speed/wps": 12862.863634567677, "speed/FLOPS": 202028948807771.62, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.38180315494537354, "optim/lr": 0.0012132, "optim/total_tokens": 530317312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498635, "loss/out": 3.801234722137451, "created_at": "2025-01-14T15:50:52.591396+00:00"} {"global_step": 2024, "acc_step": 0, "speed/wps": 12874.075160102813, "speed/FLOPS": 202205041222551.5, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24413250386714935, "optim/lr": 0.0012138000000000001, "optim/total_tokens": 530579456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433167, "loss/out": 3.773433208465576, "created_at": "2025-01-14T15:50:57.685452+00:00"} {"global_step": 2025, "acc_step": 0, "speed/wps": 12867.752495318107, "speed/FLOPS": 202105735083855.8, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.34490105509757996, "optim/lr": 0.0012144, "optim/total_tokens": 530841600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 3.823971748352051, "created_at": "2025-01-14T15:51:02.788873+00:00"} {"global_step": 2026, "acc_step": 0, "speed/wps": 12854.06501427455, "speed/FLOPS": 201890754385496.78, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.30223312973976135, "optim/lr": 0.0012150000000000002, "optim/total_tokens": 531103744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 271907, "loss/out": 3.853055238723755, "created_at": "2025-01-14T15:51:07.891269+00:00"} {"global_step": 2027, "acc_step": 0, "speed/wps": 12858.875241966614, "speed/FLOPS": 201966305621346.2, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3277544379234314, "optim/lr": 0.0012156, "optim/total_tokens": 531365888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 3.8854928016662598, "created_at": "2025-01-14T15:51:12.991476+00:00"} {"global_step": 2028, "acc_step": 0, "speed/wps": 12858.691548628545, "speed/FLOPS": 201963420465050.7, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2133903056383133, "optim/lr": 0.0012162, "optim/total_tokens": 531628032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293489, "loss/out": 3.837979793548584, "created_at": "2025-01-14T15:51:18.096597+00:00"} {"global_step": 2029, "acc_step": 0, "speed/wps": 12850.791958900152, "speed/FLOPS": 201839346553192.47, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2153678685426712, "optim/lr": 0.0012168, "optim/total_tokens": 531890176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457652, "loss/out": 3.805476188659668, "created_at": "2025-01-14T15:51:23.202404+00:00"} {"global_step": 2030, "acc_step": 0, "speed/wps": 12861.591687278722, "speed/FLOPS": 202008971127759.84, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.185044527053833, "optim/lr": 0.0012174, "optim/total_tokens": 532152320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 383767, "loss/out": 3.798647403717041, "created_at": "2025-01-14T15:51:28.303436+00:00"} {"global_step": 2031, "acc_step": 0, "speed/wps": 12862.235046885915, "speed/FLOPS": 202019075974459.56, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.201181560754776, "optim/lr": 0.0012180000000000001, "optim/total_tokens": 532414464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 3.6683998107910156, "created_at": "2025-01-14T15:51:33.402643+00:00"} {"global_step": 2032, "acc_step": 0, "speed/wps": 12861.538633031028, "speed/FLOPS": 202008137837895.66, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1881561577320099, "optim/lr": 0.0012186, "optim/total_tokens": 532676608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423603, "loss/out": 3.744102954864502, "created_at": "2025-01-14T15:51:38.508222+00:00"} {"global_step": 2033, "acc_step": 0, "speed/wps": 12853.978424697516, "speed/FLOPS": 201889394377202.9, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17947334051132202, "optim/lr": 0.0012192, "optim/total_tokens": 532938752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.8159587383270264, "created_at": "2025-01-14T15:51:43.612348+00:00"} {"global_step": 2034, "acc_step": 0, "speed/wps": 12856.350893299035, "speed/FLOPS": 201926657256703.28, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.001, "optim/grad_norm": 0.16622881591320038, "optim/lr": 0.0012198, "optim/total_tokens": 533200896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503532, "loss/out": 3.8448567390441895, "created_at": "2025-01-14T15:51:48.714621+00:00"} {"global_step": 2035, "acc_step": 0, "speed/wps": 12854.719559609857, "speed/FLOPS": 201901034919427.34, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22129051387310028, "optim/lr": 0.0012204, "optim/total_tokens": 533463040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 3.833979845046997, "created_at": "2025-01-14T15:51:53.816874+00:00"} {"global_step": 2036, "acc_step": 0, "speed/wps": 12839.508380544392, "speed/FLOPS": 201662122449853.75, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21802090108394623, "optim/lr": 0.0012209999999999999, "optim/total_tokens": 533725184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 4.022181987762451, "created_at": "2025-01-14T15:51:58.924822+00:00"} {"global_step": 2037, "acc_step": 0, "speed/wps": 12831.908214657997, "speed/FLOPS": 201542751400885.56, "speed/curr_iter_time": 1.2876, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19466353952884674, "optim/lr": 0.0012216, "optim/total_tokens": 533987328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313228, "loss/out": 3.8431947231292725, "created_at": "2025-01-14T15:52:04.036877+00:00"} {"global_step": 2038, "acc_step": 0, "speed/wps": 12838.968679903957, "speed/FLOPS": 201653645709669.3, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20483490824699402, "optim/lr": 0.0012222, "optim/total_tokens": 534249472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.7844040393829346, "created_at": "2025-01-14T15:52:09.145172+00:00"} {"global_step": 2039, "acc_step": 0, "speed/wps": 12832.891186506999, "speed/FLOPS": 201558190324519.0, "speed/curr_iter_time": 1.2888, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20757076144218445, "optim/lr": 0.0012228, "optim/total_tokens": 534511616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303402, "loss/out": 3.7873036861419678, "created_at": "2025-01-14T15:52:14.259167+00:00"} {"global_step": 2040, "acc_step": 0, "speed/wps": 12833.98374212846, "speed/FLOPS": 201575350411882.78, "speed/curr_iter_time": 1.2878, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23766539990901947, "optim/lr": 0.0012234, "optim/total_tokens": 534773760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 3.8564305305480957, "created_at": "2025-01-14T15:52:19.370273+00:00"} {"global_step": 2041, "acc_step": 0, "speed/wps": 12835.028831198308, "speed/FLOPS": 201591764971827.62, "speed/curr_iter_time": 1.2884, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1793859302997589, "optim/lr": 0.001224, "optim/total_tokens": 535035904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.7808454036712646, "created_at": "2025-01-14T15:52:24.480938+00:00"} {"global_step": 2042, "acc_step": 0, "speed/wps": 12832.77928195285, "speed/FLOPS": 201556432709723.12, "speed/curr_iter_time": 1.2878, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20449785888195038, "optim/lr": 0.0012246, "optim/total_tokens": 535298048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 3.9203689098358154, "created_at": "2025-01-14T15:52:29.595420+00:00"} {"global_step": 2043, "acc_step": 0, "speed/wps": 12832.000069230213, "speed/FLOPS": 201544194103163.62, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1732541173696518, "optim/lr": 0.0012252, "optim/total_tokens": 535560192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 3.8322243690490723, "created_at": "2025-01-14T15:52:34.708674+00:00"} {"global_step": 2044, "acc_step": 0, "speed/wps": 12824.643148808796, "speed/FLOPS": 201428643558477.22, "speed/curr_iter_time": 1.2876, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1811017394065857, "optim/lr": 0.0012258, "optim/total_tokens": 535822336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 3.971242904663086, "created_at": "2025-01-14T15:52:39.829705+00:00"} {"global_step": 2045, "acc_step": 0, "speed/wps": 12833.763241586475, "speed/FLOPS": 201571887147871.44, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17478828132152557, "optim/lr": 0.0012264, "optim/total_tokens": 536084480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.62260103225708, "created_at": "2025-01-14T15:52:44.939939+00:00"} {"global_step": 2046, "acc_step": 0, "speed/wps": 12839.716781297895, "speed/FLOPS": 201665395670060.16, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1782514601945877, "optim/lr": 0.001227, "optim/total_tokens": 536346624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.6764750480651855, "created_at": "2025-01-14T15:52:50.052978+00:00"} {"global_step": 2047, "acc_step": 0, "speed/wps": 12837.837356839036, "speed/FLOPS": 201635876726333.94, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0050413608551025, "optim/lr": 0.0012276000000000001, "optim/total_tokens": 536608768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 395284, "loss/out": 3.903721809387207, "created_at": "2025-01-14T15:52:55.162335+00:00"} {"global_step": 2048, "acc_step": 0, "speed/wps": 12840.080742135578, "speed/FLOPS": 201671112175150.6, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4044308364391327, "optim/lr": 0.0012282, "optim/total_tokens": 536870912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504796, "loss/out": 3.762418508529663, "created_at": "2025-01-14T15:53:00.270488+00:00"} {"global_step": 2049, "acc_step": 0, "speed/wps": 12833.021878856156, "speed/FLOPS": 201560243027452.62, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19147177040576935, "optim/lr": 0.0012288000000000002, "optim/total_tokens": 537133056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.8158626556396484, "created_at": "2025-01-14T15:53:05.380974+00:00"} {"global_step": 2050, "acc_step": 0, "speed/wps": 12836.573023408488, "speed/FLOPS": 201616018632430.97, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.27296456694602966, "optim/lr": 0.0012294, "optim/total_tokens": 537395200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 3.711118698120117, "created_at": "2025-01-14T15:53:10.492344+00:00"} {"global_step": 2051, "acc_step": 0, "speed/wps": 12833.569741955178, "speed/FLOPS": 201568847970265.56, "speed/curr_iter_time": 1.2881, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8177413940429688, "optim/lr": 0.00123, "optim/total_tokens": 537657344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 3.7845215797424316, "created_at": "2025-01-14T15:53:15.602998+00:00"} {"global_step": 2052, "acc_step": 0, "speed/wps": 12832.007997028217, "speed/FLOPS": 201544318620125.22, "speed/curr_iter_time": 1.2881, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3345988094806671, "optim/lr": 0.0012306, "optim/total_tokens": 537919488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472504, "loss/out": 3.9653053283691406, "created_at": "2025-01-14T15:53:20.713879+00:00"} {"global_step": 2053, "acc_step": 0, "speed/wps": 12839.220000444417, "speed/FLOPS": 201657593044107.84, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.41415977478027344, "optim/lr": 0.0012312, "optim/total_tokens": 538181632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 3.8009021282196045, "created_at": "2025-01-14T15:53:25.825447+00:00"} {"global_step": 2054, "acc_step": 0, "speed/wps": 12841.65801690299, "speed/FLOPS": 201695885442775.44, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0073797702789307, "optim/lr": 0.0012318000000000001, "optim/total_tokens": 538443776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355325, "loss/out": 3.8448503017425537, "created_at": "2025-01-14T15:53:30.932625+00:00"} {"global_step": 2055, "acc_step": 0, "speed/wps": 12834.603054345796, "speed/FLOPS": 201585077561280.25, "speed/curr_iter_time": 1.2888, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21910440921783447, "optim/lr": 0.0012324, "optim/total_tokens": 538705920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482522, "loss/out": 3.78598690032959, "created_at": "2025-01-14T15:53:36.046848+00:00"} {"global_step": 2056, "acc_step": 0, "speed/wps": 12831.056345167739, "speed/FLOPS": 201529371619949.72, "speed/curr_iter_time": 1.2889, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0030816793441772, "optim/lr": 0.001233, "optim/total_tokens": 538968064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 3.713758707046509, "created_at": "2025-01-14T15:53:41.158246+00:00"} {"global_step": 2057, "acc_step": 0, "speed/wps": 12837.201784371588, "speed/FLOPS": 201625894187364.56, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22155995666980743, "optim/lr": 0.0012336, "optim/total_tokens": 539230208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326583, "loss/out": 3.9075565338134766, "created_at": "2025-01-14T15:53:46.266991+00:00"} {"global_step": 2058, "acc_step": 0, "speed/wps": 12835.32072678809, "speed/FLOPS": 201596349593171.75, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6527043581008911, "optim/lr": 0.0012342, "optim/total_tokens": 539492352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474123, "loss/out": 3.8785171508789062, "created_at": "2025-01-14T15:53:51.376561+00:00"} {"global_step": 2059, "acc_step": 0, "speed/wps": 12837.829796570733, "speed/FLOPS": 201635757981931.34, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0178030729293823, "optim/lr": 0.0012348, "optim/total_tokens": 539754496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433690, "loss/out": 4.105562210083008, "created_at": "2025-01-14T15:53:56.485131+00:00"} {"global_step": 2060, "acc_step": 0, "speed/wps": 12839.559215029487, "speed/FLOPS": 201662920875296.06, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1887702941894531, "optim/lr": 0.0012354, "optim/total_tokens": 540016640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.951158046722412, "created_at": "2025-01-14T15:54:01.595096+00:00"} {"global_step": 2061, "acc_step": 0, "speed/wps": 12836.286386120097, "speed/FLOPS": 201611516599935.88, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25896328687667847, "optim/lr": 0.001236, "optim/total_tokens": 540278784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.7938222885131836, "created_at": "2025-01-14T15:54:06.709249+00:00"} {"global_step": 2062, "acc_step": 0, "speed/wps": 12842.874835073086, "speed/FLOPS": 201714997244220.2, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.30364280939102173, "optim/lr": 0.0012366, "optim/total_tokens": 540540928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292963, "loss/out": 3.8644795417785645, "created_at": "2025-01-14T15:54:11.821165+00:00"} {"global_step": 2063, "acc_step": 0, "speed/wps": 12832.880540547747, "speed/FLOPS": 201558023115099.88, "speed/curr_iter_time": 1.2875, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.3178711235523224, "optim/lr": 0.0012372, "optim/total_tokens": 540803072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.863590955734253, "created_at": "2025-01-14T15:54:16.932061+00:00"} {"global_step": 2064, "acc_step": 0, "speed/wps": 12841.313725828875, "speed/FLOPS": 201690477878350.94, "speed/curr_iter_time": 1.2877, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23649518191814423, "optim/lr": 0.0012378, "optim/total_tokens": 541065216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 378681, "loss/out": 3.8583321571350098, "created_at": "2025-01-14T15:54:22.041811+00:00"} {"global_step": 2065, "acc_step": 0, "speed/wps": 12839.620353290558, "speed/FLOPS": 201663881135698.97, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22865861654281616, "optim/lr": 0.0012384, "optim/total_tokens": 541327360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464271, "loss/out": 3.8291876316070557, "created_at": "2025-01-14T15:54:27.149663+00:00"} {"global_step": 2066, "acc_step": 0, "speed/wps": 12833.145445064998, "speed/FLOPS": 201562183804560.25, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2783831059932709, "optim/lr": 0.0012389999999999999, "optim/total_tokens": 541589504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 3.881446599960327, "created_at": "2025-01-14T15:54:32.264573+00:00"} {"global_step": 2067, "acc_step": 0, "speed/wps": 12835.488467057363, "speed/FLOPS": 201598984184599.97, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20579345524311066, "optim/lr": 0.0012396, "optim/total_tokens": 541851648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 3.7439942359924316, "created_at": "2025-01-14T15:54:37.376272+00:00"} {"global_step": 2068, "acc_step": 0, "speed/wps": 12834.378885508204, "speed/FLOPS": 201581556681645.6, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2387319952249527, "optim/lr": 0.0012402, "optim/total_tokens": 542113792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 3.902967929840088, "created_at": "2025-01-14T15:54:42.489571+00:00"} {"global_step": 2069, "acc_step": 0, "speed/wps": 12841.853675203069, "speed/FLOPS": 201698958525240.84, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2260216772556305, "optim/lr": 0.0012408, "optim/total_tokens": 542375936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348659, "loss/out": 3.642641305923462, "created_at": "2025-01-14T15:54:47.599097+00:00"} {"global_step": 2070, "acc_step": 0, "speed/wps": 12832.102471488333, "speed/FLOPS": 201545802471342.84, "speed/curr_iter_time": 1.2878, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.22445769608020782, "optim/lr": 0.0012414, "optim/total_tokens": 542638080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.8238139152526855, "created_at": "2025-01-14T15:54:52.710175+00:00"} {"global_step": 2071, "acc_step": 0, "speed/wps": 12837.472888498105, "speed/FLOPS": 201630152250207.62, "speed/curr_iter_time": 1.2874, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19300775229930878, "optim/lr": 0.001242, "optim/total_tokens": 542900224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 3.7983474731445312, "created_at": "2025-01-14T15:54:57.818848+00:00"} {"global_step": 2072, "acc_step": 0, "speed/wps": 12833.558863157477, "speed/FLOPS": 201568677103798.62, "speed/curr_iter_time": 1.2875, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20288817584514618, "optim/lr": 0.0012426000000000002, "optim/total_tokens": 543162368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305105, "loss/out": 3.8365719318389893, "created_at": "2025-01-14T15:55:02.930909+00:00"} {"global_step": 2073, "acc_step": 0, "speed/wps": 12830.228070790114, "speed/FLOPS": 201516362432680.84, "speed/curr_iter_time": 1.2875, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22699406743049622, "optim/lr": 0.0012432, "optim/total_tokens": 543424512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 3.807421922683716, "created_at": "2025-01-14T15:55:08.043722+00:00"} {"global_step": 2074, "acc_step": 0, "speed/wps": 12834.73015541988, "speed/FLOPS": 201587073858303.9, "speed/curr_iter_time": 1.2878, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21725232899188995, "optim/lr": 0.0012438000000000002, "optim/total_tokens": 543686656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475987, "loss/out": 3.8320813179016113, "created_at": "2025-01-14T15:55:13.153525+00:00"} {"global_step": 2075, "acc_step": 0, "speed/wps": 12835.738192719547, "speed/FLOPS": 201602906469283.6, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18314461410045624, "optim/lr": 0.0012444, "optim/total_tokens": 543948800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 3.7538561820983887, "created_at": "2025-01-14T15:55:18.262963+00:00"} {"global_step": 2076, "acc_step": 0, "speed/wps": 12834.303616554611, "speed/FLOPS": 201580374479299.03, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24151338636875153, "optim/lr": 0.001245, "optim/total_tokens": 544210944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.7734460830688477, "created_at": "2025-01-14T15:55:23.373322+00:00"} {"global_step": 2077, "acc_step": 0, "speed/wps": 12831.815610294378, "speed/FLOPS": 201541296922103.2, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21137084066867828, "optim/lr": 0.0012456000000000001, "optim/total_tokens": 544473088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326870, "loss/out": 3.8227572441101074, "created_at": "2025-01-14T15:55:28.485204+00:00"} {"global_step": 2078, "acc_step": 0, "speed/wps": 12838.880304835962, "speed/FLOPS": 201652257657786.4, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2030320167541504, "optim/lr": 0.0012462, "optim/total_tokens": 544735232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.787208080291748, "created_at": "2025-01-14T15:55:33.598508+00:00"} {"global_step": 2079, "acc_step": 0, "speed/wps": 12812.017346808096, "speed/FLOPS": 201230337988386.16, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22090350091457367, "optim/lr": 0.0012468000000000002, "optim/total_tokens": 544997376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353982, "loss/out": 3.782899856567383, "created_at": "2025-01-14T15:55:38.717477+00:00"} {"global_step": 2080, "acc_step": 0, "speed/wps": 12836.125630661809, "speed/FLOPS": 201608991714562.3, "speed/curr_iter_time": 1.288, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19858291745185852, "optim/lr": 0.0012474, "optim/total_tokens": 545259520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494003, "loss/out": 3.810903310775757, "created_at": "2025-01-14T15:55:43.827881+00:00"} {"global_step": 2081, "acc_step": 0, "speed/wps": 12838.941867650701, "speed/FLOPS": 201653224586388.03, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1963646411895752, "optim/lr": 0.001248, "optim/total_tokens": 545521664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.8472423553466797, "created_at": "2025-01-14T15:55:48.936039+00:00"} {"global_step": 2082, "acc_step": 0, "speed/wps": 12830.901635088861, "speed/FLOPS": 201526941685567.03, "speed/curr_iter_time": 1.2887, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1906125545501709, "optim/lr": 0.0012486, "optim/total_tokens": 545783808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325239, "loss/out": 3.8576555252075195, "created_at": "2025-01-14T15:55:54.050636+00:00"} {"global_step": 2083, "acc_step": 0, "speed/wps": 12842.823622743437, "speed/FLOPS": 201714192884209.94, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17750754952430725, "optim/lr": 0.0012492, "optim/total_tokens": 546045952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 3.7698330879211426, "created_at": "2025-01-14T15:55:59.159051+00:00"} {"global_step": 2084, "acc_step": 0, "speed/wps": 12836.849723473306, "speed/FLOPS": 201620364587174.7, "speed/curr_iter_time": 1.2876, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19672006368637085, "optim/lr": 0.0012498000000000001, "optim/total_tokens": 546308096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413141, "loss/out": 3.637406349182129, "created_at": "2025-01-14T15:56:04.269125+00:00"} {"global_step": 2085, "acc_step": 0, "speed/wps": 12852.225500805187, "speed/FLOPS": 201861862298704.03, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1976315677165985, "optim/lr": 0.0012504, "optim/total_tokens": 546570240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.666998863220215, "created_at": "2025-01-14T15:56:09.372015+00:00"} {"global_step": 2086, "acc_step": 0, "speed/wps": 12845.053824668337, "speed/FLOPS": 201749221270059.1, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1882791966199875, "optim/lr": 0.001251, "optim/total_tokens": 546832384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 3.7588329315185547, "created_at": "2025-01-14T15:56:14.478582+00:00"} {"global_step": 2087, "acc_step": 0, "speed/wps": 12841.487570383837, "speed/FLOPS": 201693208345974.06, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25348836183547974, "optim/lr": 0.0012516, "optim/total_tokens": 547094528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.8636465072631836, "created_at": "2025-01-14T15:56:19.589064+00:00"} {"global_step": 2088, "acc_step": 0, "speed/wps": 12840.324674216236, "speed/FLOPS": 201674943463675.1, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.21229518949985504, "optim/lr": 0.0012522, "optim/total_tokens": 547356672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479622, "loss/out": 3.782625198364258, "created_at": "2025-01-14T15:56:24.700226+00:00"} {"global_step": 2089, "acc_step": 0, "speed/wps": 12846.081771394136, "speed/FLOPS": 201765366585934.56, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.19019715487957, "optim/lr": 0.0012528, "optim/total_tokens": 547618816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336239, "loss/out": 3.8386166095733643, "created_at": "2025-01-14T15:56:29.807953+00:00"} {"global_step": 2090, "acc_step": 0, "speed/wps": 12836.733993800015, "speed/FLOPS": 201618546893626.4, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16570957005023956, "optim/lr": 0.0012534, "optim/total_tokens": 547880960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.751650094985962, "created_at": "2025-01-14T15:56:34.918416+00:00"} {"global_step": 2091, "acc_step": 0, "speed/wps": 12843.307011771107, "speed/FLOPS": 201721785173135.62, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1683775633573532, "optim/lr": 0.001254, "optim/total_tokens": 548143104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331435, "loss/out": 3.836855173110962, "created_at": "2025-01-14T15:56:40.026165+00:00"} {"global_step": 2092, "acc_step": 0, "speed/wps": 12840.959321075821, "speed/FLOPS": 201684911464701.06, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2079785019159317, "optim/lr": 0.0012546, "optim/total_tokens": 548405248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470119, "loss/out": 3.8769028186798096, "created_at": "2025-01-14T15:56:45.135706+00:00"} {"global_step": 2093, "acc_step": 0, "speed/wps": 12844.319322705733, "speed/FLOPS": 201737684907424.9, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17891161143779755, "optim/lr": 0.0012552, "optim/total_tokens": 548667392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475174, "loss/out": 3.7976596355438232, "created_at": "2025-01-14T15:56:50.244561+00:00"} {"global_step": 2094, "acc_step": 0, "speed/wps": 12837.994499078604, "speed/FLOPS": 201638344861142.44, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.20771563053131104, "optim/lr": 0.0012558, "optim/total_tokens": 548929536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.7933707237243652, "created_at": "2025-01-14T15:56:55.354874+00:00"} {"global_step": 2095, "acc_step": 0, "speed/wps": 12841.17767410087, "speed/FLOPS": 201688340999007.97, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.17385725677013397, "optim/lr": 0.0012564, "optim/total_tokens": 549191680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.80204701423645, "created_at": "2025-01-14T15:57:00.463171+00:00"} {"global_step": 2096, "acc_step": 0, "speed/wps": 12848.8949045289, "speed/FLOPS": 201809550707466.06, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20794638991355896, "optim/lr": 0.0012569999999999999, "optim/total_tokens": 549453824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309558, "loss/out": 3.891866683959961, "created_at": "2025-01-14T15:57:05.577264+00:00"} {"global_step": 2097, "acc_step": 0, "speed/wps": 12843.458692470364, "speed/FLOPS": 201724167526948.56, "speed/curr_iter_time": 1.288, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2218911498785019, "optim/lr": 0.0012576, "optim/total_tokens": 549715968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 3.8429861068725586, "created_at": "2025-01-14T15:57:10.687074+00:00"} {"global_step": 2098, "acc_step": 0, "speed/wps": 12829.557494027687, "speed/FLOPS": 201505830103157.9, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.20444895327091217, "optim/lr": 0.0012582000000000001, "optim/total_tokens": 549978112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 3.6930196285247803, "created_at": "2025-01-14T15:57:15.798882+00:00"} {"global_step": 2099, "acc_step": 0, "speed/wps": 12839.129405851198, "speed/FLOPS": 201656170131531.34, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.18710601329803467, "optim/lr": 0.0012588, "optim/total_tokens": 550240256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328746, "loss/out": 3.734769821166992, "created_at": "2025-01-14T15:57:20.910542+00:00"} {"global_step": 2100, "acc_step": 0, "speed/wps": 12839.054938535513, "speed/FLOPS": 201655000520009.7, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21521049737930298, "optim/lr": 0.0012594000000000001, "optim/total_tokens": 550502400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444641, "loss/out": 3.6589696407318115, "created_at": "2025-01-14T15:57:26.019808+00:00"} {"global_step": 2101, "acc_step": 0, "speed/wps": 12840.938524932673, "speed/FLOPS": 201684584832696.03, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22515365481376648, "optim/lr": 0.00126, "optim/total_tokens": 550764544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477559, "loss/out": 3.72653865814209, "created_at": "2025-01-14T15:57:31.127374+00:00"} {"global_step": 2102, "acc_step": 0, "speed/wps": 12839.751639582433, "speed/FLOPS": 201665943167319.12, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21743519604206085, "optim/lr": 0.0012606000000000002, "optim/total_tokens": 551026688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478047, "loss/out": 3.891848087310791, "created_at": "2025-01-14T15:57:36.237116+00:00"} {"global_step": 2103, "acc_step": 0, "speed/wps": 12843.00436431364, "speed/FLOPS": 201717031678934.9, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19080400466918945, "optim/lr": 0.0012612, "optim/total_tokens": 551288832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.8380300998687744, "created_at": "2025-01-14T15:57:41.344008+00:00"} {"global_step": 2104, "acc_step": 0, "speed/wps": 12842.077432607992, "speed/FLOPS": 201702472942760.06, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18075449764728546, "optim/lr": 0.0012618, "optim/total_tokens": 551550976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 3.79097843170166, "created_at": "2025-01-14T15:57:46.452177+00:00"} {"global_step": 2105, "acc_step": 0, "speed/wps": 12836.552871366895, "speed/FLOPS": 201615702116925.97, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17111064493656158, "optim/lr": 0.0012624000000000001, "optim/total_tokens": 551813120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.847097635269165, "created_at": "2025-01-14T15:57:51.561221+00:00"} {"global_step": 2106, "acc_step": 0, "speed/wps": 12843.324463530636, "speed/FLOPS": 201722059277000.25, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15986134111881256, "optim/lr": 0.001263, "optim/total_tokens": 552075264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 3.7855324745178223, "created_at": "2025-01-14T15:57:56.670033+00:00"} {"global_step": 2107, "acc_step": 0, "speed/wps": 12839.034476514276, "speed/FLOPS": 201654679135848.4, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2157817780971527, "optim/lr": 0.0012636000000000001, "optim/total_tokens": 552337408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 3.788217306137085, "created_at": "2025-01-14T15:58:01.777906+00:00"} {"global_step": 2108, "acc_step": 0, "speed/wps": 12841.608604686671, "speed/FLOPS": 201695109356018.66, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2895689904689789, "optim/lr": 0.0012642, "optim/total_tokens": 552599552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366319, "loss/out": 3.8310537338256836, "created_at": "2025-01-14T15:58:06.884883+00:00"} {"global_step": 2109, "acc_step": 0, "speed/wps": 12824.84751449625, "speed/FLOPS": 201431853402426.56, "speed/curr_iter_time": 1.288, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21474888920783997, "optim/lr": 0.0012648, "optim/total_tokens": 552861696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 3.790034770965576, "created_at": "2025-01-14T15:58:12.001390+00:00"} {"global_step": 2110, "acc_step": 0, "speed/wps": 12833.597311176782, "speed/FLOPS": 201569280982774.94, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18644346296787262, "optim/lr": 0.0012654, "optim/total_tokens": 553123840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.7662429809570312, "created_at": "2025-01-14T15:58:17.114878+00:00"} {"global_step": 2111, "acc_step": 0, "speed/wps": 12846.326309117461, "speed/FLOPS": 201769207386909.44, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2454610913991928, "optim/lr": 0.001266, "optim/total_tokens": 553385984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305274, "loss/out": 3.828065872192383, "created_at": "2025-01-14T15:58:22.220866+00:00"} {"global_step": 2112, "acc_step": 0, "speed/wps": 12840.129666980203, "speed/FLOPS": 201671880607065.2, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20450855791568756, "optim/lr": 0.0012666, "optim/total_tokens": 553648128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.6263318061828613, "created_at": "2025-01-14T15:58:27.333004+00:00"} {"global_step": 2113, "acc_step": 0, "speed/wps": 12836.223312742586, "speed/FLOPS": 201610525945870.2, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2065107822418213, "optim/lr": 0.0012672, "optim/total_tokens": 553910272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482522, "loss/out": 3.630491256713867, "created_at": "2025-01-14T15:58:32.443482+00:00"} {"global_step": 2114, "acc_step": 0, "speed/wps": 12838.81392887861, "speed/FLOPS": 201651215132166.06, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21676640212535858, "optim/lr": 0.0012678, "optim/total_tokens": 554172416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 3.7079501152038574, "created_at": "2025-01-14T15:58:37.553969+00:00"} {"global_step": 2115, "acc_step": 0, "speed/wps": 12841.28040098515, "speed/FLOPS": 201689954465887.1, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.19084274768829346, "optim/lr": 0.0012684, "optim/total_tokens": 554434560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476195, "loss/out": 3.7505288124084473, "created_at": "2025-01-14T15:58:42.665190+00:00"} {"global_step": 2116, "acc_step": 0, "speed/wps": 12840.560093757782, "speed/FLOPS": 201678641051075.44, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18156687915325165, "optim/lr": 0.001269, "optim/total_tokens": 554696704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297061, "loss/out": 3.674091100692749, "created_at": "2025-01-14T15:58:47.772530+00:00"} {"global_step": 2117, "acc_step": 0, "speed/wps": 12841.491530771596, "speed/FLOPS": 201693270549305.97, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.001, "optim/grad_norm": 0.19699855148792267, "optim/lr": 0.0012696, "optim/total_tokens": 554958848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466155, "loss/out": 3.737243175506592, "created_at": "2025-01-14T15:58:52.888956+00:00"} {"global_step": 2118, "acc_step": 0, "speed/wps": 12843.138893630028, "speed/FLOPS": 201719144646711.44, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.162157341837883, "optim/lr": 0.0012702, "optim/total_tokens": 555220992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 347191, "loss/out": 3.628356456756592, "created_at": "2025-01-14T15:58:57.995987+00:00"} {"global_step": 2119, "acc_step": 0, "speed/wps": 12844.95534928575, "speed/FLOPS": 201747674578856.2, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20211881399154663, "optim/lr": 0.0012707999999999999, "optim/total_tokens": 555483136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.7558441162109375, "created_at": "2025-01-14T15:59:03.105801+00:00"} {"global_step": 2120, "acc_step": 0, "speed/wps": 12826.0672130732, "speed/FLOPS": 201451010444619.06, "speed/curr_iter_time": 1.2887, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1908315122127533, "optim/lr": 0.0012714, "optim/total_tokens": 555745280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.765028238296509, "created_at": "2025-01-14T15:59:08.218903+00:00"} {"global_step": 2121, "acc_step": 0, "speed/wps": 12831.29155365421, "speed/FLOPS": 201533065892440.06, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21353323757648468, "optim/lr": 0.001272, "optim/total_tokens": 556007424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 3.778355836868286, "created_at": "2025-01-14T15:59:13.330885+00:00"} {"global_step": 2122, "acc_step": 0, "speed/wps": 12838.213081538583, "speed/FLOPS": 201641777999040.16, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2494349181652069, "optim/lr": 0.0012726, "optim/total_tokens": 556269568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 3.6486568450927734, "created_at": "2025-01-14T15:59:18.443155+00:00"} {"global_step": 2123, "acc_step": 0, "speed/wps": 12822.193492731458, "speed/FLOPS": 201390168343602.4, "speed/curr_iter_time": 1.288, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.25202450156211853, "optim/lr": 0.0012732, "optim/total_tokens": 556531712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.748223066329956, "created_at": "2025-01-14T15:59:23.560724+00:00"} {"global_step": 2124, "acc_step": 0, "speed/wps": 12828.915167949777, "speed/FLOPS": 201495741489455.8, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0018, "optim/grad_norm": 0.3112453818321228, "optim/lr": 0.0012738, "optim/total_tokens": 556793856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478047, "loss/out": 3.7692904472351074, "created_at": "2025-01-14T15:59:28.684155+00:00"} {"global_step": 2125, "acc_step": 0, "speed/wps": 12842.205380181835, "speed/FLOPS": 201704482535225.66, "speed/curr_iter_time": 1.2875, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19658520817756653, "optim/lr": 0.0012744, "optim/total_tokens": 557056000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454458, "loss/out": 3.752847909927368, "created_at": "2025-01-14T15:59:33.795449+00:00"} {"global_step": 2126, "acc_step": 0, "speed/wps": 12830.1987311518, "speed/FLOPS": 201515901613344.25, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23843052983283997, "optim/lr": 0.001275, "optim/total_tokens": 557318144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.7413010597229004, "created_at": "2025-01-14T15:59:38.910249+00:00"} {"global_step": 2127, "acc_step": 0, "speed/wps": 12831.499659675459, "speed/FLOPS": 201536334483471.44, "speed/curr_iter_time": 1.2876, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18663956224918365, "optim/lr": 0.0012756, "optim/total_tokens": 557580288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476715, "loss/out": 3.809739112854004, "created_at": "2025-01-14T15:59:44.025545+00:00"} {"global_step": 2128, "acc_step": 0, "speed/wps": 12840.159433954652, "speed/FLOPS": 201672348138303.03, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2014269083738327, "optim/lr": 0.0012762000000000001, "optim/total_tokens": 557842432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 3.700308322906494, "created_at": "2025-01-14T15:59:49.137355+00:00"} {"global_step": 2129, "acc_step": 0, "speed/wps": 12833.583462186321, "speed/FLOPS": 201569063465349.25, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.19319893419742584, "optim/lr": 0.0012768, "optim/total_tokens": 558104576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 383639, "loss/out": 3.8141019344329834, "created_at": "2025-01-14T15:59:54.248952+00:00"} {"global_step": 2130, "acc_step": 0, "speed/wps": 12830.614071936421, "speed/FLOPS": 201522425111104.75, "speed/curr_iter_time": 1.2882, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17380213737487793, "optim/lr": 0.0012774000000000001, "optim/total_tokens": 558366720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 3.8213322162628174, "created_at": "2025-01-14T15:59:59.361176+00:00"} {"global_step": 2131, "acc_step": 0, "speed/wps": 12852.23870698017, "speed/FLOPS": 201862069719829.34, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.001, "optim/grad_norm": 0.20853887498378754, "optim/lr": 0.001278, "optim/total_tokens": 558628864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367755, "loss/out": 3.742597818374634, "created_at": "2025-01-14T16:00:04.463936+00:00"} {"global_step": 2132, "acc_step": 0, "speed/wps": 12844.830216753919, "speed/FLOPS": 201745709200493.66, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2224186360836029, "optim/lr": 0.0012786000000000002, "optim/total_tokens": 558891008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.7276158332824707, "created_at": "2025-01-14T16:00:09.573333+00:00"} {"global_step": 2133, "acc_step": 0, "speed/wps": 12836.362989602243, "speed/FLOPS": 201612719762887.22, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.16173075139522552, "optim/lr": 0.0012792, "optim/total_tokens": 559153152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 504269, "loss/out": 3.764918804168701, "created_at": "2025-01-14T16:00:14.685126+00:00"} {"global_step": 2134, "acc_step": 0, "speed/wps": 12845.155699121482, "speed/FLOPS": 201750821348335.88, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1877245157957077, "optim/lr": 0.0012798, "optim/total_tokens": 559415296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 3.7743656635284424, "created_at": "2025-01-14T16:00:19.794097+00:00"} {"global_step": 2135, "acc_step": 0, "speed/wps": 12841.675309510916, "speed/FLOPS": 201696157046945.44, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19187481701374054, "optim/lr": 0.0012804000000000001, "optim/total_tokens": 559677440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468777, "loss/out": 3.817762851715088, "created_at": "2025-01-14T16:00:24.902928+00:00"} {"global_step": 2136, "acc_step": 0, "speed/wps": 12823.888303159056, "speed/FLOPS": 201416787670281.44, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1885017454624176, "optim/lr": 0.001281, "optim/total_tokens": 559939584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464846, "loss/out": 3.761601448059082, "created_at": "2025-01-14T16:00:30.020204+00:00"} {"global_step": 2137, "acc_step": 0, "speed/wps": 12850.180627004744, "speed/FLOPS": 201829744745717.0, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17543824017047882, "optim/lr": 0.0012816000000000001, "optim/total_tokens": 560201728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.663691282272339, "created_at": "2025-01-14T16:00:35.127555+00:00"} {"global_step": 2138, "acc_step": 0, "speed/wps": 12839.266119306365, "speed/FLOPS": 201658317404208.72, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22113493084907532, "optim/lr": 0.0012822, "optim/total_tokens": 560463872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.854008197784424, "created_at": "2025-01-14T16:00:40.240560+00:00"} {"global_step": 2139, "acc_step": 0, "speed/wps": 12847.950739397844, "speed/FLOPS": 201794721296663.97, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21765018999576569, "optim/lr": 0.0012828, "optim/total_tokens": 560726016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300465, "loss/out": 3.7651267051696777, "created_at": "2025-01-14T16:00:45.345574+00:00"} {"global_step": 2140, "acc_step": 0, "speed/wps": 12842.68206844797, "speed/FLOPS": 201711969579483.62, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21084976196289062, "optim/lr": 0.0012834, "optim/total_tokens": 560988160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 3.7547898292541504, "created_at": "2025-01-14T16:00:50.452581+00:00"} {"global_step": 2141, "acc_step": 0, "speed/wps": 12852.067603907568, "speed/FLOPS": 201859382310953.84, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18674568831920624, "optim/lr": 0.001284, "optim/total_tokens": 561250304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360703, "loss/out": 3.7656795978546143, "created_at": "2025-01-14T16:00:55.563873+00:00"} {"global_step": 2142, "acc_step": 0, "speed/wps": 12841.98366662876, "speed/FLOPS": 201701000219208.25, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20395229756832123, "optim/lr": 0.0012846, "optim/total_tokens": 561512448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 506365, "loss/out": 3.6903083324432373, "created_at": "2025-01-14T16:01:00.673694+00:00"} {"global_step": 2143, "acc_step": 0, "speed/wps": 12846.980446500073, "speed/FLOPS": 201779481513382.3, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.19910167157649994, "optim/lr": 0.0012852, "optim/total_tokens": 561774592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.814011335372925, "created_at": "2025-01-14T16:01:05.783507+00:00"} {"global_step": 2144, "acc_step": 0, "speed/wps": 12843.027389247025, "speed/FLOPS": 201717393317153.28, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2299925982952118, "optim/lr": 0.0012858, "optim/total_tokens": 562036736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 3.8738763332366943, "created_at": "2025-01-14T16:01:10.894628+00:00"} {"global_step": 2145, "acc_step": 0, "speed/wps": 12838.436932269642, "speed/FLOPS": 201645293882374.6, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22715231776237488, "optim/lr": 0.0012864, "optim/total_tokens": 562298880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499296, "loss/out": 3.8371517658233643, "created_at": "2025-01-14T16:01:16.006248+00:00"} {"global_step": 2146, "acc_step": 0, "speed/wps": 12846.82333281102, "speed/FLOPS": 201777013826998.9, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20075829327106476, "optim/lr": 0.001287, "optim/total_tokens": 562561024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.713656425476074, "created_at": "2025-01-14T16:01:21.112087+00:00"} {"global_step": 2147, "acc_step": 0, "speed/wps": 12844.591308994186, "speed/FLOPS": 201741956825833.4, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17766794562339783, "optim/lr": 0.0012876, "optim/total_tokens": 562823168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.7161736488342285, "created_at": "2025-01-14T16:01:26.220748+00:00"} {"global_step": 2148, "acc_step": 0, "speed/wps": 12841.705936687544, "speed/FLOPS": 201696638088840.53, "speed/curr_iter_time": 1.2875, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19975432753562927, "optim/lr": 0.0012882, "optim/total_tokens": 563085312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364445, "loss/out": 3.658965587615967, "created_at": "2025-01-14T16:01:31.330762+00:00"} {"global_step": 2149, "acc_step": 0, "speed/wps": 12850.63642052034, "speed/FLOPS": 201836903609205.4, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18677756190299988, "optim/lr": 0.0012887999999999999, "optim/total_tokens": 563347456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.839461088180542, "created_at": "2025-01-14T16:01:36.435412+00:00"} {"global_step": 2150, "acc_step": 0, "speed/wps": 12845.961560293392, "speed/FLOPS": 201763478505410.03, "speed/curr_iter_time": 1.2874, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2019590139389038, "optim/lr": 0.0012894, "optim/total_tokens": 563609600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 3.7566399574279785, "created_at": "2025-01-14T16:01:41.545441+00:00"} {"global_step": 2151, "acc_step": 0, "speed/wps": 12847.537629713774, "speed/FLOPS": 201788232841403.47, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20640018582344055, "optim/lr": 0.00129, "optim/total_tokens": 563871744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475383, "loss/out": 3.6678013801574707, "created_at": "2025-01-14T16:01:46.652154+00:00"} {"global_step": 2152, "acc_step": 0, "speed/wps": 12850.30177125961, "speed/FLOPS": 201831647482709.34, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18730413913726807, "optim/lr": 0.0012906, "optim/total_tokens": 564133888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.767129421234131, "created_at": "2025-01-14T16:01:51.755876+00:00"} {"global_step": 2153, "acc_step": 0, "speed/wps": 12848.089836481486, "speed/FLOPS": 201796906007502.0, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20826934278011322, "optim/lr": 0.0012912, "optim/total_tokens": 564396032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310904, "loss/out": 3.7781639099121094, "created_at": "2025-01-14T16:01:56.861395+00:00"} {"global_step": 2154, "acc_step": 0, "speed/wps": 12852.149696101807, "speed/FLOPS": 201860671681670.88, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.19662930071353912, "optim/lr": 0.0012918, "optim/total_tokens": 564658176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472504, "loss/out": 3.649975061416626, "created_at": "2025-01-14T16:02:01.965195+00:00"} {"global_step": 2155, "acc_step": 0, "speed/wps": 12850.004146499166, "speed/FLOPS": 201826972876867.28, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1980496644973755, "optim/lr": 0.0012924, "optim/total_tokens": 564920320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336478, "loss/out": 3.6817893981933594, "created_at": "2025-01-14T16:02:07.074159+00:00"} {"global_step": 2156, "acc_step": 0, "speed/wps": 12844.548206804462, "speed/FLOPS": 201741279846715.16, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19695709645748138, "optim/lr": 0.001293, "optim/total_tokens": 565182464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 3.7085936069488525, "created_at": "2025-01-14T16:02:12.180268+00:00"} {"global_step": 2157, "acc_step": 0, "speed/wps": 12852.499467595115, "speed/FLOPS": 201866165323609.7, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21003544330596924, "optim/lr": 0.0012936000000000002, "optim/total_tokens": 565444608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308499, "loss/out": 3.763643503189087, "created_at": "2025-01-14T16:02:17.284058+00:00"} {"global_step": 2158, "acc_step": 0, "speed/wps": 12840.596744405648, "speed/FLOPS": 201679216699863.44, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18673540651798248, "optim/lr": 0.0012942000000000001, "optim/total_tokens": 565706752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461368, "loss/out": 3.6913933753967285, "created_at": "2025-01-14T16:02:22.396888+00:00"} {"global_step": 2159, "acc_step": 0, "speed/wps": 12846.733461136886, "speed/FLOPS": 201775602268860.03, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.1991024762392044, "optim/lr": 0.0012948, "optim/total_tokens": 565968896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380268, "loss/out": 3.6501524448394775, "created_at": "2025-01-14T16:02:27.503377+00:00"} {"global_step": 2160, "acc_step": 0, "speed/wps": 12849.674933955053, "speed/FLOPS": 201821802141476.56, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18422353267669678, "optim/lr": 0.0012954000000000002, "optim/total_tokens": 566231040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 3.8035593032836914, "created_at": "2025-01-14T16:02:32.611657+00:00"} {"global_step": 2161, "acc_step": 0, "speed/wps": 12841.0781778917, "speed/FLOPS": 201686778274320.84, "speed/curr_iter_time": 1.2876, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19102025032043457, "optim/lr": 0.001296, "optim/total_tokens": 566493184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.6267499923706055, "created_at": "2025-01-14T16:02:37.722454+00:00"} {"global_step": 2162, "acc_step": 0, "speed/wps": 12830.02911268225, "speed/FLOPS": 201513237522199.75, "speed/curr_iter_time": 1.2894, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19967404007911682, "optim/lr": 0.0012966, "optim/total_tokens": 566755328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310904, "loss/out": 3.7305538654327393, "created_at": "2025-01-14T16:02:42.834326+00:00"} {"global_step": 2163, "acc_step": 0, "speed/wps": 12836.85899205361, "speed/FLOPS": 201620510162964.72, "speed/curr_iter_time": 1.2874, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16051322221755981, "optim/lr": 0.0012972, "optim/total_tokens": 567017472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 3.7239112854003906, "created_at": "2025-01-14T16:02:47.945772+00:00"} {"global_step": 2164, "acc_step": 0, "speed/wps": 12848.98566514225, "speed/FLOPS": 201810976227616.88, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18690061569213867, "optim/lr": 0.0012978, "optim/total_tokens": 567279616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460308, "loss/out": 3.66398286819458, "created_at": "2025-01-14T16:02:53.050581+00:00"} {"global_step": 2165, "acc_step": 0, "speed/wps": 12848.782063788332, "speed/FLOPS": 201807778388576.03, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17445679008960724, "optim/lr": 0.0012984000000000001, "optim/total_tokens": 567541760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.711923122406006, "created_at": "2025-01-14T16:02:58.155545+00:00"} {"global_step": 2166, "acc_step": 0, "speed/wps": 12845.50652979262, "speed/FLOPS": 201756331626117.97, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1685209572315216, "optim/lr": 0.001299, "optim/total_tokens": 567803904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.51902437210083, "created_at": "2025-01-14T16:03:03.264003+00:00"} {"global_step": 2167, "acc_step": 0, "speed/wps": 12848.82325651709, "speed/FLOPS": 201808425376985.88, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.16487233340740204, "optim/lr": 0.0012996, "optim/total_tokens": 568066048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 3.8380517959594727, "created_at": "2025-01-14T16:03:08.368090+00:00"} {"global_step": 2168, "acc_step": 0, "speed/wps": 12838.429628949016, "speed/FLOPS": 201645179173688.34, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1776803880929947, "optim/lr": 0.0013002, "optim/total_tokens": 568328192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473004, "loss/out": 3.7418696880340576, "created_at": "2025-01-14T16:03:13.477575+00:00"} {"global_step": 2169, "acc_step": 0, "speed/wps": 12850.473130042, "speed/FLOPS": 201834338907858.44, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15961572527885437, "optim/lr": 0.0013008, "optim/total_tokens": 568590336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309558, "loss/out": 3.8473591804504395, "created_at": "2025-01-14T16:03:18.585325+00:00"} {"global_step": 2170, "acc_step": 0, "speed/wps": 12852.444084568613, "speed/FLOPS": 201865295457074.22, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15656156837940216, "optim/lr": 0.0013014, "optim/total_tokens": 568852480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.719461441040039, "created_at": "2025-01-14T16:03:23.692060+00:00"} {"global_step": 2171, "acc_step": 0, "speed/wps": 12851.26771939382, "speed/FLOPS": 201846819025508.94, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18840229511260986, "optim/lr": 0.001302, "optim/total_tokens": 569114624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495289, "loss/out": 3.691347360610962, "created_at": "2025-01-14T16:03:28.795402+00:00"} {"global_step": 2172, "acc_step": 0, "speed/wps": 12848.79084519822, "speed/FLOPS": 201807916312685.22, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.001, "optim/grad_norm": 0.18432225286960602, "optim/lr": 0.0013026, "optim/total_tokens": 569376768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 3.7027297019958496, "created_at": "2025-01-14T16:03:33.899878+00:00"} {"global_step": 2173, "acc_step": 0, "speed/wps": 12846.315964491167, "speed/FLOPS": 201769044910338.47, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19328974187374115, "optim/lr": 0.0013032, "optim/total_tokens": 569638912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327688, "loss/out": 3.815748929977417, "created_at": "2025-01-14T16:03:39.005075+00:00"} {"global_step": 2174, "acc_step": 0, "speed/wps": 12851.173319259358, "speed/FLOPS": 201845336341677.84, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1701943576335907, "optim/lr": 0.0013038, "optim/total_tokens": 569901056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461622, "loss/out": 3.8376283645629883, "created_at": "2025-01-14T16:03:44.108432+00:00"} {"global_step": 2175, "acc_step": 0, "speed/wps": 12850.573759872404, "speed/FLOPS": 201835919437624.8, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.170224130153656, "optim/lr": 0.0013044, "optim/total_tokens": 570163200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444068, "loss/out": 3.8203043937683105, "created_at": "2025-01-14T16:03:49.212908+00:00"} {"global_step": 2176, "acc_step": 0, "speed/wps": 12846.51853317541, "speed/FLOPS": 201772226529883.06, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20167425274848938, "optim/lr": 0.001305, "optim/total_tokens": 570425344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.6896727085113525, "created_at": "2025-01-14T16:03:54.319588+00:00"} {"global_step": 2177, "acc_step": 0, "speed/wps": 12852.19772708244, "speed/FLOPS": 201861426074223.22, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20836971700191498, "optim/lr": 0.0013055999999999999, "optim/total_tokens": 570687488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 3.6539440155029297, "created_at": "2025-01-14T16:03:59.426316+00:00"} {"global_step": 2178, "acc_step": 0, "speed/wps": 12846.983246943135, "speed/FLOPS": 201779525498188.8, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18459923565387726, "optim/lr": 0.0013062, "optim/total_tokens": 570949632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.729860544204712, "created_at": "2025-01-14T16:04:04.535641+00:00"} {"global_step": 2179, "acc_step": 0, "speed/wps": 12852.424108016585, "speed/FLOPS": 201864981697873.94, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.22061479091644287, "optim/lr": 0.0013067999999999999, "optim/total_tokens": 571211776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475696, "loss/out": 3.8740382194519043, "created_at": "2025-01-14T16:04:09.648493+00:00"} {"global_step": 2180, "acc_step": 0, "speed/wps": 12853.66936450481, "speed/FLOPS": 201884540162183.28, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18082039058208466, "optim/lr": 0.0013074, "optim/total_tokens": 571473920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 399826, "loss/out": 3.8680102825164795, "created_at": "2025-01-14T16:04:14.752967+00:00"} {"global_step": 2181, "acc_step": 0, "speed/wps": 12844.95189321389, "speed/FLOPS": 201747620296498.5, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20300093293190002, "optim/lr": 0.001308, "optim/total_tokens": 571736064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502937, "loss/out": 3.80018949508667, "created_at": "2025-01-14T16:04:19.859221+00:00"} {"global_step": 2182, "acc_step": 0, "speed/wps": 12836.171050607352, "speed/FLOPS": 201609705097220.62, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21437270939350128, "optim/lr": 0.0013086, "optim/total_tokens": 571998208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.8030457496643066, "created_at": "2025-01-14T16:04:24.970258+00:00"} {"global_step": 2183, "acc_step": 0, "speed/wps": 12851.01905305743, "speed/FLOPS": 201842913378992.56, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21345320343971252, "optim/lr": 0.0013092000000000002, "optim/total_tokens": 572260352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494768, "loss/out": 3.8037595748901367, "created_at": "2025-01-14T16:04:30.074677+00:00"} {"global_step": 2184, "acc_step": 0, "speed/wps": 12853.513932550946, "speed/FLOPS": 201882098889761.94, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21219943463802338, "optim/lr": 0.0013098, "optim/total_tokens": 572522496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.824375867843628, "created_at": "2025-01-14T16:04:35.179475+00:00"} {"global_step": 2185, "acc_step": 0, "speed/wps": 12852.09896373903, "speed/FLOPS": 201859874860199.9, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.18864938616752625, "optim/lr": 0.0013104000000000002, "optim/total_tokens": 572784640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.7560887336730957, "created_at": "2025-01-14T16:04:40.284134+00:00"} {"global_step": 2186, "acc_step": 0, "speed/wps": 12848.007265991264, "speed/FLOPS": 201795609124489.3, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.24097363650798798, "optim/lr": 0.001311, "optim/total_tokens": 573046784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.6886050701141357, "created_at": "2025-01-14T16:04:45.390674+00:00"} {"global_step": 2187, "acc_step": 0, "speed/wps": 12844.295362393499, "speed/FLOPS": 201737308577796.38, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21143995225429535, "optim/lr": 0.0013116, "optim/total_tokens": 573308928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319751, "loss/out": 3.6763856410980225, "created_at": "2025-01-14T16:04:50.497282+00:00"} {"global_step": 2188, "acc_step": 0, "speed/wps": 12847.443155362673, "speed/FLOPS": 201786748991898.06, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1816408336162567, "optim/lr": 0.0013122000000000001, "optim/total_tokens": 573571072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 3.773688316345215, "created_at": "2025-01-14T16:04:55.605640+00:00"} {"global_step": 2189, "acc_step": 0, "speed/wps": 12849.233746001815, "speed/FLOPS": 201814872678415.2, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.16168835759162903, "optim/lr": 0.0013128, "optim/total_tokens": 573833216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357724, "loss/out": 3.7032811641693115, "created_at": "2025-01-14T16:05:00.717354+00:00"} {"global_step": 2190, "acc_step": 0, "speed/wps": 12846.067241164377, "speed/FLOPS": 201765138368709.6, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.18349702656269073, "optim/lr": 0.0013134000000000002, "optim/total_tokens": 574095360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 3.781867504119873, "created_at": "2025-01-14T16:05:05.822541+00:00"} {"global_step": 2191, "acc_step": 0, "speed/wps": 12847.949479712233, "speed/FLOPS": 201794701511570.47, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17593102157115936, "optim/lr": 0.001314, "optim/total_tokens": 574357504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 3.7366912364959717, "created_at": "2025-01-14T16:05:10.927156+00:00"} {"global_step": 2192, "acc_step": 0, "speed/wps": 12850.438898752356, "speed/FLOPS": 201833801258415.84, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24989053606987, "optim/lr": 0.0013146, "optim/total_tokens": 574619648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 3.8178906440734863, "created_at": "2025-01-14T16:05:16.032741+00:00"} {"global_step": 2193, "acc_step": 0, "speed/wps": 12843.903314518913, "speed/FLOPS": 201731150927195.75, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.26963531970977783, "optim/lr": 0.0013152, "optim/total_tokens": 574881792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.7386529445648193, "created_at": "2025-01-14T16:05:21.142344+00:00"} {"global_step": 2194, "acc_step": 0, "speed/wps": 12847.04501649018, "speed/FLOPS": 201780495673805.4, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31950077414512634, "optim/lr": 0.0013158, "optim/total_tokens": 575143936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 281268, "loss/out": 3.7297747135162354, "created_at": "2025-01-14T16:05:26.247793+00:00"} {"global_step": 2195, "acc_step": 0, "speed/wps": 12847.19193733162, "speed/FLOPS": 201782803267510.62, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22835981845855713, "optim/lr": 0.0013164000000000001, "optim/total_tokens": 575406080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461605, "loss/out": 3.796539783477783, "created_at": "2025-01-14T16:05:31.353079+00:00"} {"global_step": 2196, "acc_step": 0, "speed/wps": 12850.081012352628, "speed/FLOPS": 201828180160720.66, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.26070138812065125, "optim/lr": 0.001317, "optim/total_tokens": 575668224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313070, "loss/out": 3.868885040283203, "created_at": "2025-01-14T16:05:36.456811+00:00"} {"global_step": 2197, "acc_step": 0, "speed/wps": 12849.780319272662, "speed/FLOPS": 201823457362704.22, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23611362278461456, "optim/lr": 0.0013176, "optim/total_tokens": 575930368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.7589588165283203, "created_at": "2025-01-14T16:05:41.567977+00:00"} {"global_step": 2198, "acc_step": 0, "speed/wps": 12846.772443134814, "speed/FLOPS": 201776214534704.34, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2338474839925766, "optim/lr": 0.0013182, "optim/total_tokens": 576192512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288491, "loss/out": 3.897124767303467, "created_at": "2025-01-14T16:05:46.673030+00:00"} {"global_step": 2199, "acc_step": 0, "speed/wps": 12850.438520934931, "speed/FLOPS": 201833795324274.0, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2301064431667328, "optim/lr": 0.0013188, "optim/total_tokens": 576454656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461913, "loss/out": 3.7045137882232666, "created_at": "2025-01-14T16:05:51.779087+00:00"} {"global_step": 2200, "acc_step": 0, "speed/wps": 12843.174548719458, "speed/FLOPS": 201719704658886.16, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21395833790302277, "optim/lr": 0.0013194, "optim/total_tokens": 576716800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 414278, "loss/out": 3.680865526199341, "created_at": "2025-01-14T16:05:56.887033+00:00"} {"global_step": 2201, "acc_step": 0, "speed/wps": 12843.041054700056, "speed/FLOPS": 201717607951871.44, "speed/curr_iter_time": 1.2882, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20414644479751587, "optim/lr": 0.00132, "optim/total_tokens": 576978944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.697831392288208, "created_at": "2025-01-14T16:06:01.996206+00:00"} {"global_step": 2202, "acc_step": 0, "speed/wps": 12846.508586933192, "speed/FLOPS": 201772070310481.47, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17571677267551422, "optim/lr": 0.0013206, "optim/total_tokens": 577241088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 3.785125732421875, "created_at": "2025-01-14T16:06:07.103695+00:00"} {"global_step": 2203, "acc_step": 0, "speed/wps": 12849.836434908377, "speed/FLOPS": 201824338735874.06, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.18083889782428741, "optim/lr": 0.0013212, "optim/total_tokens": 577503232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 3.7538180351257324, "created_at": "2025-01-14T16:06:12.212998+00:00"} {"global_step": 2204, "acc_step": 0, "speed/wps": 12841.95227880299, "speed/FLOPS": 201700507230273.16, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18735070526599884, "optim/lr": 0.0013218, "optim/total_tokens": 577765376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469353, "loss/out": 3.6925079822540283, "created_at": "2025-01-14T16:06:17.321809+00:00"} {"global_step": 2205, "acc_step": 0, "speed/wps": 12848.403400073215, "speed/FLOPS": 201801830954591.62, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.19367074966430664, "optim/lr": 0.0013224, "optim/total_tokens": 578027520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350820, "loss/out": 3.801194906234741, "created_at": "2025-01-14T16:06:22.426142+00:00"} {"global_step": 2206, "acc_step": 0, "speed/wps": 12855.733369867336, "speed/FLOPS": 201916958202642.3, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18998368084430695, "optim/lr": 0.001323, "optim/total_tokens": 578289664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 3.8165178298950195, "created_at": "2025-01-14T16:06:27.530462+00:00"} {"global_step": 2207, "acc_step": 0, "speed/wps": 12853.186859668767, "speed/FLOPS": 201876961760702.1, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18184787034988403, "optim/lr": 0.0013235999999999999, "optim/total_tokens": 578551808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433167, "loss/out": 3.722193479537964, "created_at": "2025-01-14T16:06:32.633162+00:00"} {"global_step": 2208, "acc_step": 0, "speed/wps": 12848.476713737036, "speed/FLOPS": 201802982446425.47, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17883329093456268, "optim/lr": 0.0013242, "optim/total_tokens": 578813952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 3.881610870361328, "created_at": "2025-01-14T16:06:37.739298+00:00"} {"global_step": 2209, "acc_step": 0, "speed/wps": 12855.485289262835, "speed/FLOPS": 201913061755850.16, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16082707047462463, "optim/lr": 0.0013248000000000001, "optim/total_tokens": 579076096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450471, "loss/out": 3.71933913230896, "created_at": "2025-01-14T16:06:42.841026+00:00"} {"global_step": 2210, "acc_step": 0, "speed/wps": 12854.112019393015, "speed/FLOPS": 201891492665474.06, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.158547505736351, "optim/lr": 0.0013254, "optim/total_tokens": 579338240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469875, "loss/out": 3.710484743118286, "created_at": "2025-01-14T16:06:47.949496+00:00"} {"global_step": 2211, "acc_step": 0, "speed/wps": 12850.305960017075, "speed/FLOPS": 201831713272901.28, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1628720611333847, "optim/lr": 0.0013260000000000001, "optim/total_tokens": 579600384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424340, "loss/out": 3.8248729705810547, "created_at": "2025-01-14T16:06:53.054714+00:00"} {"global_step": 2212, "acc_step": 0, "speed/wps": 12848.029499561282, "speed/FLOPS": 201795958333258.6, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20609818398952484, "optim/lr": 0.0013266, "optim/total_tokens": 579862528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.768782138824463, "created_at": "2025-01-14T16:06:58.163503+00:00"} {"global_step": 2213, "acc_step": 0, "speed/wps": 12849.386575637098, "speed/FLOPS": 201817273077847.22, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.252567857503891, "optim/lr": 0.0013272000000000002, "optim/total_tokens": 580124672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 3.5173802375793457, "created_at": "2025-01-14T16:07:03.272533+00:00"} {"global_step": 2214, "acc_step": 0, "speed/wps": 12850.289146337669, "speed/FLOPS": 201831449190961.8, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3050445020198822, "optim/lr": 0.0013278, "optim/total_tokens": 580386816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 3.6798176765441895, "created_at": "2025-01-14T16:07:08.384020+00:00"} {"global_step": 2215, "acc_step": 0, "speed/wps": 12848.82890593345, "speed/FLOPS": 201808514108832.94, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23677441477775574, "optim/lr": 0.0013284000000000002, "optim/total_tokens": 580648960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469353, "loss/out": 3.753883123397827, "created_at": "2025-01-14T16:07:13.490232+00:00"} {"global_step": 2216, "acc_step": 0, "speed/wps": 12850.020680194579, "speed/FLOPS": 201827232561273.44, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21379831433296204, "optim/lr": 0.001329, "optim/total_tokens": 580911104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349184, "loss/out": 3.7052001953125, "created_at": "2025-01-14T16:07:18.595402+00:00"} {"global_step": 2217, "acc_step": 0, "speed/wps": 12850.994588976184, "speed/FLOPS": 201842529136979.44, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19663305580615997, "optim/lr": 0.0013296, "optim/total_tokens": 581173248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.716902732849121, "created_at": "2025-01-14T16:07:23.698630+00:00"} {"global_step": 2218, "acc_step": 0, "speed/wps": 12845.701524512257, "speed/FLOPS": 201759394286140.7, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20646175742149353, "optim/lr": 0.0013302000000000001, "optim/total_tokens": 581435392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.725332498550415, "created_at": "2025-01-14T16:07:28.804044+00:00"} {"global_step": 2219, "acc_step": 0, "speed/wps": 12835.300756606894, "speed/FLOPS": 201596035934034.16, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.1795610785484314, "optim/lr": 0.0013308, "optim/total_tokens": 581697536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 3.747864246368408, "created_at": "2025-01-14T16:07:33.919589+00:00"} {"global_step": 2220, "acc_step": 0, "speed/wps": 12841.953614651478, "speed/FLOPS": 201700528211609.3, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.001, "optim/grad_norm": 0.21810993552207947, "optim/lr": 0.0013314, "optim/total_tokens": 581959680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 3.806292772293091, "created_at": "2025-01-14T16:07:39.027948+00:00"} {"global_step": 2221, "acc_step": 0, "speed/wps": 12843.989034631146, "speed/FLOPS": 201732497279350.16, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18798606097698212, "optim/lr": 0.001332, "optim/total_tokens": 582221824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329565, "loss/out": 3.898542642593384, "created_at": "2025-01-14T16:07:44.135529+00:00"} {"global_step": 2222, "acc_step": 0, "speed/wps": 12838.83941008147, "speed/FLOPS": 201651615349471.8, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.001, "optim/grad_norm": 0.177215114235878, "optim/lr": 0.0013326, "optim/total_tokens": 582483968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447263, "loss/out": 3.682384490966797, "created_at": "2025-01-14T16:07:49.245431+00:00"} {"global_step": 2223, "acc_step": 0, "speed/wps": 12847.484864924252, "speed/FLOPS": 201787404097877.6, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.172418013215065, "optim/lr": 0.0013332, "optim/total_tokens": 582746112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300225, "loss/out": 3.7060537338256836, "created_at": "2025-01-14T16:07:54.352162+00:00"} {"global_step": 2224, "acc_step": 0, "speed/wps": 12851.656170521957, "speed/FLOPS": 201852920184264.12, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1819399893283844, "optim/lr": 0.0013338, "optim/total_tokens": 583008256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 3.7708261013031006, "created_at": "2025-01-14T16:07:59.455725+00:00"} {"global_step": 2225, "acc_step": 0, "speed/wps": 12850.825309661428, "speed/FLOPS": 201839870372727.03, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22288841009140015, "optim/lr": 0.0013344, "optim/total_tokens": 583270400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310208, "loss/out": 3.7755417823791504, "created_at": "2025-01-14T16:08:04.560809+00:00"} {"global_step": 2226, "acc_step": 0, "speed/wps": 12849.164994146873, "speed/FLOPS": 201813792836058.38, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17048907279968262, "optim/lr": 0.001335, "optim/total_tokens": 583532544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468540, "loss/out": 3.7178220748901367, "created_at": "2025-01-14T16:08:09.665372+00:00"} {"global_step": 2227, "acc_step": 0, "speed/wps": 12846.921304429907, "speed/FLOPS": 201778552605900.62, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2239496409893036, "optim/lr": 0.0013356, "optim/total_tokens": 583794688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483949, "loss/out": 3.696016311645508, "created_at": "2025-01-14T16:08:14.771990+00:00"} {"global_step": 2228, "acc_step": 0, "speed/wps": 12847.655946859357, "speed/FLOPS": 201790091174759.62, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20883503556251526, "optim/lr": 0.0013362, "optim/total_tokens": 584056832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.7376747131347656, "created_at": "2025-01-14T16:08:19.876855+00:00"} {"global_step": 2229, "acc_step": 0, "speed/wps": 12851.324728152244, "speed/FLOPS": 201847714426398.22, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21007230877876282, "optim/lr": 0.0013368, "optim/total_tokens": 584318976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.717353343963623, "created_at": "2025-01-14T16:08:24.981788+00:00"} {"global_step": 2230, "acc_step": 0, "speed/wps": 12851.443347420505, "speed/FLOPS": 201849577505006.22, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16495563089847565, "optim/lr": 0.0013373999999999999, "optim/total_tokens": 584581120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 3.7202277183532715, "created_at": "2025-01-14T16:08:30.091403+00:00"} {"global_step": 2231, "acc_step": 0, "speed/wps": 12851.978365368263, "speed/FLOPS": 201857980697067.47, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17815876007080078, "optim/lr": 0.001338, "optim/total_tokens": 584843264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 3.743929386138916, "created_at": "2025-01-14T16:08:35.198574+00:00"} {"global_step": 2232, "acc_step": 0, "speed/wps": 12852.140428900842, "speed/FLOPS": 201860526127545.25, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18727707862854004, "optim/lr": 0.0013386, "optim/total_tokens": 585105408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318166, "loss/out": 3.7075119018554688, "created_at": "2025-01-14T16:08:40.301735+00:00"} {"global_step": 2233, "acc_step": 0, "speed/wps": 12851.434989548752, "speed/FLOPS": 201849446233145.44, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.001, "optim/grad_norm": 0.25589364767074585, "optim/lr": 0.0013392, "optim/total_tokens": 585367552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461385, "loss/out": 3.6293182373046875, "created_at": "2025-01-14T16:08:45.406655+00:00"} {"global_step": 2234, "acc_step": 0, "speed/wps": 12851.726819726857, "speed/FLOPS": 201854029827109.25, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1798027902841568, "optim/lr": 0.0013398, "optim/total_tokens": 585629696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 386688, "loss/out": 3.7274363040924072, "created_at": "2025-01-14T16:08:50.515208+00:00"} {"global_step": 2235, "acc_step": 0, "speed/wps": 12857.209206988076, "speed/FLOPS": 201940138252636.28, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20792442560195923, "optim/lr": 0.0013404, "optim/total_tokens": 585891840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.7990262508392334, "created_at": "2025-01-14T16:08:55.617560+00:00"} {"global_step": 2236, "acc_step": 0, "speed/wps": 12855.836200593407, "speed/FLOPS": 201918573300499.47, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19513411819934845, "optim/lr": 0.001341, "optim/total_tokens": 586153984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322976, "loss/out": 3.791994571685791, "created_at": "2025-01-14T16:09:00.720507+00:00"} {"global_step": 2237, "acc_step": 0, "speed/wps": 12851.619966039401, "speed/FLOPS": 201852351543115.4, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19586487114429474, "optim/lr": 0.0013416, "optim/total_tokens": 586416128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450214, "loss/out": 3.5743422508239746, "created_at": "2025-01-14T16:09:05.828748+00:00"} {"global_step": 2238, "acc_step": 0, "speed/wps": 12848.55883513631, "speed/FLOPS": 201804272275847.7, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19465424120426178, "optim/lr": 0.0013422, "optim/total_tokens": 586678272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482092, "loss/out": 3.659942388534546, "created_at": "2025-01-14T16:09:10.941209+00:00"} {"global_step": 2239, "acc_step": 0, "speed/wps": 12847.295820650608, "speed/FLOPS": 201784434897785.97, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1802370399236679, "optim/lr": 0.0013428000000000001, "optim/total_tokens": 586940416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473241, "loss/out": 3.818808078765869, "created_at": "2025-01-14T16:09:16.050223+00:00"} {"global_step": 2240, "acc_step": 0, "speed/wps": 12849.642701358616, "speed/FLOPS": 201821295884257.28, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17887113988399506, "optim/lr": 0.0013434, "optim/total_tokens": 587202560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.6281073093414307, "created_at": "2025-01-14T16:09:21.159047+00:00"} {"global_step": 2241, "acc_step": 0, "speed/wps": 12851.108226642486, "speed/FLOPS": 201844313972683.22, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.173277348279953, "optim/lr": 0.0013440000000000001, "optim/total_tokens": 587464704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314942, "loss/out": 3.5932555198669434, "created_at": "2025-01-14T16:09:26.262349+00:00"} {"global_step": 2242, "acc_step": 0, "speed/wps": 12849.127055769794, "speed/FLOPS": 201813196961713.53, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8616575598716736, "optim/lr": 0.0013446, "optim/total_tokens": 587726848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463485, "loss/out": 3.6886539459228516, "created_at": "2025-01-14T16:09:31.367500+00:00"} {"global_step": 2243, "acc_step": 0, "speed/wps": 12846.285967279784, "speed/FLOPS": 201768573762913.22, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1809411495923996, "optim/lr": 0.0013452000000000002, "optim/total_tokens": 587988992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363200, "loss/out": 3.659695863723755, "created_at": "2025-01-14T16:09:36.472783+00:00"} {"global_step": 2244, "acc_step": 0, "speed/wps": 12846.95103019454, "speed/FLOPS": 201779019489881.7, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18022622168064117, "optim/lr": 0.0013458, "optim/total_tokens": 588251136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.7687010765075684, "created_at": "2025-01-14T16:09:41.581657+00:00"} {"global_step": 2245, "acc_step": 0, "speed/wps": 12850.169414970112, "speed/FLOPS": 201829568645306.2, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.20079892873764038, "optim/lr": 0.0013464, "optim/total_tokens": 588513280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.746608257293701, "created_at": "2025-01-14T16:09:46.691362+00:00"} {"global_step": 2246, "acc_step": 0, "speed/wps": 12846.251378377718, "speed/FLOPS": 201768030496676.16, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18294395506381989, "optim/lr": 0.001347, "optim/total_tokens": 588775424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 3.7341737747192383, "created_at": "2025-01-14T16:09:51.797651+00:00"} {"global_step": 2247, "acc_step": 0, "speed/wps": 12852.09237252822, "speed/FLOPS": 201859771336176.84, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1876072734594345, "optim/lr": 0.0013476, "optim/total_tokens": 589037568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 3.7477495670318604, "created_at": "2025-01-14T16:09:56.902555+00:00"} {"global_step": 2248, "acc_step": 0, "speed/wps": 12852.99901152646, "speed/FLOPS": 201874011347496.75, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17124472558498383, "optim/lr": 0.0013482000000000001, "optim/total_tokens": 589299712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323502, "loss/out": 3.7092249393463135, "created_at": "2025-01-14T16:10:02.009823+00:00"} {"global_step": 2249, "acc_step": 0, "speed/wps": 12845.533948662123, "speed/FLOPS": 201756762277141.0, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1687486171722412, "optim/lr": 0.0013488, "optim/total_tokens": 589561856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 3.748434066772461, "created_at": "2025-01-14T16:10:07.118532+00:00"} {"global_step": 2250, "acc_step": 0, "speed/wps": 12850.609660422795, "speed/FLOPS": 201836483305101.22, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16553477942943573, "optim/lr": 0.0013494, "optim/total_tokens": 589824000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472260, "loss/out": 3.687375068664551, "created_at": "2025-01-14T16:10:12.222048+00:00"} {"global_step": 2251, "acc_step": 0, "speed/wps": 12844.120836957096, "speed/FLOPS": 201734567415994.56, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.306326299905777, "optim/lr": 0.00135, "optim/total_tokens": 590086144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478574, "loss/out": 3.77425479888916, "created_at": "2025-01-14T16:10:17.337624+00:00"} {"global_step": 2252, "acc_step": 0, "speed/wps": 12852.269438553923, "speed/FLOPS": 201862552401424.75, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17842231690883636, "optim/lr": 0.0013506, "optim/total_tokens": 590348288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334076, "loss/out": 3.679896116256714, "created_at": "2025-01-14T16:10:22.448525+00:00"} {"global_step": 2253, "acc_step": 0, "speed/wps": 12851.7577248907, "speed/FLOPS": 201854515235175.56, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.178252175450325, "optim/lr": 0.0013512, "optim/total_tokens": 590610432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.6741628646850586, "created_at": "2025-01-14T16:10:27.555233+00:00"} {"global_step": 2254, "acc_step": 0, "speed/wps": 12853.169158066034, "speed/FLOPS": 201876683732706.6, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18881359696388245, "optim/lr": 0.0013518, "optim/total_tokens": 590872576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.7274186611175537, "created_at": "2025-01-14T16:10:32.657774+00:00"} {"global_step": 2255, "acc_step": 0, "speed/wps": 12854.07370194226, "speed/FLOPS": 201890890837256.28, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20052838325500488, "optim/lr": 0.0013524, "optim/total_tokens": 591134720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464271, "loss/out": 3.7194111347198486, "created_at": "2025-01-14T16:10:37.759781+00:00"} {"global_step": 2256, "acc_step": 0, "speed/wps": 12848.375687439988, "speed/FLOPS": 201801395689605.66, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21838250756263733, "optim/lr": 0.001353, "optim/total_tokens": 591396864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.7365241050720215, "created_at": "2025-01-14T16:10:42.864320+00:00"} {"global_step": 2257, "acc_step": 0, "speed/wps": 12851.73712377879, "speed/FLOPS": 201854191666404.12, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20928849279880524, "optim/lr": 0.0013536, "optim/total_tokens": 591659008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.7804622650146484, "created_at": "2025-01-14T16:10:47.972140+00:00"} {"global_step": 2258, "acc_step": 0, "speed/wps": 12857.469607864883, "speed/FLOPS": 201944228206234.8, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2417166829109192, "optim/lr": 0.0013542, "optim/total_tokens": 591921152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.688274383544922, "created_at": "2025-01-14T16:10:53.073063+00:00"} {"global_step": 2259, "acc_step": 0, "speed/wps": 12852.112122707904, "speed/FLOPS": 201860081539888.38, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21105770766735077, "optim/lr": 0.0013548, "optim/total_tokens": 592183296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320571, "loss/out": 3.862440824508667, "created_at": "2025-01-14T16:10:58.183428+00:00"} {"global_step": 2260, "acc_step": 0, "speed/wps": 12850.237168528161, "speed/FLOPS": 201830632808038.5, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1966339349746704, "optim/lr": 0.0013553999999999999, "optim/total_tokens": 592445440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 3.7753050327301025, "created_at": "2025-01-14T16:11:03.287473+00:00"} {"global_step": 2261, "acc_step": 0, "speed/wps": 12854.090450389935, "speed/FLOPS": 201891153894641.78, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20317737758159637, "optim/lr": 0.001356, "optim/total_tokens": 592707584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402220, "loss/out": 3.526531219482422, "created_at": "2025-01-14T16:11:08.392382+00:00"} {"global_step": 2262, "acc_step": 0, "speed/wps": 12851.563607156178, "speed/FLOPS": 201851466349409.2, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17515744268894196, "optim/lr": 0.0013566, "optim/total_tokens": 592969728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460578, "loss/out": 3.7311294078826904, "created_at": "2025-01-14T16:11:13.497935+00:00"} {"global_step": 2263, "acc_step": 0, "speed/wps": 12851.220199964282, "speed/FLOPS": 201846072667569.5, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18863821029663086, "optim/lr": 0.0013572, "optim/total_tokens": 593231872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296475, "loss/out": 3.77109956741333, "created_at": "2025-01-14T16:11:18.601650+00:00"} {"global_step": 2264, "acc_step": 0, "speed/wps": 12850.841516912544, "speed/FLOPS": 201840124929876.9, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16685757040977478, "optim/lr": 0.0013578, "optim/total_tokens": 593494016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464535, "loss/out": 3.746227741241455, "created_at": "2025-01-14T16:11:23.708001+00:00"} {"global_step": 2265, "acc_step": 0, "speed/wps": 12850.75340778551, "speed/FLOPS": 201838741054962.22, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18184441328048706, "optim/lr": 0.0013584, "optim/total_tokens": 593756160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355325, "loss/out": 3.689788341522217, "created_at": "2025-01-14T16:11:28.840221+00:00"} {"global_step": 2266, "acc_step": 0, "speed/wps": 12850.458188590823, "speed/FLOPS": 201834104231835.97, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1704631745815277, "optim/lr": 0.001359, "optim/total_tokens": 594018304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 3.699087619781494, "created_at": "2025-01-14T16:11:33.944840+00:00"} {"global_step": 2267, "acc_step": 0, "speed/wps": 12851.682176849901, "speed/FLOPS": 201853328649380.94, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1997620165348053, "optim/lr": 0.0013596, "optim/total_tokens": 594280448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 3.792039394378662, "created_at": "2025-01-14T16:11:39.048395+00:00"} {"global_step": 2268, "acc_step": 0, "speed/wps": 12844.119071635307, "speed/FLOPS": 201734539689190.06, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.16921639442443848, "optim/lr": 0.0013602000000000002, "optim/total_tokens": 594542592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299469, "loss/out": 3.6806623935699463, "created_at": "2025-01-14T16:11:44.155618+00:00"} {"global_step": 2269, "acc_step": 0, "speed/wps": 12854.898175127903, "speed/FLOPS": 201903840321581.66, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1693093180656433, "optim/lr": 0.0013608000000000001, "optim/total_tokens": 594804736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452081, "loss/out": 3.762852668762207, "created_at": "2025-01-14T16:11:49.257573+00:00"} {"global_step": 2270, "acc_step": 0, "speed/wps": 12853.650447726573, "speed/FLOPS": 201884243048186.7, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19188973307609558, "optim/lr": 0.0013614, "optim/total_tokens": 595066880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364158, "loss/out": 3.7134292125701904, "created_at": "2025-01-14T16:11:54.361851+00:00"} {"global_step": 2271, "acc_step": 0, "speed/wps": 12852.288506193772, "speed/FLOPS": 201862851884910.25, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18308936059474945, "optim/lr": 0.0013620000000000001, "optim/total_tokens": 595329024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 3.719458818435669, "created_at": "2025-01-14T16:11:59.465583+00:00"} {"global_step": 2272, "acc_step": 0, "speed/wps": 12842.18675024902, "speed/FLOPS": 201704189926530.6, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15815122425556183, "optim/lr": 0.0013626, "optim/total_tokens": 595591168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 3.697300434112549, "created_at": "2025-01-14T16:12:04.577158+00:00"} {"global_step": 2273, "acc_step": 0, "speed/wps": 12850.153033420154, "speed/FLOPS": 201829311350553.47, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1750333309173584, "optim/lr": 0.0013632000000000002, "optim/total_tokens": 595853312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465057, "loss/out": 3.6021153926849365, "created_at": "2025-01-14T16:12:09.685879+00:00"} {"global_step": 2274, "acc_step": 0, "speed/wps": 12836.46843298682, "speed/FLOPS": 201614375896136.38, "speed/curr_iter_time": 1.2875, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.16446955502033234, "optim/lr": 0.0013638, "optim/total_tokens": 596115456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456330, "loss/out": 3.6250998973846436, "created_at": "2025-01-14T16:12:14.797706+00:00"} {"global_step": 2275, "acc_step": 0, "speed/wps": 12841.778798974057, "speed/FLOPS": 201697782491173.5, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20466116070747375, "optim/lr": 0.0013644, "optim/total_tokens": 596377600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287957, "loss/out": 3.727595806121826, "created_at": "2025-01-14T16:12:19.907107+00:00"} {"global_step": 2276, "acc_step": 0, "speed/wps": 12849.649238436152, "speed/FLOPS": 201821398558042.88, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18364185094833374, "optim/lr": 0.0013650000000000001, "optim/total_tokens": 596639744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 3.7672476768493652, "created_at": "2025-01-14T16:12:25.014379+00:00"} {"global_step": 2277, "acc_step": 0, "speed/wps": 12854.712134429157, "speed/FLOPS": 201900918296761.16, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1745261549949646, "optim/lr": 0.0013656, "optim/total_tokens": 596901888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294022, "loss/out": 3.678290605545044, "created_at": "2025-01-14T16:12:30.117298+00:00"} {"global_step": 2278, "acc_step": 0, "speed/wps": 12849.46474805061, "speed/FLOPS": 201818500883021.56, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16757284104824066, "optim/lr": 0.0013662000000000001, "optim/total_tokens": 597164032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460815, "loss/out": 3.647265672683716, "created_at": "2025-01-14T16:12:35.224819+00:00"} {"global_step": 2279, "acc_step": 0, "speed/wps": 12841.172153262803, "speed/FLOPS": 201688254286659.7, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20476293563842773, "optim/lr": 0.0013668, "optim/total_tokens": 597426176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419060, "loss/out": 3.7119333744049072, "created_at": "2025-01-14T16:12:40.336260+00:00"} {"global_step": 2280, "acc_step": 0, "speed/wps": 12843.925512788526, "speed/FLOPS": 201731499581523.66, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19409584999084473, "optim/lr": 0.0013674, "optim/total_tokens": 597688320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.6501731872558594, "created_at": "2025-01-14T16:12:45.445229+00:00"} {"global_step": 2281, "acc_step": 0, "speed/wps": 12840.57534491936, "speed/FLOPS": 201678880591525.3, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.19279174506664276, "optim/lr": 0.001368, "optim/total_tokens": 597950464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484902, "loss/out": 3.68735408782959, "created_at": "2025-01-14T16:12:50.555773+00:00"} {"global_step": 2282, "acc_step": 0, "speed/wps": 12850.260620676021, "speed/FLOPS": 201831001156247.25, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1679277867078781, "optim/lr": 0.0013686, "optim/total_tokens": 598212608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327402, "loss/out": 3.7407310009002686, "created_at": "2025-01-14T16:12:55.663891+00:00"} {"global_step": 2283, "acc_step": 0, "speed/wps": 12846.820105580202, "speed/FLOPS": 201776963138904.22, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16444209218025208, "optim/lr": 0.0013691999999999999, "optim/total_tokens": 598474752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464799, "loss/out": 3.6905837059020996, "created_at": "2025-01-14T16:13:00.769622+00:00"} {"global_step": 2284, "acc_step": 0, "speed/wps": 12846.346716996288, "speed/FLOPS": 201769527920689.78, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16515913605690002, "optim/lr": 0.0013698, "optim/total_tokens": 598736896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364445, "loss/out": 3.651582956314087, "created_at": "2025-01-14T16:13:05.878184+00:00"} {"global_step": 2285, "acc_step": 0, "speed/wps": 12847.183385620934, "speed/FLOPS": 201782668951143.03, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1572452038526535, "optim/lr": 0.0013704, "optim/total_tokens": 598999040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.703709602355957, "created_at": "2025-01-14T16:13:10.990573+00:00"} {"global_step": 2286, "acc_step": 0, "speed/wps": 12849.097803252049, "speed/FLOPS": 201812737510725.2, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16519469022750854, "optim/lr": 0.001371, "optim/total_tokens": 599261184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.735285997390747, "created_at": "2025-01-14T16:13:16.097228+00:00"} {"global_step": 2287, "acc_step": 0, "speed/wps": 12848.667810364217, "speed/FLOPS": 201805983881549.62, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16519616544246674, "optim/lr": 0.0013716, "optim/total_tokens": 599523328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 3.780165433883667, "created_at": "2025-01-14T16:13:21.202877+00:00"} {"global_step": 2288, "acc_step": 0, "speed/wps": 12851.118933376269, "speed/FLOPS": 201844482136649.88, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16987460851669312, "optim/lr": 0.0013721999999999999, "optim/total_tokens": 599785472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.657518148422241, "created_at": "2025-01-14T16:13:26.307881+00:00"} {"global_step": 2289, "acc_step": 0, "speed/wps": 12849.223094026496, "speed/FLOPS": 201814705374505.44, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17493046820163727, "optim/lr": 0.0013728, "optim/total_tokens": 600047616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 3.7128915786743164, "created_at": "2025-01-14T16:13:31.413825+00:00"} {"global_step": 2290, "acc_step": 0, "speed/wps": 12855.690491140718, "speed/FLOPS": 201916284733319.25, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1710362285375595, "optim/lr": 0.0013733999999999999, "optim/total_tokens": 600309760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 3.7004590034484863, "created_at": "2025-01-14T16:13:36.519500+00:00"} {"global_step": 2291, "acc_step": 0, "speed/wps": 12852.526054809912, "speed/FLOPS": 201866582912352.72, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.001, "optim/grad_norm": 0.19264131784439087, "optim/lr": 0.001374, "optim/total_tokens": 600571904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.7184550762176514, "created_at": "2025-01-14T16:13:41.628439+00:00"} {"global_step": 2292, "acc_step": 0, "speed/wps": 12844.674241196906, "speed/FLOPS": 201743259390037.9, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20639532804489136, "optim/lr": 0.0013746, "optim/total_tokens": 600834048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 3.8130340576171875, "created_at": "2025-01-14T16:13:46.739196+00:00"} {"global_step": 2293, "acc_step": 0, "speed/wps": 12848.628312466259, "speed/FLOPS": 201805363512786.4, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.20283819735050201, "optim/lr": 0.0013752, "optim/total_tokens": 601096192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302050, "loss/out": 3.67508602142334, "created_at": "2025-01-14T16:13:51.846988+00:00"} {"global_step": 2294, "acc_step": 0, "speed/wps": 12847.528343360866, "speed/FLOPS": 201788086986470.25, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.20951035618782043, "optim/lr": 0.0013758000000000002, "optim/total_tokens": 601358336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.7616219520568848, "created_at": "2025-01-14T16:13:56.956182+00:00"} {"global_step": 2295, "acc_step": 0, "speed/wps": 12850.572194596076, "speed/FLOPS": 201835894852809.16, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1914716362953186, "optim/lr": 0.0013764, "optim/total_tokens": 601620480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 3.7741177082061768, "created_at": "2025-01-14T16:14:02.065356+00:00"} {"global_step": 2296, "acc_step": 0, "speed/wps": 12843.638950255452, "speed/FLOPS": 201726998723163.53, "speed/curr_iter_time": 1.2873, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17702490091323853, "optim/lr": 0.0013770000000000002, "optim/total_tokens": 601882624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477242, "loss/out": 3.6518638134002686, "created_at": "2025-01-14T16:14:07.176169+00:00"} {"global_step": 2297, "acc_step": 0, "speed/wps": 12854.238950662198, "speed/FLOPS": 201893486295478.44, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20767822861671448, "optim/lr": 0.0013776, "optim/total_tokens": 602144768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.5820107460021973, "created_at": "2025-01-14T16:14:12.280581+00:00"} {"global_step": 2298, "acc_step": 0, "speed/wps": 12845.91050391151, "speed/FLOPS": 201762676594773.75, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20443284511566162, "optim/lr": 0.0013782, "optim/total_tokens": 602406912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364976, "loss/out": 3.710087299346924, "created_at": "2025-01-14T16:14:17.390539+00:00"} {"global_step": 2299, "acc_step": 0, "speed/wps": 12853.666664452381, "speed/FLOPS": 201884497754149.6, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.19426921010017395, "optim/lr": 0.0013788000000000001, "optim/total_tokens": 602669056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 3.679112672805786, "created_at": "2025-01-14T16:14:22.494542+00:00"} {"global_step": 2300, "acc_step": 0, "speed/wps": 12851.39850466724, "speed/FLOPS": 201848873187945.8, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18034975230693817, "optim/lr": 0.0013794, "optim/total_tokens": 602931200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477030, "loss/out": 3.776301622390747, "created_at": "2025-01-14T16:14:27.602311+00:00"} {"global_step": 2301, "acc_step": 0, "speed/wps": 12857.474158384457, "speed/FLOPS": 201944299678397.8, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20181962847709656, "optim/lr": 0.0013800000000000002, "optim/total_tokens": 603193344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 3.73923921585083, "created_at": "2025-01-14T16:14:32.706038+00:00"} {"global_step": 2302, "acc_step": 0, "speed/wps": 12850.05816861757, "speed/FLOPS": 201827821368471.0, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1549251824617386, "optim/lr": 0.0013806, "optim/total_tokens": 603455488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357724, "loss/out": 3.774709939956665, "created_at": "2025-01-14T16:14:37.812012+00:00"} {"global_step": 2303, "acc_step": 0, "speed/wps": 12849.363760180282, "speed/FLOPS": 201816914729746.2, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18512344360351562, "optim/lr": 0.0013812, "optim/total_tokens": 603717632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 3.6431453227996826, "created_at": "2025-01-14T16:14:42.919344+00:00"} {"global_step": 2304, "acc_step": 0, "speed/wps": 12851.875467251539, "speed/FLOPS": 201856364540747.5, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1805448830127716, "optim/lr": 0.0013818, "optim/total_tokens": 603979776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431521, "loss/out": 3.7574381828308105, "created_at": "2025-01-14T16:14:48.024330+00:00"} {"global_step": 2305, "acc_step": 0, "speed/wps": 12849.704315807563, "speed/FLOPS": 201822263623845.12, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17202389240264893, "optim/lr": 0.0013824, "optim/total_tokens": 604241920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.700460910797119, "created_at": "2025-01-14T16:14:53.128691+00:00"} {"global_step": 2306, "acc_step": 0, "speed/wps": 12847.804655046344, "speed/FLOPS": 201792426841182.75, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16599568724632263, "optim/lr": 0.0013830000000000001, "optim/total_tokens": 604504064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 3.785665988922119, "created_at": "2025-01-14T16:14:58.233970+00:00"} {"global_step": 2307, "acc_step": 0, "speed/wps": 12846.329843320287, "speed/FLOPS": 201769262896421.34, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19136157631874084, "optim/lr": 0.0013836, "optim/total_tokens": 604766208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469590, "loss/out": 3.845062017440796, "created_at": "2025-01-14T16:15:03.340743+00:00"} {"global_step": 2308, "acc_step": 0, "speed/wps": 12848.154541954344, "speed/FLOPS": 201797922295867.88, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16369621455669403, "optim/lr": 0.0013842, "optim/total_tokens": 605028352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.527644395828247, "created_at": "2025-01-14T16:15:08.445743+00:00"} {"global_step": 2309, "acc_step": 0, "speed/wps": 12850.96482337439, "speed/FLOPS": 201842061627301.06, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.18386466801166534, "optim/lr": 0.0013848, "optim/total_tokens": 605290496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287957, "loss/out": 3.7492218017578125, "created_at": "2025-01-14T16:15:13.548875+00:00"} {"global_step": 2310, "acc_step": 0, "speed/wps": 12850.600569102207, "speed/FLOPS": 201836340513418.6, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18595558404922485, "optim/lr": 0.0013854, "optim/total_tokens": 605552640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 3.755615711212158, "created_at": "2025-01-14T16:15:18.658333+00:00"} {"global_step": 2311, "acc_step": 0, "speed/wps": 12850.07697392165, "speed/FLOPS": 201828116731612.78, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19461581110954285, "optim/lr": 0.001386, "optim/total_tokens": 605814784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421661, "loss/out": 3.659048557281494, "created_at": "2025-01-14T16:15:23.762082+00:00"} {"global_step": 2312, "acc_step": 0, "speed/wps": 12851.228834500376, "speed/FLOPS": 201846208284823.97, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19707472622394562, "optim/lr": 0.0013866, "optim/total_tokens": 606076928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481474, "loss/out": 3.73895525932312, "created_at": "2025-01-14T16:15:28.870459+00:00"} {"global_step": 2313, "acc_step": 0, "speed/wps": 12852.455927336476, "speed/FLOPS": 201865481464017.16, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22770796716213226, "optim/lr": 0.0013871999999999999, "optim/total_tokens": 606339072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.657461643218994, "created_at": "2025-01-14T16:15:33.973211+00:00"} {"global_step": 2314, "acc_step": 0, "speed/wps": 12849.980216596401, "speed/FLOPS": 201826597024861.22, "speed/curr_iter_time": 1.2869, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.226649671792984, "optim/lr": 0.0013878, "optim/total_tokens": 606601216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 3.7076640129089355, "created_at": "2025-01-14T16:15:39.077067+00:00"} {"global_step": 2315, "acc_step": 0, "speed/wps": 12848.377228721876, "speed/FLOPS": 201801419897555.66, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2067478746175766, "optim/lr": 0.0013884, "optim/total_tokens": 606863360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473347, "loss/out": 3.7472410202026367, "created_at": "2025-01-14T16:15:44.181480+00:00"} {"global_step": 2316, "acc_step": 0, "speed/wps": 12851.90655151965, "speed/FLOPS": 201856852761892.44, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17400561273097992, "optim/lr": 0.001389, "optim/total_tokens": 607125504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 3.7210564613342285, "created_at": "2025-01-14T16:15:49.286735+00:00"} {"global_step": 2317, "acc_step": 0, "speed/wps": 12848.333017303312, "speed/FLOPS": 201800725496474.25, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1650160700082779, "optim/lr": 0.0013896, "optim/total_tokens": 607387648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.6887450218200684, "created_at": "2025-01-14T16:15:54.391326+00:00"} {"global_step": 2318, "acc_step": 0, "speed/wps": 12853.230076200269, "speed/FLOPS": 201877640535715.06, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18826799094676971, "optim/lr": 0.0013902, "optim/total_tokens": 607649792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 3.7690064907073975, "created_at": "2025-01-14T16:15:59.493988+00:00"} {"global_step": 2319, "acc_step": 0, "speed/wps": 12847.693975068685, "speed/FLOPS": 201790688460043.38, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21326470375061035, "optim/lr": 0.0013908, "optim/total_tokens": 607911936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 3.766505241394043, "created_at": "2025-01-14T16:16:04.600443+00:00"} {"global_step": 2320, "acc_step": 0, "speed/wps": 12852.245854663515, "speed/FLOPS": 201862181984018.22, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21195152401924133, "optim/lr": 0.0013914, "optim/total_tokens": 608174080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473839, "loss/out": 3.8386433124542236, "created_at": "2025-01-14T16:16:09.708943+00:00"} {"global_step": 2321, "acc_step": 0, "speed/wps": 12847.739047654957, "speed/FLOPS": 201791396386947.0, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2031337320804596, "optim/lr": 0.001392, "optim/total_tokens": 608436224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 3.7126290798187256, "created_at": "2025-01-14T16:16:14.818238+00:00"} {"global_step": 2322, "acc_step": 0, "speed/wps": 12848.533922951714, "speed/FLOPS": 201803880995755.7, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17974568903446198, "optim/lr": 0.0013926000000000001, "optim/total_tokens": 608698368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310665, "loss/out": 3.6305696964263916, "created_at": "2025-01-14T16:16:19.922705+00:00"} {"global_step": 2323, "acc_step": 0, "speed/wps": 12853.818211932592, "speed/FLOPS": 201886878015574.38, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16643264889717102, "optim/lr": 0.0013932, "optim/total_tokens": 608960512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453945, "loss/out": 3.8875508308410645, "created_at": "2025-01-14T16:16:25.024970+00:00"} {"global_step": 2324, "acc_step": 0, "speed/wps": 12850.877468242912, "speed/FLOPS": 201840689594922.66, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17465461790561676, "optim/lr": 0.0013938000000000002, "optim/total_tokens": 609222656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 395808, "loss/out": 3.7192163467407227, "created_at": "2025-01-14T16:16:30.130787+00:00"} {"global_step": 2325, "acc_step": 0, "speed/wps": 12853.14284052133, "speed/FLOPS": 201876270379502.94, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15677915513515472, "optim/lr": 0.0013944, "optim/total_tokens": 609484800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 3.716932773590088, "created_at": "2025-01-14T16:16:35.239576+00:00"} {"global_step": 2326, "acc_step": 0, "speed/wps": 12847.888937683338, "speed/FLOPS": 201793750615813.2, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1473051905632019, "optim/lr": 0.0013950000000000002, "optim/total_tokens": 609746944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.64432692527771, "created_at": "2025-01-14T16:16:40.344082+00:00"} {"global_step": 2327, "acc_step": 0, "speed/wps": 12849.211660769553, "speed/FLOPS": 201814525799494.22, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16531004011631012, "optim/lr": 0.0013956, "optim/total_tokens": 610009088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.7109928131103516, "created_at": "2025-01-14T16:16:45.448042+00:00"} {"global_step": 2328, "acc_step": 0, "speed/wps": 12854.934186732933, "speed/FLOPS": 201904405933323.8, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1546061784029007, "optim/lr": 0.0013962, "optim/total_tokens": 610271232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 3.7747886180877686, "created_at": "2025-01-14T16:16:50.553276+00:00"} {"global_step": 2329, "acc_step": 0, "speed/wps": 12855.801611914201, "speed/FLOPS": 201918030037762.72, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15893875062465668, "optim/lr": 0.0013968000000000001, "optim/total_tokens": 610533376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.7511961460113525, "created_at": "2025-01-14T16:16:55.655801+00:00"} {"global_step": 2330, "acc_step": 0, "speed/wps": 12855.74761899865, "speed/FLOPS": 201917182004829.97, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15976694226264954, "optim/lr": 0.0013974, "optim/total_tokens": 610795520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443307, "loss/out": 3.7950639724731445, "created_at": "2025-01-14T16:17:00.757442+00:00"} {"global_step": 2331, "acc_step": 0, "speed/wps": 12849.422821502681, "speed/FLOPS": 201817842368973.25, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18175645172595978, "optim/lr": 0.0013980000000000002, "optim/total_tokens": 611057664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.7006640434265137, "created_at": "2025-01-14T16:17:05.863473+00:00"} {"global_step": 2332, "acc_step": 0, "speed/wps": 12852.641233697694, "speed/FLOPS": 201868391955057.72, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18299980461597443, "optim/lr": 0.0013986, "optim/total_tokens": 611319808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 3.72170090675354, "created_at": "2025-01-14T16:17:10.971761+00:00"} {"global_step": 2333, "acc_step": 0, "speed/wps": 12849.813892390255, "speed/FLOPS": 201823984674650.3, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16792969405651093, "optim/lr": 0.0013992, "optim/total_tokens": 611581952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464799, "loss/out": 3.823556661605835, "created_at": "2025-01-14T16:17:16.079803+00:00"} {"global_step": 2334, "acc_step": 0, "speed/wps": 12852.625122848824, "speed/FLOPS": 201868138912037.6, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1971767544746399, "optim/lr": 0.0013998, "optim/total_tokens": 611844096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479622, "loss/out": 3.7275540828704834, "created_at": "2025-01-14T16:17:21.188000+00:00"} {"global_step": 2335, "acc_step": 0, "speed/wps": 12858.320514713434, "speed/FLOPS": 201957592867559.84, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18873077630996704, "optim/lr": 0.0014004, "optim/total_tokens": 612106240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302869, "loss/out": 3.637510299682617, "created_at": "2025-01-14T16:17:26.288493+00:00"} {"global_step": 2336, "acc_step": 0, "speed/wps": 12855.43668180496, "speed/FLOPS": 201912298308930.6, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1920626014471054, "optim/lr": 0.0014010000000000001, "optim/total_tokens": 612368384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471738, "loss/out": 3.6740405559539795, "created_at": "2025-01-14T16:17:31.392288+00:00"} {"global_step": 2337, "acc_step": 0, "speed/wps": 12852.907088490058, "speed/FLOPS": 201872567569894.56, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20455902814865112, "optim/lr": 0.0014016, "optim/total_tokens": 612630528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348249, "loss/out": 3.751481771469116, "created_at": "2025-01-14T16:17:36.495626+00:00"} {"global_step": 2338, "acc_step": 0, "speed/wps": 12852.57320595421, "speed/FLOPS": 201867323485866.38, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1583845466375351, "optim/lr": 0.0014022, "optim/total_tokens": 612892672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.632227659225464, "created_at": "2025-01-14T16:17:41.602230+00:00"} {"global_step": 2339, "acc_step": 0, "speed/wps": 12854.859688608538, "speed/FLOPS": 201903235837909.03, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1930711567401886, "optim/lr": 0.0014028, "optim/total_tokens": 613154816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417484, "loss/out": 3.6929609775543213, "created_at": "2025-01-14T16:17:46.706876+00:00"} {"global_step": 2340, "acc_step": 0, "speed/wps": 12855.82801550516, "speed/FLOPS": 201918444742441.06, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18900373578071594, "optim/lr": 0.0014034, "optim/total_tokens": 613416960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.6000232696533203, "created_at": "2025-01-14T16:17:51.813412+00:00"} {"global_step": 2341, "acc_step": 0, "speed/wps": 12857.121792783657, "speed/FLOPS": 201938765292436.16, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.206788569688797, "optim/lr": 0.001404, "optim/total_tokens": 613679104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 389443, "loss/out": 3.7989578247070312, "created_at": "2025-01-14T16:17:56.914176+00:00"} {"global_step": 2342, "acc_step": 0, "speed/wps": 12852.838494515765, "speed/FLOPS": 201871490207270.28, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.19730974733829498, "optim/lr": 0.0014046, "optim/total_tokens": 613941248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451559, "loss/out": 3.6086487770080566, "created_at": "2025-01-14T16:18:02.022407+00:00"} {"global_step": 2343, "acc_step": 0, "speed/wps": 12859.423723849684, "speed/FLOPS": 201974920283012.88, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22381839156150818, "optim/lr": 0.0014052, "optim/total_tokens": 614203392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 347841, "loss/out": 3.6838231086730957, "created_at": "2025-01-14T16:18:07.122956+00:00"} {"global_step": 2344, "acc_step": 0, "speed/wps": 12849.74773183511, "speed/FLOPS": 201822945532217.62, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.001, "optim/grad_norm": 0.18630972504615784, "optim/lr": 0.0014058, "optim/total_tokens": 614465536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 3.7319071292877197, "created_at": "2025-01-14T16:18:12.227059+00:00"} {"global_step": 2345, "acc_step": 0, "speed/wps": 12847.33134636348, "speed/FLOPS": 201784992877924.06, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18451619148254395, "optim/lr": 0.0014064, "optim/total_tokens": 614727680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 3.7798972129821777, "created_at": "2025-01-14T16:18:17.332821+00:00"} {"global_step": 2346, "acc_step": 0, "speed/wps": 12854.344133778048, "speed/FLOPS": 201895138340849.6, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.19531206786632538, "optim/lr": 0.001407, "optim/total_tokens": 614989824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.6665148735046387, "created_at": "2025-01-14T16:18:22.434751+00:00"} {"global_step": 2347, "acc_step": 0, "speed/wps": 12858.920846549592, "speed/FLOPS": 201967021903989.1, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19533921778202057, "optim/lr": 0.0014076, "optim/total_tokens": 615251968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 3.7357177734375, "created_at": "2025-01-14T16:18:27.536136+00:00"} {"global_step": 2348, "acc_step": 0, "speed/wps": 12855.502194044437, "speed/FLOPS": 201913327268674.8, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23854295909404755, "optim/lr": 0.0014082, "optim/total_tokens": 615514112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 3.6404223442077637, "created_at": "2025-01-14T16:18:32.639390+00:00"} {"global_step": 2349, "acc_step": 0, "speed/wps": 12853.637769266676, "speed/FLOPS": 201884043915551.94, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.20397429168224335, "optim/lr": 0.0014088, "optim/total_tokens": 615776256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.691910743713379, "created_at": "2025-01-14T16:18:37.745459+00:00"} {"global_step": 2350, "acc_step": 0, "speed/wps": 12849.141621076937, "speed/FLOPS": 201813425729877.03, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.18392159044742584, "optim/lr": 0.0014094, "optim/total_tokens": 616038400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325819, "loss/out": 3.791898727416992, "created_at": "2025-01-14T16:18:42.850519+00:00"} {"global_step": 2351, "acc_step": 0, "speed/wps": 12854.506834632999, "speed/FLOPS": 201897693781351.03, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.19770672917366028, "optim/lr": 0.00141, "optim/total_tokens": 616300544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 3.7286839485168457, "created_at": "2025-01-14T16:18:47.955320+00:00"} {"global_step": 2352, "acc_step": 0, "speed/wps": 12854.86423728089, "speed/FLOPS": 201903307281058.84, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16716152429580688, "optim/lr": 0.0014106000000000001, "optim/total_tokens": 616562688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318319, "loss/out": 3.745600700378418, "created_at": "2025-01-14T16:18:53.057551+00:00"} {"global_step": 2353, "acc_step": 0, "speed/wps": 12851.22438932052, "speed/FLOPS": 201846138467166.0, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.19014757871627808, "optim/lr": 0.0014112, "optim/total_tokens": 616824832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462150, "loss/out": 3.7692806720733643, "created_at": "2025-01-14T16:18:58.162955+00:00"} {"global_step": 2354, "acc_step": 0, "speed/wps": 12853.694651137834, "speed/FLOPS": 201884937323502.38, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19395092129707336, "optim/lr": 0.0014118000000000002, "optim/total_tokens": 617086976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296535, "loss/out": 3.6516215801239014, "created_at": "2025-01-14T16:19:03.265179+00:00"} {"global_step": 2355, "acc_step": 0, "speed/wps": 12848.231610886545, "speed/FLOPS": 201799132769351.9, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.001, "optim/grad_norm": 0.22855713963508606, "optim/lr": 0.0014124, "optim/total_tokens": 617349120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481190, "loss/out": 3.6530840396881104, "created_at": "2025-01-14T16:19:08.369851+00:00"} {"global_step": 2356, "acc_step": 0, "speed/wps": 12861.411839642004, "speed/FLOPS": 202006146373488.62, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.27762141823768616, "optim/lr": 0.001413, "optim/total_tokens": 617611264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456556, "loss/out": 3.7190213203430176, "created_at": "2025-01-14T16:19:13.471339+00:00"} {"global_step": 2357, "acc_step": 0, "speed/wps": 12846.943611640272, "speed/FLOPS": 201778902971292.75, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.22262008488178253, "optim/lr": 0.0014136, "optim/total_tokens": 617873408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481190, "loss/out": 3.7197108268737793, "created_at": "2025-01-14T16:19:18.580043+00:00"} {"global_step": 2358, "acc_step": 0, "speed/wps": 12844.296752651835, "speed/FLOPS": 201737330413713.94, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20999060571193695, "optim/lr": 0.0014142, "optim/total_tokens": 618135552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356143, "loss/out": 3.6750926971435547, "created_at": "2025-01-14T16:19:23.687401+00:00"} {"global_step": 2359, "acc_step": 0, "speed/wps": 12842.851208298787, "speed/FLOPS": 201714626153263.22, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.17393755912780762, "optim/lr": 0.0014148000000000001, "optim/total_tokens": 618397696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 3.689889907836914, "created_at": "2025-01-14T16:19:28.797360+00:00"} {"global_step": 2360, "acc_step": 0, "speed/wps": 12853.035683916858, "speed/FLOPS": 201874587337781.1, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.17339183390140533, "optim/lr": 0.0014154, "optim/total_tokens": 618659840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.6461641788482666, "created_at": "2025-01-14T16:19:33.904764+00:00"} {"global_step": 2361, "acc_step": 0, "speed/wps": 12858.827668610445, "speed/FLOPS": 201965558416414.9, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.15216219425201416, "optim/lr": 0.001416, "optim/total_tokens": 618921984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478609, "loss/out": 3.773329973220825, "created_at": "2025-01-14T16:19:39.004997+00:00"} {"global_step": 2362, "acc_step": 0, "speed/wps": 12852.094848926075, "speed/FLOPS": 201859810231408.1, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16669809818267822, "optim/lr": 0.0014166, "optim/total_tokens": 619184128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.596851110458374, "created_at": "2025-01-14T16:19:44.107902+00:00"} {"global_step": 2363, "acc_step": 0, "speed/wps": 12854.87454166775, "speed/FLOPS": 201903469125614.2, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.17289306223392487, "optim/lr": 0.0014172, "optim/total_tokens": 619446272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325000, "loss/out": 3.6228270530700684, "created_at": "2025-01-14T16:19:49.214125+00:00"} {"global_step": 2364, "acc_step": 0, "speed/wps": 12857.943511150688, "speed/FLOPS": 201951671508549.75, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.16360054910182953, "optim/lr": 0.0014178, "optim/total_tokens": 619708416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463770, "loss/out": 3.8122739791870117, "created_at": "2025-01-14T16:19:54.319082+00:00"} {"global_step": 2365, "acc_step": 0, "speed/wps": 12854.75843979951, "speed/FLOPS": 201901645586232.8, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1523570567369461, "optim/lr": 0.0014184, "optim/total_tokens": 619970560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365740, "loss/out": 3.634669303894043, "created_at": "2025-01-14T16:19:59.421616+00:00"} {"global_step": 2366, "acc_step": 0, "speed/wps": 12855.554665379355, "speed/FLOPS": 201914151403092.88, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1712205857038498, "optim/lr": 0.001419, "optim/total_tokens": 620232704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 3.83546781539917, "created_at": "2025-01-14T16:20:04.530391+00:00"} {"global_step": 2367, "acc_step": 0, "speed/wps": 12853.339611970388, "speed/FLOPS": 201879360945502.7, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1880507618188858, "optim/lr": 0.0014196, "optim/total_tokens": 620494848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430991, "loss/out": 3.8033528327941895, "created_at": "2025-01-14T16:20:09.637355+00:00"} {"global_step": 2368, "acc_step": 0, "speed/wps": 12851.682507797532, "speed/FLOPS": 201853333847368.22, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.21318911015987396, "optim/lr": 0.0014202, "optim/total_tokens": 620756992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457380, "loss/out": 3.68936824798584, "created_at": "2025-01-14T16:20:14.744239+00:00"} {"global_step": 2369, "acc_step": 0, "speed/wps": 12851.588176676845, "speed/FLOPS": 201851852247493.3, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.20506234467029572, "optim/lr": 0.0014208, "optim/total_tokens": 621019136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358542, "loss/out": 3.736762523651123, "created_at": "2025-01-14T16:20:19.849358+00:00"} {"global_step": 2370, "acc_step": 0, "speed/wps": 12855.21226135637, "speed/FLOPS": 201908773477399.47, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21738147735595703, "optim/lr": 0.0014214, "optim/total_tokens": 621281280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461913, "loss/out": 3.7171120643615723, "created_at": "2025-01-14T16:20:24.956479+00:00"} {"global_step": 2371, "acc_step": 0, "speed/wps": 12857.636130318004, "speed/FLOPS": 201946843670186.9, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.23829218745231628, "optim/lr": 0.0014219999999999999, "optim/total_tokens": 621543424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489532, "loss/out": 3.7400224208831787, "created_at": "2025-01-14T16:20:30.058742+00:00"} {"global_step": 2372, "acc_step": 0, "speed/wps": 12847.472604414856, "speed/FLOPS": 201787211529729.84, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1856350600719452, "optim/lr": 0.0014226, "optim/total_tokens": 621805568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.706362724304199, "created_at": "2025-01-14T16:20:35.167005+00:00"} {"global_step": 2373, "acc_step": 0, "speed/wps": 12852.359465164895, "speed/FLOPS": 201863966393059.56, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15895970165729523, "optim/lr": 0.0014232, "optim/total_tokens": 622067712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.6085782051086426, "created_at": "2025-01-14T16:20:40.272466+00:00"} {"global_step": 2374, "acc_step": 0, "speed/wps": 12849.444102669995, "speed/FLOPS": 201818176618949.28, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.16134777665138245, "optim/lr": 0.0014238, "optim/total_tokens": 622329856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322598, "loss/out": 3.7513926029205322, "created_at": "2025-01-14T16:20:45.379723+00:00"} {"global_step": 2375, "acc_step": 0, "speed/wps": 12859.196687388718, "speed/FLOPS": 201971354363412.6, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16733253002166748, "optim/lr": 0.0014244, "optim/total_tokens": 622592000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.7735776901245117, "created_at": "2025-01-14T16:20:50.481076+00:00"} {"global_step": 2376, "acc_step": 0, "speed/wps": 12856.365874293893, "speed/FLOPS": 201926892553813.56, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16707970201969147, "optim/lr": 0.001425, "optim/total_tokens": 622854144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307679, "loss/out": 3.6776084899902344, "created_at": "2025-01-14T16:20:55.586616+00:00"} {"global_step": 2377, "acc_step": 0, "speed/wps": 12853.348650820812, "speed/FLOPS": 201879502913069.28, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.17827700078487396, "optim/lr": 0.0014256, "optim/total_tokens": 623116288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466683, "loss/out": 3.61759090423584, "created_at": "2025-01-14T16:21:00.690614+00:00"} {"global_step": 2378, "acc_step": 0, "speed/wps": 12857.399452275606, "speed/FLOPS": 201943126316301.34, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1856130212545395, "optim/lr": 0.0014262, "optim/total_tokens": 623378432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445933, "loss/out": 3.724970817565918, "created_at": "2025-01-14T16:21:05.796731+00:00"} {"global_step": 2379, "acc_step": 0, "speed/wps": 12852.00257969079, "speed/FLOPS": 201858361016276.72, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17948251962661743, "optim/lr": 0.0014268000000000002, "optim/total_tokens": 623640576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 3.7407519817352295, "created_at": "2025-01-14T16:21:10.904592+00:00"} {"global_step": 2380, "acc_step": 0, "speed/wps": 12845.690500854247, "speed/FLOPS": 201759221144443.56, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18570536375045776, "optim/lr": 0.0014274000000000001, "optim/total_tokens": 623902720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 3.652960777282715, "created_at": "2025-01-14T16:21:16.013291+00:00"} {"global_step": 2381, "acc_step": 0, "speed/wps": 12859.072062081517, "speed/FLOPS": 201969396951709.9, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18940883874893188, "optim/lr": 0.001428, "optim/total_tokens": 624164864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 3.6509475708007812, "created_at": "2025-01-14T16:21:21.113368+00:00"} {"global_step": 2382, "acc_step": 0, "speed/wps": 12860.641231827556, "speed/FLOPS": 201994042918836.62, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1704830825328827, "optim/lr": 0.0014286000000000001, "optim/total_tokens": 624427008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427855, "loss/out": 3.7996764183044434, "created_at": "2025-01-14T16:21:26.216779+00:00"} {"global_step": 2383, "acc_step": 0, "speed/wps": 12858.149045260945, "speed/FLOPS": 201954899704184.78, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18812225759029388, "optim/lr": 0.0014292, "optim/total_tokens": 624689152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499510, "loss/out": 3.685408115386963, "created_at": "2025-01-14T16:21:31.317505+00:00"} {"global_step": 2384, "acc_step": 0, "speed/wps": 12849.535474057328, "speed/FLOPS": 201819611732144.78, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1618839055299759, "optim/lr": 0.0014298000000000002, "optim/total_tokens": 624951296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 3.6778340339660645, "created_at": "2025-01-14T16:21:36.421344+00:00"} {"global_step": 2385, "acc_step": 0, "speed/wps": 12852.556888698893, "speed/FLOPS": 201867067200949.38, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19483648240566254, "optim/lr": 0.0014304, "optim/total_tokens": 625213440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 3.6942920684814453, "created_at": "2025-01-14T16:21:41.527233+00:00"} {"global_step": 2386, "acc_step": 0, "speed/wps": 12857.519320762622, "speed/FLOPS": 201945009015606.47, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.17807355523109436, "optim/lr": 0.001431, "optim/total_tokens": 625475584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.7437305450439453, "created_at": "2025-01-14T16:21:46.632790+00:00"} {"global_step": 2387, "acc_step": 0, "speed/wps": 12852.354944086299, "speed/FLOPS": 201863895383307.62, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17730075120925903, "optim/lr": 0.0014316, "optim/total_tokens": 625737728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325525, "loss/out": 3.780823230743408, "created_at": "2025-01-14T16:21:51.736753+00:00"} {"global_step": 2388, "acc_step": 0, "speed/wps": 12849.230135120979, "speed/FLOPS": 201814815964569.78, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21420074999332428, "optim/lr": 0.0014322, "optim/total_tokens": 625999872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470403, "loss/out": 3.7898073196411133, "created_at": "2025-01-14T16:21:56.842455+00:00"} {"global_step": 2389, "acc_step": 0, "speed/wps": 12851.667760693737, "speed/FLOPS": 201853102223838.5, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17651082575321198, "optim/lr": 0.0014328000000000001, "optim/total_tokens": 626262016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325381, "loss/out": 3.5166029930114746, "created_at": "2025-01-14T16:22:01.946651+00:00"} {"global_step": 2390, "acc_step": 0, "speed/wps": 12857.136954109741, "speed/FLOPS": 201939003421896.16, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2018708437681198, "optim/lr": 0.0014334, "optim/total_tokens": 626524160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449411, "loss/out": 3.6206116676330566, "created_at": "2025-01-14T16:22:07.048558+00:00"} {"global_step": 2391, "acc_step": 0, "speed/wps": 12853.050535108947, "speed/FLOPS": 201874820596160.53, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1492658108472824, "optim/lr": 0.001434, "optim/total_tokens": 626786304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423001, "loss/out": 3.7294530868530273, "created_at": "2025-01-14T16:22:12.156253+00:00"} {"global_step": 2392, "acc_step": 0, "speed/wps": 12851.410345508299, "speed/FLOPS": 201849059164625.66, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20386667549610138, "optim/lr": 0.0014346, "optim/total_tokens": 627048448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.7826454639434814, "created_at": "2025-01-14T16:22:17.261818+00:00"} {"global_step": 2393, "acc_step": 0, "speed/wps": 12855.471294354329, "speed/FLOPS": 201912841946581.06, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.18517504632472992, "optim/lr": 0.0014352, "optim/total_tokens": 627310592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 3.74007511138916, "created_at": "2025-01-14T16:22:22.366034+00:00"} {"global_step": 2394, "acc_step": 0, "speed/wps": 12853.120666723655, "speed/FLOPS": 201875922109540.44, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19026045501232147, "optim/lr": 0.0014358, "optim/total_tokens": 627572736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.6598060131073, "created_at": "2025-01-14T16:22:27.469851+00:00"} {"global_step": 2395, "acc_step": 0, "speed/wps": 12853.361558790808, "speed/FLOPS": 201879705650475.47, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.21417100727558136, "optim/lr": 0.0014364, "optim/total_tokens": 627834880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440339, "loss/out": 3.688941478729248, "created_at": "2025-01-14T16:22:32.573040+00:00"} {"global_step": 2396, "acc_step": 0, "speed/wps": 12861.647276112642, "speed/FLOPS": 202009844226783.62, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.19935165345668793, "optim/lr": 0.001437, "optim/total_tokens": 628097024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 3.809203624725342, "created_at": "2025-01-14T16:22:37.673411+00:00"} {"global_step": 2397, "acc_step": 0, "speed/wps": 12854.862116758479, "speed/FLOPS": 201903273975340.47, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.24109800159931183, "optim/lr": 0.0014376, "optim/total_tokens": 628359168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 3.6867103576660156, "created_at": "2025-01-14T16:22:42.778425+00:00"} {"global_step": 2398, "acc_step": 0, "speed/wps": 12851.565297070525, "speed/FLOPS": 201851492891836.2, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.23699592053890228, "optim/lr": 0.0014382, "optim/total_tokens": 628621312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 3.787261962890625, "created_at": "2025-01-14T16:22:47.886846+00:00"} {"global_step": 2399, "acc_step": 0, "speed/wps": 12843.75124823755, "speed/FLOPS": 201728762517285.97, "speed/curr_iter_time": 1.2868, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2724912464618683, "optim/lr": 0.0014388, "optim/total_tokens": 628883456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.6820170879364014, "created_at": "2025-01-14T16:22:52.993429+00:00"} {"global_step": 2400, "acc_step": 0, "speed/wps": 12844.329514071116, "speed/FLOPS": 201737844976822.66, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21673206984996796, "optim/lr": 0.0014394, "optim/total_tokens": 629145600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295655, "loss/out": 3.585764169692993, "created_at": "2025-01-14T16:22:58.102670+00:00"} {"global_step": 2401, "acc_step": 0, "speed/wps": 12835.71476779501, "speed/FLOPS": 201602538548654.53, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1852361559867859, "optim/lr": 0.0014399999999999999, "optim/total_tokens": 629407744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461913, "loss/out": 3.659735679626465, "created_at": "2025-01-14T16:23:03.214316+00:00"} {"global_step": 2402, "acc_step": 0, "speed/wps": 12856.853041074584, "speed/FLOPS": 201934544177541.62, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.16637380421161652, "optim/lr": 0.0014406, "optim/total_tokens": 629669888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 3.7699954509735107, "created_at": "2025-01-14T16:23:08.317248+00:00"} {"global_step": 2403, "acc_step": 0, "speed/wps": 12848.941571535875, "speed/FLOPS": 201810283676937.8, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.2025434672832489, "optim/lr": 0.0014412, "optim/total_tokens": 629932032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 3.7929294109344482, "created_at": "2025-01-14T16:23:13.424219+00:00"} {"global_step": 2404, "acc_step": 0, "speed/wps": 12843.63564492125, "speed/FLOPS": 201726946808347.9, "speed/curr_iter_time": 1.2884, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.15002253651618958, "optim/lr": 0.0014418, "optim/total_tokens": 630194176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.6275320053100586, "created_at": "2025-01-14T16:23:18.531678+00:00"} {"global_step": 2405, "acc_step": 0, "speed/wps": 12853.880547336821, "speed/FLOPS": 201887857078755.62, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1676110178232193, "optim/lr": 0.0014424000000000001, "optim/total_tokens": 630456320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324468, "loss/out": 3.8276119232177734, "created_at": "2025-01-14T16:23:23.637315+00:00"} {"global_step": 2406, "acc_step": 0, "speed/wps": 12851.868294160808, "speed/FLOPS": 201856251877500.7, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1745540052652359, "optim/lr": 0.001443, "optim/total_tokens": 630718464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477242, "loss/out": 3.6997876167297363, "created_at": "2025-01-14T16:23:28.740903+00:00"} {"global_step": 2407, "acc_step": 0, "speed/wps": 12852.796574119213, "speed/FLOPS": 201870831789837.78, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20204992592334747, "optim/lr": 0.0014436000000000002, "optim/total_tokens": 630980608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304859, "loss/out": 3.6758317947387695, "created_at": "2025-01-14T16:23:33.845878+00:00"} {"global_step": 2408, "acc_step": 0, "speed/wps": 12848.96686303216, "speed/FLOPS": 201810680914641.1, "speed/curr_iter_time": 1.2871, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.17768247425556183, "optim/lr": 0.0014442, "optim/total_tokens": 631242752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 3.8144686222076416, "created_at": "2025-01-14T16:23:38.953913+00:00"} {"global_step": 2409, "acc_step": 0, "speed/wps": 12852.610882986013, "speed/FLOPS": 201867915255424.53, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17102034389972687, "optim/lr": 0.0014448, "optim/total_tokens": 631504896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465633, "loss/out": 3.7827606201171875, "created_at": "2025-01-14T16:23:44.056739+00:00"} {"global_step": 2410, "acc_step": 0, "speed/wps": 12854.930070104288, "speed/FLOPS": 201904341276014.1, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1914941966533661, "optim/lr": 0.0014454000000000001, "optim/total_tokens": 631767040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.7166645526885986, "created_at": "2025-01-14T16:23:49.159125+00:00"} {"global_step": 2411, "acc_step": 0, "speed/wps": 12856.057887749328, "speed/FLOPS": 201922055201913.3, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1877245008945465, "optim/lr": 0.001446, "optim/total_tokens": 632029184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324468, "loss/out": 3.6303048133850098, "created_at": "2025-01-14T16:23:54.263037+00:00"} {"global_step": 2412, "acc_step": 0, "speed/wps": 12853.182375571156, "speed/FLOPS": 201876891331787.3, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18206541240215302, "optim/lr": 0.0014466000000000001, "optim/total_tokens": 632291328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 3.6665592193603516, "created_at": "2025-01-14T16:23:59.372187+00:00"} {"global_step": 2413, "acc_step": 0, "speed/wps": 12859.347862409353, "speed/FLOPS": 201973728774847.94, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1764543652534485, "optim/lr": 0.0014472, "optim/total_tokens": 632553472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350820, "loss/out": 3.6850991249084473, "created_at": "2025-01-14T16:24:04.477603+00:00"} {"global_step": 2414, "acc_step": 0, "speed/wps": 12855.600824162304, "speed/FLOPS": 201914876390207.9, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21253308653831482, "optim/lr": 0.0014478, "optim/total_tokens": 632815616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.651665210723877, "created_at": "2025-01-14T16:24:09.582539+00:00"} {"global_step": 2415, "acc_step": 0, "speed/wps": 12846.352495566265, "speed/FLOPS": 201769618681071.88, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19671766459941864, "optim/lr": 0.0014484, "optim/total_tokens": 633077760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 3.6377596855163574, "created_at": "2025-01-14T16:24:14.687812+00:00"} {"global_step": 2416, "acc_step": 0, "speed/wps": 12853.146049800449, "speed/FLOPS": 201876320785641.5, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.43154770135879517, "optim/lr": 0.001449, "optim/total_tokens": 633339904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 3.557344436645508, "created_at": "2025-01-14T16:24:19.790273+00:00"} {"global_step": 2417, "acc_step": 0, "speed/wps": 12860.244674168009, "speed/FLOPS": 201987814435866.12, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2520911395549774, "optim/lr": 0.0014496, "optim/total_tokens": 633602048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 3.808492422103882, "created_at": "2025-01-14T16:24:24.892379+00:00"} {"global_step": 2418, "acc_step": 0, "speed/wps": 12849.421711692912, "speed/FLOPS": 201817824937885.78, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0057494640350342, "optim/lr": 0.0014502, "optim/total_tokens": 633864192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477559, "loss/out": 3.6993138790130615, "created_at": "2025-01-14T16:24:30.001374+00:00"} {"global_step": 2419, "acc_step": 0, "speed/wps": 12851.472960332461, "speed/FLOPS": 201850042616480.9, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22958961129188538, "optim/lr": 0.0014508, "optim/total_tokens": 634126336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480426, "loss/out": 3.743595600128174, "created_at": "2025-01-14T16:24:35.106868+00:00"} {"global_step": 2420, "acc_step": 0, "speed/wps": 12854.948765208566, "speed/FLOPS": 201904634908316.56, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21342040598392487, "optim/lr": 0.0014514, "optim/total_tokens": 634388480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321630, "loss/out": 3.7593023777008057, "created_at": "2025-01-14T16:24:40.213428+00:00"} {"global_step": 2421, "acc_step": 0, "speed/wps": 12847.216440883842, "speed/FLOPS": 201783188129469.7, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17714987695217133, "optim/lr": 0.001452, "optim/total_tokens": 634650624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.7385005950927734, "created_at": "2025-01-14T16:24:45.318164+00:00"} {"global_step": 2422, "acc_step": 0, "speed/wps": 12856.660129438136, "speed/FLOPS": 201931514235202.84, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19846391677856445, "optim/lr": 0.0014526, "optim/total_tokens": 634912768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462701, "loss/out": 3.804230213165283, "created_at": "2025-01-14T16:24:50.421404+00:00"} {"global_step": 2423, "acc_step": 0, "speed/wps": 12857.980665205352, "speed/FLOPS": 201952255064033.25, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18073152005672455, "optim/lr": 0.0014532, "optim/total_tokens": 635174912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.653593063354492, "created_at": "2025-01-14T16:24:55.521867+00:00"} {"global_step": 2424, "acc_step": 0, "speed/wps": 12856.569563385674, "speed/FLOPS": 201930091770898.12, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19054654240608215, "optim/lr": 0.0014537999999999999, "optim/total_tokens": 635437056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328460, "loss/out": 3.6947078704833984, "created_at": "2025-01-14T16:25:00.624512+00:00"} {"global_step": 2425, "acc_step": 0, "speed/wps": 12854.09456648084, "speed/FLOPS": 201891218543505.56, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.21995501220226288, "optim/lr": 0.0014544, "optim/total_tokens": 635699200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.8162102699279785, "created_at": "2025-01-14T16:25:05.726671+00:00"} {"global_step": 2426, "acc_step": 0, "speed/wps": 12850.691581451902, "speed/FLOPS": 201837769987434.62, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20459313690662384, "optim/lr": 0.001455, "optim/total_tokens": 635961344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 282855, "loss/out": 3.6549205780029297, "created_at": "2025-01-14T16:25:10.835455+00:00"} {"global_step": 2427, "acc_step": 0, "speed/wps": 12850.492089096897, "speed/FLOPS": 201834636685868.03, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2183780074119568, "optim/lr": 0.0014556, "optim/total_tokens": 636223488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450462, "loss/out": 3.6831719875335693, "created_at": "2025-01-14T16:25:15.942061+00:00"} {"global_step": 2428, "acc_step": 0, "speed/wps": 12854.467505117867, "speed/FLOPS": 201897076057271.47, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18581265211105347, "optim/lr": 0.0014562, "optim/total_tokens": 636485632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413178, "loss/out": 3.8029839992523193, "created_at": "2025-01-14T16:25:21.047319+00:00"} {"global_step": 2429, "acc_step": 0, "speed/wps": 12857.504259553494, "speed/FLOPS": 201944772458620.88, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1676809936761856, "optim/lr": 0.0014568, "optim/total_tokens": 636747776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 3.6044654846191406, "created_at": "2025-01-14T16:25:26.147993+00:00"} {"global_step": 2430, "acc_step": 0, "speed/wps": 12857.948010318134, "speed/FLOPS": 201951742174157.0, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1593324840068817, "optim/lr": 0.0014574, "optim/total_tokens": 637009920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 3.7266130447387695, "created_at": "2025-01-14T16:25:31.250890+00:00"} {"global_step": 2431, "acc_step": 0, "speed/wps": 12849.860185943086, "speed/FLOPS": 201824711778511.5, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.141632080078125, "optim/lr": 0.001458, "optim/total_tokens": 637272064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 3.5758707523345947, "created_at": "2025-01-14T16:25:36.357469+00:00"} {"global_step": 2432, "acc_step": 0, "speed/wps": 12847.038376516832, "speed/FLOPS": 201780391383899.7, "speed/curr_iter_time": 1.2878, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0041570663452148, "optim/lr": 0.0014586, "optim/total_tokens": 637534208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 3.625338554382324, "created_at": "2025-01-14T16:25:41.465919+00:00"} {"global_step": 2433, "acc_step": 0, "speed/wps": 12851.16886942399, "speed/FLOPS": 201845266450898.66, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1634787768125534, "optim/lr": 0.0014592000000000001, "optim/total_tokens": 637796352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 3.6289610862731934, "created_at": "2025-01-14T16:25:46.573667+00:00"} {"global_step": 2434, "acc_step": 0, "speed/wps": 12853.099502497924, "speed/FLOPS": 201875589696292.66, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17491966485977173, "optim/lr": 0.0014598, "optim/total_tokens": 638058496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.544705390930176, "created_at": "2025-01-14T16:25:51.678871+00:00"} {"global_step": 2435, "acc_step": 0, "speed/wps": 12844.047624570032, "speed/FLOPS": 201733417514851.38, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23433908820152283, "optim/lr": 0.0014604000000000002, "optim/total_tokens": 638320640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 3.7565202713012695, "created_at": "2025-01-14T16:25:56.789629+00:00"} {"global_step": 2436, "acc_step": 0, "speed/wps": 12856.541087348533, "speed/FLOPS": 201929644515604.62, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.28381574153900146, "optim/lr": 0.001461, "optim/total_tokens": 638582784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.675685167312622, "created_at": "2025-01-14T16:26:01.892965+00:00"} {"global_step": 2437, "acc_step": 0, "speed/wps": 12853.001917889693, "speed/FLOPS": 201874056995925.1, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.001, "optim/grad_norm": 0.2461012750864029, "optim/lr": 0.0014616000000000002, "optim/total_tokens": 638844928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317507, "loss/out": 3.7129883766174316, "created_at": "2025-01-14T16:26:06.997879+00:00"} {"global_step": 2438, "acc_step": 0, "speed/wps": 12857.858612568743, "speed/FLOPS": 201950338059657.44, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22181235253810883, "optim/lr": 0.0014622, "optim/total_tokens": 639107072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466155, "loss/out": 3.648817539215088, "created_at": "2025-01-14T16:26:12.101879+00:00"} {"global_step": 2439, "acc_step": 0, "speed/wps": 12845.528173173423, "speed/FLOPS": 201756671565154.6, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1852959394454956, "optim/lr": 0.0014628, "optim/total_tokens": 639369216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 391836, "loss/out": 3.6588621139526367, "created_at": "2025-01-14T16:26:17.210571+00:00"} {"global_step": 2440, "acc_step": 0, "speed/wps": 12853.305665918126, "speed/FLOPS": 201878827776105.16, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17028947174549103, "optim/lr": 0.0014634000000000001, "optim/total_tokens": 639631360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.718801498413086, "created_at": "2025-01-14T16:26:22.315428+00:00"} {"global_step": 2441, "acc_step": 0, "speed/wps": 12851.407667528263, "speed/FLOPS": 201849017103269.25, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1824169158935547, "optim/lr": 0.001464, "optim/total_tokens": 639893504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.652129888534546, "created_at": "2025-01-14T16:26:27.418583+00:00"} {"global_step": 2442, "acc_step": 0, "speed/wps": 12851.477431502064, "speed/FLOPS": 201850112842343.56, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1830766201019287, "optim/lr": 0.0014646000000000001, "optim/total_tokens": 640155648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 3.6892709732055664, "created_at": "2025-01-14T16:26:32.525288+00:00"} {"global_step": 2443, "acc_step": 0, "speed/wps": 12857.248832834453, "speed/FLOPS": 201940760631005.25, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20000199973583221, "optim/lr": 0.0014652, "optim/total_tokens": 640417792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 3.6745922565460205, "created_at": "2025-01-14T16:26:37.627658+00:00"} {"global_step": 2444, "acc_step": 0, "speed/wps": 12853.430407213475, "speed/FLOPS": 201880787009561.1, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1483556479215622, "optim/lr": 0.0014658, "optim/total_tokens": 640679936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325000, "loss/out": 3.6511988639831543, "created_at": "2025-01-14T16:26:42.736057+00:00"} {"global_step": 2445, "acc_step": 0, "speed/wps": 12855.455500506097, "speed/FLOPS": 201912593882491.56, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20532245934009552, "optim/lr": 0.0014664, "optim/total_tokens": 640942080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455808, "loss/out": 3.726896047592163, "created_at": "2025-01-14T16:26:47.839375+00:00"} {"global_step": 2446, "acc_step": 0, "speed/wps": 12858.5393857374, "speed/FLOPS": 201961030537751.53, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19880281388759613, "optim/lr": 0.001467, "optim/total_tokens": 641204224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300932, "loss/out": 3.692183017730713, "created_at": "2025-01-14T16:26:52.942070+00:00"} {"global_step": 2447, "acc_step": 0, "speed/wps": 12848.378291432435, "speed/FLOPS": 201801436588885.28, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16742433607578278, "optim/lr": 0.0014676000000000001, "optim/total_tokens": 641466368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.6818175315856934, "created_at": "2025-01-14T16:26:58.046425+00:00"} {"global_step": 2448, "acc_step": 0, "speed/wps": 12858.993117129336, "speed/FLOPS": 201968157012754.16, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.20357745885849, "optim/lr": 0.0014682, "optim/total_tokens": 641728512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464298, "loss/out": 3.602170944213867, "created_at": "2025-01-14T16:27:03.147179+00:00"} {"global_step": 2449, "acc_step": 0, "speed/wps": 12853.696825276385, "speed/FLOPS": 201884971471335.9, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1932532787322998, "optim/lr": 0.0014688, "optim/total_tokens": 641990656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 3.7602460384368896, "created_at": "2025-01-14T16:27:08.250507+00:00"} {"global_step": 2450, "acc_step": 0, "speed/wps": 12852.966360809176, "speed/FLOPS": 201873498523115.03, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1720113754272461, "optim/lr": 0.0014694, "optim/total_tokens": 642252800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330862, "loss/out": 3.678846836090088, "created_at": "2025-01-14T16:27:13.358100+00:00"} {"global_step": 2451, "acc_step": 0, "speed/wps": 12855.362555963258, "speed/FLOPS": 201911134060727.06, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3303690552711487, "optim/lr": 0.00147, "optim/total_tokens": 642514944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.7919631004333496, "created_at": "2025-01-14T16:27:18.459640+00:00"} {"global_step": 2452, "acc_step": 0, "speed/wps": 12858.785763180724, "speed/FLOPS": 201964900234057.03, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2138756662607193, "optim/lr": 0.0014706, "optim/total_tokens": 642777088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302343, "loss/out": 3.6712560653686523, "created_at": "2025-01-14T16:27:23.562198+00:00"} {"global_step": 2453, "acc_step": 0, "speed/wps": 12852.31588125464, "speed/FLOPS": 201863281847858.47, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18080562353134155, "optim/lr": 0.0014712, "optim/total_tokens": 643039232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 3.695253849029541, "created_at": "2025-01-14T16:27:28.665723+00:00"} {"global_step": 2454, "acc_step": 0, "speed/wps": 12858.651944434832, "speed/FLOPS": 201962798426766.56, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1914263367652893, "optim/lr": 0.0014717999999999999, "optim/total_tokens": 643301376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360123, "loss/out": 3.6906039714813232, "created_at": "2025-01-14T16:27:33.769392+00:00"} {"global_step": 2455, "acc_step": 0, "speed/wps": 12856.901128357822, "speed/FLOPS": 201935299454403.72, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17049185931682587, "optim/lr": 0.0014724, "optim/total_tokens": 643563520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466155, "loss/out": 3.632699966430664, "created_at": "2025-01-14T16:27:38.874070+00:00"} {"global_step": 2456, "acc_step": 0, "speed/wps": 12854.900084313245, "speed/FLOPS": 201903870307960.97, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1941910833120346, "optim/lr": 0.001473, "optim/total_tokens": 643825664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 3.631540060043335, "created_at": "2025-01-14T16:27:43.975787+00:00"} {"global_step": 2457, "acc_step": 0, "speed/wps": 12850.55515493344, "speed/FLOPS": 201835627221492.5, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.2124902755022049, "optim/lr": 0.0014736, "optim/total_tokens": 644087808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 3.7071425914764404, "created_at": "2025-01-14T16:27:49.079500+00:00"} {"global_step": 2458, "acc_step": 0, "speed/wps": 12853.66492468007, "speed/FLOPS": 201884470428634.75, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17353390157222748, "optim/lr": 0.0014742, "optim/total_tokens": 644349952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.652296781539917, "created_at": "2025-01-14T16:27:54.182699+00:00"} {"global_step": 2459, "acc_step": 0, "speed/wps": 12852.08627425586, "speed/FLOPS": 201859675554429.53, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.18155521154403687, "optim/lr": 0.0014748, "optim/total_tokens": 644612096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479707, "loss/out": 3.6562695503234863, "created_at": "2025-01-14T16:27:59.294597+00:00"} {"global_step": 2460, "acc_step": 0, "speed/wps": 12858.526039729939, "speed/FLOPS": 201960820920365.03, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.15990377962589264, "optim/lr": 0.0014754, "optim/total_tokens": 644874240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501606, "loss/out": 3.665194511413574, "created_at": "2025-01-14T16:28:04.394883+00:00"} {"global_step": 2461, "acc_step": 0, "speed/wps": 12861.513247354806, "speed/FLOPS": 202007739120967.0, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2004203498363495, "optim/lr": 0.001476, "optim/total_tokens": 645136384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 3.6849827766418457, "created_at": "2025-01-14T16:28:09.499464+00:00"} {"global_step": 2462, "acc_step": 0, "speed/wps": 12856.60110491927, "speed/FLOPS": 201930587174025.84, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1543688178062439, "optim/lr": 0.0014766, "optim/total_tokens": 645398528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 3.713284969329834, "created_at": "2025-01-14T16:28:14.601415+00:00"} {"global_step": 2463, "acc_step": 0, "speed/wps": 12855.164642298821, "speed/FLOPS": 201908025554665.22, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17203256487846375, "optim/lr": 0.0014772000000000001, "optim/total_tokens": 645660672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.577645778656006, "created_at": "2025-01-14T16:28:19.703081+00:00"} {"global_step": 2464, "acc_step": 0, "speed/wps": 12848.683106598057, "speed/FLOPS": 201806224129921.3, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.1920345276594162, "optim/lr": 0.0014778, "optim/total_tokens": 645922816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.703742027282715, "created_at": "2025-01-14T16:28:24.808000+00:00"} {"global_step": 2465, "acc_step": 0, "speed/wps": 12853.567735313587, "speed/FLOPS": 201882943936081.97, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16984835267066956, "optim/lr": 0.0014784000000000002, "optim/total_tokens": 646184960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 3.821507453918457, "created_at": "2025-01-14T16:28:29.910306+00:00"} {"global_step": 2466, "acc_step": 0, "speed/wps": 12860.63087118522, "speed/FLOPS": 201993880190711.72, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2830020487308502, "optim/lr": 0.001479, "optim/total_tokens": 646447104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.7880680561065674, "created_at": "2025-01-14T16:28:35.010952+00:00"} {"global_step": 2467, "acc_step": 0, "speed/wps": 12857.99312668672, "speed/FLOPS": 201952450788722.03, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17309649288654327, "optim/lr": 0.0014796000000000002, "optim/total_tokens": 646709248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462126, "loss/out": 3.7093400955200195, "created_at": "2025-01-14T16:28:40.117689+00:00"} {"global_step": 2468, "acc_step": 0, "speed/wps": 12857.214386900168, "speed/FLOPS": 201940219610273.62, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16744789481163025, "optim/lr": 0.0014802, "optim/total_tokens": 646971392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 3.5236968994140625, "created_at": "2025-01-14T16:28:45.220348+00:00"} {"global_step": 2469, "acc_step": 0, "speed/wps": 12854.812957468248, "speed/FLOPS": 201902501861135.47, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1496133953332901, "optim/lr": 0.0014808, "optim/total_tokens": 647233536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320810, "loss/out": 3.7041091918945312, "created_at": "2025-01-14T16:28:50.324322+00:00"} {"global_step": 2470, "acc_step": 0, "speed/wps": 12851.279611627859, "speed/FLOPS": 201847005809386.16, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17299672961235046, "optim/lr": 0.0014814000000000001, "optim/total_tokens": 647495680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454458, "loss/out": 3.7127208709716797, "created_at": "2025-01-14T16:28:55.428925+00:00"} {"global_step": 2471, "acc_step": 0, "speed/wps": 12856.19059555111, "speed/FLOPS": 201924139560299.9, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19612127542495728, "optim/lr": 0.001482, "optim/total_tokens": 647757824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330091, "loss/out": 3.6164848804473877, "created_at": "2025-01-14T16:29:00.535855+00:00"} {"global_step": 2472, "acc_step": 0, "speed/wps": 12856.705542136833, "speed/FLOPS": 201932227504039.47, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20034724473953247, "optim/lr": 0.0014826, "optim/total_tokens": 648019968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 3.7033333778381348, "created_at": "2025-01-14T16:29:05.638748+00:00"} {"global_step": 2473, "acc_step": 0, "speed/wps": 12857.907372260392, "speed/FLOPS": 201951103897617.6, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.23608362674713135, "optim/lr": 0.0014832, "optim/total_tokens": 648282112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301104, "loss/out": 3.808763265609741, "created_at": "2025-01-14T16:29:10.741415+00:00"} {"global_step": 2474, "acc_step": 0, "speed/wps": 12853.895353522295, "speed/FLOPS": 201888089630244.3, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2845735549926758, "optim/lr": 0.0014838, "optim/total_tokens": 648544256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 3.664541721343994, "created_at": "2025-01-14T16:29:15.848443+00:00"} {"global_step": 2475, "acc_step": 0, "speed/wps": 12858.813417350053, "speed/FLOPS": 201965334580787.16, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20381984114646912, "optim/lr": 0.0014844, "optim/total_tokens": 648806400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318319, "loss/out": 3.776279926300049, "created_at": "2025-01-14T16:29:20.950364+00:00"} {"global_step": 2476, "acc_step": 0, "speed/wps": 12857.908584561572, "speed/FLOPS": 201951122938473.5, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18329459428787231, "optim/lr": 0.001485, "optim/total_tokens": 649068544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 3.801785469055176, "created_at": "2025-01-14T16:29:26.055271+00:00"} {"global_step": 2477, "acc_step": 0, "speed/wps": 12855.570407802783, "speed/FLOPS": 201914398659485.16, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.20809708535671234, "optim/lr": 0.0014856, "optim/total_tokens": 649330688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350526, "loss/out": 3.6425986289978027, "created_at": "2025-01-14T16:29:31.156640+00:00"} {"global_step": 2478, "acc_step": 0, "speed/wps": 12858.782400701257, "speed/FLOPS": 201964847421696.47, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2545667588710785, "optim/lr": 0.0014862, "optim/total_tokens": 649592832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450225, "loss/out": 3.6788454055786133, "created_at": "2025-01-14T16:29:36.258385+00:00"} {"global_step": 2479, "acc_step": 0, "speed/wps": 12855.259173882316, "speed/FLOPS": 201909510303084.0, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.23649227619171143, "optim/lr": 0.0014868, "optim/total_tokens": 649854976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 397678, "loss/out": 3.7356958389282227, "created_at": "2025-01-14T16:29:41.363309+00:00"} {"global_step": 2480, "acc_step": 0, "speed/wps": 12851.840298997504, "speed/FLOPS": 201855812174991.62, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6664302349090576, "optim/lr": 0.0014874, "optim/total_tokens": 650117120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 3.6408767700195312, "created_at": "2025-01-14T16:29:46.469687+00:00"} {"global_step": 2481, "acc_step": 0, "speed/wps": 12841.823452755278, "speed/FLOPS": 201698483840168.3, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20750556886196136, "optim/lr": 0.001488, "optim/total_tokens": 650379264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.7487313747406006, "created_at": "2025-01-14T16:29:51.577543+00:00"} {"global_step": 2482, "acc_step": 0, "speed/wps": 12847.166005465071, "speed/FLOPS": 201782395971912.34, "speed/curr_iter_time": 1.2879, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22662316262722015, "optim/lr": 0.0014885999999999999, "optim/total_tokens": 650641408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 3.7144947052001953, "created_at": "2025-01-14T16:29:56.685131+00:00"} {"global_step": 2483, "acc_step": 0, "speed/wps": 12848.294377720407, "speed/FLOPS": 201800118608728.38, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18120981752872467, "optim/lr": 0.0014892, "optim/total_tokens": 650903552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 3.799740791320801, "created_at": "2025-01-14T16:30:01.791647+00:00"} {"global_step": 2484, "acc_step": 0, "speed/wps": 12853.938645257418, "speed/FLOPS": 201888769586433.5, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2023555040359497, "optim/lr": 0.0014898, "optim/total_tokens": 651165696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335820, "loss/out": 3.6665990352630615, "created_at": "2025-01-14T16:30:06.894963+00:00"} {"global_step": 2485, "acc_step": 0, "speed/wps": 12855.831655930497, "speed/FLOPS": 201918501920323.47, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15840716660022736, "optim/lr": 0.0014904, "optim/total_tokens": 651427840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 3.6956567764282227, "created_at": "2025-01-14T16:30:12.004964+00:00"} {"global_step": 2486, "acc_step": 0, "speed/wps": 12848.181244910429, "speed/FLOPS": 201798341702486.9, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1657075434923172, "optim/lr": 0.001491, "optim/total_tokens": 651689984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 393877, "loss/out": 3.532965898513794, "created_at": "2025-01-14T16:30:17.110301+00:00"} {"global_step": 2487, "acc_step": 0, "speed/wps": 12847.707380699476, "speed/FLOPS": 201790899013896.22, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1703498512506485, "optim/lr": 0.0014916, "optim/total_tokens": 651952128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457902, "loss/out": 3.7079858779907227, "created_at": "2025-01-14T16:30:22.218974+00:00"} {"global_step": 2488, "acc_step": 0, "speed/wps": 12857.299046753398, "speed/FLOPS": 201941549309603.4, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18630021810531616, "optim/lr": 0.0014922, "optim/total_tokens": 652214272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338354, "loss/out": 3.709859848022461, "created_at": "2025-01-14T16:30:27.323271+00:00"} {"global_step": 2489, "acc_step": 0, "speed/wps": 12847.878595231761, "speed/FLOPS": 201793588173399.2, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0009, "optim/grad_norm": 6.3550310134887695, "optim/lr": 0.0014928, "optim/total_tokens": 652476416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 3.695618152618408, "created_at": "2025-01-14T16:30:32.427957+00:00"} {"global_step": 2490, "acc_step": 0, "speed/wps": 12850.412545453504, "speed/FLOPS": 201833387343644.16, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2889552116394043, "optim/lr": 0.0014934000000000002, "optim/total_tokens": 652738560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 3.6646950244903564, "created_at": "2025-01-14T16:30:37.532485+00:00"} {"global_step": 2491, "acc_step": 0, "speed/wps": 12848.91744851503, "speed/FLOPS": 201809904791746.94, "speed/curr_iter_time": 1.2875, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.5328795909881592, "optim/lr": 0.001494, "optim/total_tokens": 653000704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 3.7393598556518555, "created_at": "2025-01-14T16:30:42.640111+00:00"} {"global_step": 2492, "acc_step": 0, "speed/wps": 12847.419211541459, "speed/FLOPS": 201786372921283.62, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.3906784057617188, "optim/lr": 0.0014946, "optim/total_tokens": 653262848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 4.805240631103516, "created_at": "2025-01-14T16:30:47.744719+00:00"} {"global_step": 2493, "acc_step": 0, "speed/wps": 12848.149648496606, "speed/FLOPS": 201797845437389.78, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5876500606536865, "optim/lr": 0.0014952000000000001, "optim/total_tokens": 653524992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311191, "loss/out": 3.9404847621917725, "created_at": "2025-01-14T16:30:52.851873+00:00"} {"global_step": 2494, "acc_step": 0, "speed/wps": 12837.185841008275, "speed/FLOPS": 201625643774935.28, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.5983476042747498, "optim/lr": 0.0014958, "optim/total_tokens": 653787136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457095, "loss/out": 3.80133056640625, "created_at": "2025-01-14T16:30:57.961506+00:00"} {"global_step": 2495, "acc_step": 0, "speed/wps": 12848.87201808836, "speed/FLOPS": 201809191244468.1, "speed/curr_iter_time": 1.287, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8395048379898071, "optim/lr": 0.0014964000000000002, "optim/total_tokens": 654049280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 383188, "loss/out": 3.9795403480529785, "created_at": "2025-01-14T16:31:03.066122+00:00"} {"global_step": 2496, "acc_step": 0, "speed/wps": 12854.19777955482, "speed/FLOPS": 201892839646662.03, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5703949928283691, "optim/lr": 0.001497, "optim/total_tokens": 654311424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 3.9754114151000977, "created_at": "2025-01-14T16:31:08.170140+00:00"} {"global_step": 2497, "acc_step": 0, "speed/wps": 12854.048596962662, "speed/FLOPS": 201890496529053.84, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3202474117279053, "optim/lr": 0.0014976, "optim/total_tokens": 654573568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477763, "loss/out": 3.9135918617248535, "created_at": "2025-01-14T16:31:13.272879+00:00"} {"global_step": 2498, "acc_step": 0, "speed/wps": 12852.659396304443, "speed/FLOPS": 201868677223754.75, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.3661975860595703, "optim/lr": 0.0014982, "optim/total_tokens": 654835712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 3.9207730293273926, "created_at": "2025-01-14T16:31:18.375448+00:00"} {"global_step": 2499, "acc_step": 0, "speed/wps": 12856.82183186621, "speed/FLOPS": 201934053994038.16, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2917717695236206, "optim/lr": 0.0014988, "optim/total_tokens": 655097856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 3.852125406265259, "created_at": "2025-01-14T16:31:23.480402+00:00"} {"global_step": 2500, "acc_step": 0, "speed/wps": 12856.219297869624, "speed/FLOPS": 201924590369653.28, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2373313158750534, "optim/lr": 0.0014994000000000001, "optim/total_tokens": 655360000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 3.810549020767212, "created_at": "2025-01-14T16:31:28.584506+00:00"} {"global_step": 2501, "acc_step": 0, "speed/wps": 3415.325083355492, "speed/FLOPS": 53642373582569.34, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2681095600128174, "optim/lr": 0.0015, "optim/total_tokens": 655622144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 386163, "loss/out": 3.766374349594116, "created_at": "2025-01-14T16:31:47.779196+00:00"} {"global_step": 2502, "acc_step": 0, "speed/wps": 12927.328032532114, "speed/FLOPS": 203041450761168.78, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.27027571201324463, "optim/lr": 0.0015006, "optim/total_tokens": 655884288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.7177741527557373, "created_at": "2025-01-14T16:31:52.858127+00:00"} {"global_step": 2503, "acc_step": 0, "speed/wps": 12918.763850197933, "speed/FLOPS": 202906938509190.7, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2839069366455078, "optim/lr": 0.0015011999999999998, "optim/total_tokens": 656146432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 3.731931686401367, "created_at": "2025-01-14T16:31:57.937087+00:00"} {"global_step": 2504, "acc_step": 0, "speed/wps": 12904.522952869189, "speed/FLOPS": 202683265647599.16, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2523937523365021, "optim/lr": 0.0015018000000000002, "optim/total_tokens": 656408576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356143, "loss/out": 3.738882064819336, "created_at": "2025-01-14T16:32:03.020554+00:00"} {"global_step": 2505, "acc_step": 0, "speed/wps": 12883.811849418289, "speed/FLOPS": 202357969307864.25, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.006317138671875, "optim/lr": 0.0015024, "optim/total_tokens": 656670720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464298, "loss/out": 3.7740182876586914, "created_at": "2025-01-14T16:32:08.114177+00:00"} {"global_step": 2506, "acc_step": 0, "speed/wps": 12866.48761029405, "speed/FLOPS": 202085868326418.8, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.22967559099197388, "optim/lr": 0.001503, "optim/total_tokens": 656932864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287239, "loss/out": 3.7767372131347656, "created_at": "2025-01-14T16:32:13.215765+00:00"} {"global_step": 2507, "acc_step": 0, "speed/wps": 12851.555781919702, "speed/FLOPS": 201851343443317.3, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.2247307002544403, "optim/lr": 0.0015036, "optim/total_tokens": 657195008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470403, "loss/out": 3.8455796241760254, "created_at": "2025-01-14T16:32:18.322928+00:00"} {"global_step": 2508, "acc_step": 0, "speed/wps": 12858.24481900916, "speed/FLOPS": 201956403962507.44, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2074177861213684, "optim/lr": 0.0015041999999999998, "optim/total_tokens": 657457152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.662616491317749, "created_at": "2025-01-14T16:32:23.424259+00:00"} {"global_step": 2509, "acc_step": 0, "speed/wps": 12862.441529342323, "speed/FLOPS": 202022319065174.06, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.212988942861557, "optim/lr": 0.0015048000000000001, "optim/total_tokens": 657719296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.6790647506713867, "created_at": "2025-01-14T16:32:28.524355+00:00"} {"global_step": 2510, "acc_step": 0, "speed/wps": 12864.276219566631, "speed/FLOPS": 202051135396275.4, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.18840570747852325, "optim/lr": 0.0015054, "optim/total_tokens": 657981440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330091, "loss/out": 3.6433048248291016, "created_at": "2025-01-14T16:32:33.625386+00:00"} {"global_step": 2511, "acc_step": 0, "speed/wps": 12863.309128566398, "speed/FLOPS": 202035945903194.44, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.19675947725772858, "optim/lr": 0.001506, "optim/total_tokens": 658243584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 3.716273784637451, "created_at": "2025-01-14T16:32:38.725166+00:00"} {"global_step": 2512, "acc_step": 0, "speed/wps": 12863.07264605252, "speed/FLOPS": 202032231620352.97, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18394480645656586, "optim/lr": 0.0015065999999999999, "optim/total_tokens": 658505728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454173, "loss/out": 3.7025351524353027, "created_at": "2025-01-14T16:32:43.825746+00:00"} {"global_step": 2513, "acc_step": 0, "speed/wps": 12859.809488492067, "speed/FLOPS": 201980979246817.66, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.20299039781093597, "optim/lr": 0.0015072, "optim/total_tokens": 658767872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.786255359649658, "created_at": "2025-01-14T16:32:48.928464+00:00"} {"global_step": 2514, "acc_step": 0, "speed/wps": 12855.108679730492, "speed/FLOPS": 201907146585628.97, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19530139863491058, "optim/lr": 0.0015078000000000001, "optim/total_tokens": 659030016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320038, "loss/out": 3.7762556076049805, "created_at": "2025-01-14T16:32:54.035128+00:00"} {"global_step": 2515, "acc_step": 0, "speed/wps": 12852.04164531255, "speed/FLOPS": 201858974595548.28, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.18728779256343842, "optim/lr": 0.0015084, "optim/total_tokens": 659292160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457665, "loss/out": 3.691807746887207, "created_at": "2025-01-14T16:32:59.140778+00:00"} {"global_step": 2516, "acc_step": 0, "speed/wps": 12861.121033879263, "speed/FLOPS": 202001578869375.8, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.001, "optim/grad_norm": 0.182379812002182, "optim/lr": 0.0015090000000000001, "optim/total_tokens": 659554304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353457, "loss/out": 3.5896406173706055, "created_at": "2025-01-14T16:33:04.242297+00:00"} {"global_step": 2517, "acc_step": 0, "speed/wps": 12856.567522164265, "speed/FLOPS": 201930059710710.97, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19170993566513062, "optim/lr": 0.0015096, "optim/total_tokens": 659816448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465057, "loss/out": 3.6929092407226562, "created_at": "2025-01-14T16:33:09.344749+00:00"} {"global_step": 2518, "acc_step": 0, "speed/wps": 12850.284178508085, "speed/FLOPS": 201831371164371.7, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1783660501241684, "optim/lr": 0.0015102, "optim/total_tokens": 660078592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 3.6146187782287598, "created_at": "2025-01-14T16:33:14.449499+00:00"} {"global_step": 2519, "acc_step": 0, "speed/wps": 12850.934095692472, "speed/FLOPS": 201841579006832.25, "speed/curr_iter_time": 1.2866, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.21125856041908264, "optim/lr": 0.0015108, "optim/total_tokens": 660340736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 3.8496930599212646, "created_at": "2025-01-14T16:33:19.557533+00:00"} {"global_step": 2520, "acc_step": 0, "speed/wps": 12863.574202097852, "speed/FLOPS": 202040109247255.12, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22327806055545807, "optim/lr": 0.0015114000000000002, "optim/total_tokens": 660602880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.6544666290283203, "created_at": "2025-01-14T16:33:24.658235+00:00"} {"global_step": 2521, "acc_step": 0, "speed/wps": 12858.966477303935, "speed/FLOPS": 201967738597689.28, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.25741904973983765, "optim/lr": 0.001512, "optim/total_tokens": 660865024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459507, "loss/out": 3.768439769744873, "created_at": "2025-01-14T16:33:29.760930+00:00"} {"global_step": 2522, "acc_step": 0, "speed/wps": 12857.116563614409, "speed/FLOPS": 201938683161147.47, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2643108069896698, "optim/lr": 0.0015126, "optim/total_tokens": 661127168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498989, "loss/out": 3.702025890350342, "created_at": "2025-01-14T16:33:34.866008+00:00"} {"global_step": 2523, "acc_step": 0, "speed/wps": 12842.522362159449, "speed/FLOPS": 201709461172762.97, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.18739941716194153, "optim/lr": 0.0015132, "optim/total_tokens": 661389312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 3.7876694202423096, "created_at": "2025-01-14T16:33:39.975047+00:00"} {"global_step": 2524, "acc_step": 0, "speed/wps": 12844.889442737842, "speed/FLOPS": 201746639425956.22, "speed/curr_iter_time": 1.2883, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1735931783914566, "optim/lr": 0.0015138000000000003, "optim/total_tokens": 661651456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 3.686760187149048, "created_at": "2025-01-14T16:33:45.080767+00:00"} {"global_step": 2525, "acc_step": 0, "speed/wps": 12853.506003983272, "speed/FLOPS": 201881974360711.62, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17273610830307007, "optim/lr": 0.0015144000000000002, "optim/total_tokens": 661913600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.657032012939453, "created_at": "2025-01-14T16:33:50.186850+00:00"} {"global_step": 2526, "acc_step": 0, "speed/wps": 12851.160415684059, "speed/FLOPS": 201845133673296.38, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17980915307998657, "optim/lr": 0.001515, "optim/total_tokens": 662175744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.649103879928589, "created_at": "2025-01-14T16:33:55.295387+00:00"} {"global_step": 2527, "acc_step": 0, "speed/wps": 12855.238007310272, "speed/FLOPS": 201909177852984.16, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17864762246608734, "optim/lr": 0.0015156, "optim/total_tokens": 662437888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355857, "loss/out": 3.7131893634796143, "created_at": "2025-01-14T16:34:00.408553+00:00"} {"global_step": 2528, "acc_step": 0, "speed/wps": 12842.8542108632, "speed/FLOPS": 201714673312663.28, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.18821804225444794, "optim/lr": 0.0015161999999999999, "optim/total_tokens": 662700032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470925, "loss/out": 3.8097262382507324, "created_at": "2025-01-14T16:34:05.515789+00:00"} {"global_step": 2529, "acc_step": 0, "speed/wps": 12851.400502000843, "speed/FLOPS": 201848904558814.66, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2163286805152893, "optim/lr": 0.0015168000000000002, "optim/total_tokens": 662962176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 3.850609302520752, "created_at": "2025-01-14T16:34:10.620838+00:00"} {"global_step": 2530, "acc_step": 0, "speed/wps": 12854.098433683312, "speed/FLOPS": 201891279283234.47, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15731501579284668, "optim/lr": 0.0015174000000000001, "optim/total_tokens": 663224320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 3.655447244644165, "created_at": "2025-01-14T16:34:15.723167+00:00"} {"global_step": 2531, "acc_step": 0, "speed/wps": 12860.314764448032, "speed/FLOPS": 201988915300028.03, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17400911450386047, "optim/lr": 0.001518, "optim/total_tokens": 663486464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 3.5945088863372803, "created_at": "2025-01-14T16:34:20.826462+00:00"} {"global_step": 2532, "acc_step": 0, "speed/wps": 12857.842822251581, "speed/FLOPS": 201950090051028.22, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16274084150791168, "optim/lr": 0.0015186, "optim/total_tokens": 663748608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 3.622335433959961, "created_at": "2025-01-14T16:34:25.927149+00:00"} {"global_step": 2533, "acc_step": 0, "speed/wps": 12859.812875009879, "speed/FLOPS": 201981032436733.47, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20010331273078918, "optim/lr": 0.0015191999999999998, "optim/total_tokens": 664010752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 3.6767711639404297, "created_at": "2025-01-14T16:34:31.037776+00:00"} {"global_step": 2534, "acc_step": 0, "speed/wps": 12857.07034226561, "speed/FLOPS": 201937957191349.8, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15600590407848358, "optim/lr": 0.0015198000000000002, "optim/total_tokens": 664272896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319663, "loss/out": 3.5811381340026855, "created_at": "2025-01-14T16:34:36.142762+00:00"} {"global_step": 2535, "acc_step": 0, "speed/wps": 12851.772324328671, "speed/FLOPS": 201854744539410.62, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0019, "optim/grad_norm": 0.19800902903079987, "optim/lr": 0.0015204, "optim/total_tokens": 664535040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451547, "loss/out": 3.677140474319458, "created_at": "2025-01-14T16:34:41.250999+00:00"} {"global_step": 2536, "acc_step": 0, "speed/wps": 12844.067099185128, "speed/FLOPS": 201733723390442.38, "speed/curr_iter_time": 1.288, "speed/data_load_time": 0.001, "optim/grad_norm": 0.17520000040531158, "optim/lr": 0.001521, "optim/total_tokens": 664797184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331674, "loss/out": 3.580326557159424, "created_at": "2025-01-14T16:34:46.358105+00:00"} {"global_step": 2537, "acc_step": 0, "speed/wps": 12859.780210815632, "speed/FLOPS": 201980519400677.5, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16337274014949799, "optim/lr": 0.0015216, "optim/total_tokens": 665059328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.782644271850586, "created_at": "2025-01-14T16:34:51.460565+00:00"} {"global_step": 2538, "acc_step": 0, "speed/wps": 12861.518531817344, "speed/FLOPS": 202007822120712.78, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.16328884661197662, "optim/lr": 0.0015221999999999998, "optim/total_tokens": 665321472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465870, "loss/out": 3.7927167415618896, "created_at": "2025-01-14T16:34:56.560710+00:00"} {"global_step": 2539, "acc_step": 0, "speed/wps": 12855.048910930496, "speed/FLOPS": 201906207834494.1, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.20924779772758484, "optim/lr": 0.0015228000000000002, "optim/total_tokens": 665583616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 3.727661371231079, "created_at": "2025-01-14T16:35:01.662461+00:00"} {"global_step": 2540, "acc_step": 0, "speed/wps": 12857.280398843823, "speed/FLOPS": 201941256418558.62, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.22052840888500214, "optim/lr": 0.0015234, "optim/total_tokens": 665845760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329272, "loss/out": 3.7094311714172363, "created_at": "2025-01-14T16:35:06.763684+00:00"} {"global_step": 2541, "acc_step": 0, "speed/wps": 12859.929148103407, "speed/FLOPS": 201982858665443.3, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1678677648305893, "optim/lr": 0.001524, "optim/total_tokens": 666107904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481190, "loss/out": 3.734916925430298, "created_at": "2025-01-14T16:35:11.866968+00:00"} {"global_step": 2542, "acc_step": 0, "speed/wps": 12854.889073041744, "speed/FLOPS": 201903697360810.97, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.20463857054710388, "optim/lr": 0.0015246, "optim/total_tokens": 666370048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 3.564209461212158, "created_at": "2025-01-14T16:35:16.973424+00:00"} {"global_step": 2543, "acc_step": 0, "speed/wps": 12851.976853736574, "speed/FLOPS": 201857956954814.6, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17564032971858978, "optim/lr": 0.0015252, "optim/total_tokens": 666632192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439586, "loss/out": 3.715998649597168, "created_at": "2025-01-14T16:35:22.080234+00:00"} {"global_step": 2544, "acc_step": 0, "speed/wps": 12859.453660248893, "speed/FLOPS": 201975390475299.38, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17866241931915283, "optim/lr": 0.0015258000000000001, "optim/total_tokens": 666894336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365501, "loss/out": 3.6959261894226074, "created_at": "2025-01-14T16:35:27.180468+00:00"} {"global_step": 2545, "acc_step": 0, "speed/wps": 12861.021811319008, "speed/FLOPS": 202000020442720.9, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14615288376808167, "optim/lr": 0.0015264, "optim/total_tokens": 667156480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 3.664039134979248, "created_at": "2025-01-14T16:35:32.281877+00:00"} {"global_step": 2546, "acc_step": 0, "speed/wps": 12861.278222453873, "speed/FLOPS": 202004047731939.78, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14433299005031586, "optim/lr": 0.0015270000000000001, "optim/total_tokens": 667418624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380023, "loss/out": 3.6924638748168945, "created_at": "2025-01-14T16:35:37.381496+00:00"} {"global_step": 2547, "acc_step": 0, "speed/wps": 12857.271981696005, "speed/FLOPS": 201941124215685.78, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1456688940525055, "optim/lr": 0.0015276, "optim/total_tokens": 667680768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473073, "loss/out": 3.6923632621765137, "created_at": "2025-01-14T16:35:42.486456+00:00"} {"global_step": 2548, "acc_step": 0, "speed/wps": 12865.673324118905, "speed/FLOPS": 202073078843091.7, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14416037499904633, "optim/lr": 0.0015282, "optim/total_tokens": 667942912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 396035, "loss/out": 3.697591543197632, "created_at": "2025-01-14T16:35:47.584327+00:00"} {"global_step": 2549, "acc_step": 0, "speed/wps": 12861.88838681109, "speed/FLOPS": 202013631201623.56, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.14739005267620087, "optim/lr": 0.0015288000000000003, "optim/total_tokens": 668205056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452325, "loss/out": 3.7737674713134766, "created_at": "2025-01-14T16:35:52.684312+00:00"} {"global_step": 2550, "acc_step": 0, "speed/wps": 12867.208293071459, "speed/FLOPS": 202097187639759.94, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16395573318004608, "optim/lr": 0.0015294000000000002, "optim/total_tokens": 668467200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 3.6673338413238525, "created_at": "2025-01-14T16:35:57.784192+00:00"} {"global_step": 2551, "acc_step": 0, "speed/wps": 12868.434464858145, "speed/FLOPS": 202116446352606.44, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18152645230293274, "optim/lr": 0.0015300000000000001, "optim/total_tokens": 668729344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.7279186248779297, "created_at": "2025-01-14T16:36:02.881377+00:00"} {"global_step": 2552, "acc_step": 0, "speed/wps": 12858.066970387388, "speed/FLOPS": 201953610605512.88, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15370182693004608, "optim/lr": 0.0015306, "optim/total_tokens": 668991488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 3.6965625286102295, "created_at": "2025-01-14T16:36:07.989963+00:00"} {"global_step": 2553, "acc_step": 0, "speed/wps": 12863.881690545959, "speed/FLOPS": 202044938775866.06, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17927943170070648, "optim/lr": 0.0015312, "optim/total_tokens": 669253632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.6900038719177246, "created_at": "2025-01-14T16:36:13.097208+00:00"} {"global_step": 2554, "acc_step": 0, "speed/wps": 12862.834067890064, "speed/FLOPS": 202028484422470.84, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16343428194522858, "optim/lr": 0.0015318000000000003, "optim/total_tokens": 669515776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330329, "loss/out": 3.688286304473877, "created_at": "2025-01-14T16:36:18.196091+00:00"} {"global_step": 2555, "acc_step": 0, "speed/wps": 12858.475741038455, "speed/FLOPS": 201960030910297.62, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1845637559890747, "optim/lr": 0.0015324000000000002, "optim/total_tokens": 669777920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 3.545510768890381, "created_at": "2025-01-14T16:36:23.302762+00:00"} {"global_step": 2556, "acc_step": 0, "speed/wps": 12859.671883631938, "speed/FLOPS": 201978817973400.44, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1630895435810089, "optim/lr": 0.001533, "optim/total_tokens": 670040064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319751, "loss/out": 3.726685047149658, "created_at": "2025-01-14T16:36:28.402726+00:00"} {"global_step": 2557, "acc_step": 0, "speed/wps": 12852.998689901533, "speed/FLOPS": 201874006295935.3, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.16733697056770325, "optim/lr": 0.0015336, "optim/total_tokens": 670302208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446742, "loss/out": 3.689279079437256, "created_at": "2025-01-14T16:36:33.505097+00:00"} {"global_step": 2558, "acc_step": 0, "speed/wps": 12856.844575156629, "speed/FLOPS": 201934411208666.75, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18882620334625244, "optim/lr": 0.0015341999999999999, "optim/total_tokens": 670564352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357486, "loss/out": 3.626075506210327, "created_at": "2025-01-14T16:36:38.607322+00:00"} {"global_step": 2559, "acc_step": 0, "speed/wps": 12858.52510222095, "speed/FLOPS": 201960806195498.1, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.1633436381816864, "optim/lr": 0.0015348000000000002, "optim/total_tokens": 670826496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459528, "loss/out": 3.733213424682617, "created_at": "2025-01-14T16:36:43.707612+00:00"} {"global_step": 2560, "acc_step": 0, "speed/wps": 12862.633417782114, "speed/FLOPS": 202025332936803.34, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.17576336860656738, "optim/lr": 0.0015354000000000001, "optim/total_tokens": 671088640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363341, "loss/out": 3.5562050342559814, "created_at": "2025-01-14T16:36:48.806787+00:00"} {"global_step": 2561, "acc_step": 0, "speed/wps": 12865.803541820696, "speed/FLOPS": 202075124091036.22, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14747250080108643, "optim/lr": 0.001536, "optim/total_tokens": 671350784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452325, "loss/out": 3.6556034088134766, "created_at": "2025-01-14T16:36:53.904433+00:00"} {"global_step": 2562, "acc_step": 0, "speed/wps": 12864.619137185371, "speed/FLOPS": 202056521388695.9, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.15567421913146973, "optim/lr": 0.0015366, "optim/total_tokens": 671612928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313070, "loss/out": 3.671945810317993, "created_at": "2025-01-14T16:36:59.004607+00:00"} {"global_step": 2563, "acc_step": 0, "speed/wps": 12860.870853142926, "speed/FLOPS": 201997649437127.28, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15946799516677856, "optim/lr": 0.0015371999999999999, "optim/total_tokens": 671875072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 3.670971393585205, "created_at": "2025-01-14T16:37:04.106958+00:00"} {"global_step": 2564, "acc_step": 0, "speed/wps": 12859.626701430567, "speed/FLOPS": 201978108324841.06, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.16447053849697113, "optim/lr": 0.0015377999999999998, "optim/total_tokens": 672137216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.6337075233459473, "created_at": "2025-01-14T16:37:09.207088+00:00"} {"global_step": 2565, "acc_step": 0, "speed/wps": 12855.939695298372, "speed/FLOPS": 201920198827057.6, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18066485226154327, "optim/lr": 0.0015384, "optim/total_tokens": 672399360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 3.568679094314575, "created_at": "2025-01-14T16:37:14.310514+00:00"} {"global_step": 2566, "acc_step": 0, "speed/wps": 12863.625507155635, "speed/FLOPS": 202040915063688.16, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.17721711099147797, "optim/lr": 0.001539, "optim/total_tokens": 672661504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.647958517074585, "created_at": "2025-01-14T16:37:19.413166+00:00"} {"global_step": 2567, "acc_step": 0, "speed/wps": 12856.039670910204, "speed/FLOPS": 201921769081422.3, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.16374771296977997, "optim/lr": 0.0015396, "optim/total_tokens": 672923648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.726278305053711, "created_at": "2025-01-14T16:37:24.514865+00:00"} {"global_step": 2568, "acc_step": 0, "speed/wps": 12868.715957323258, "speed/FLOPS": 202120867578582.22, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.20454274117946625, "optim/lr": 0.0015402, "optim/total_tokens": 673185792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 3.576279640197754, "created_at": "2025-01-14T16:37:29.616869+00:00"} {"global_step": 2569, "acc_step": 0, "speed/wps": 12855.844341067406, "speed/FLOPS": 201918701157829.8, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.20773786306381226, "optim/lr": 0.0015408, "optim/total_tokens": 673447936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.695554494857788, "created_at": "2025-01-14T16:37:34.725437+00:00"} {"global_step": 2570, "acc_step": 0, "speed/wps": 12865.626841285997, "speed/FLOPS": 202072348766326.78, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19636237621307373, "optim/lr": 0.0015414, "optim/total_tokens": 673710080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325000, "loss/out": 3.6184191703796387, "created_at": "2025-01-14T16:37:39.823368+00:00"} {"global_step": 2571, "acc_step": 0, "speed/wps": 12862.545553627058, "speed/FLOPS": 202023952909510.2, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.209847554564476, "optim/lr": 0.001542, "optim/total_tokens": 673972224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 3.6587953567504883, "created_at": "2025-01-14T16:37:44.922757+00:00"} {"global_step": 2572, "acc_step": 0, "speed/wps": 12864.482969821127, "speed/FLOPS": 202054382693126.88, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17965872585773468, "optim/lr": 0.0015426, "optim/total_tokens": 674234368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463248, "loss/out": 3.595242977142334, "created_at": "2025-01-14T16:37:50.024090+00:00"} {"global_step": 2573, "acc_step": 0, "speed/wps": 12858.24822819839, "speed/FLOPS": 201956457508509.0, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.001, "optim/grad_norm": 0.21385616064071655, "optim/lr": 0.0015432, "optim/total_tokens": 674496512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 3.7431907653808594, "created_at": "2025-01-14T16:37:55.128208+00:00"} {"global_step": 2574, "acc_step": 0, "speed/wps": 12858.93062873718, "speed/FLOPS": 201967175546687.3, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.19776524603366852, "optim/lr": 0.0015438, "optim/total_tokens": 674758656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310132, "loss/out": 3.649688959121704, "created_at": "2025-01-14T16:38:00.233400+00:00"} {"global_step": 2575, "acc_step": 0, "speed/wps": 12864.596923773033, "speed/FLOPS": 202056172496530.72, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.16387198865413666, "optim/lr": 0.0015444000000000002, "optim/total_tokens": 675020800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 3.7145230770111084, "created_at": "2025-01-14T16:38:05.332221+00:00"} {"global_step": 2576, "acc_step": 0, "speed/wps": 12862.322805704373, "speed/FLOPS": 202020454347296.7, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1856684386730194, "optim/lr": 0.0015450000000000001, "optim/total_tokens": 675282944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469353, "loss/out": 3.6069562435150146, "created_at": "2025-01-14T16:38:10.432006+00:00"} {"global_step": 2577, "acc_step": 0, "speed/wps": 12856.91707609813, "speed/FLOPS": 201935549935579.75, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1736612617969513, "optim/lr": 0.0015456, "optim/total_tokens": 675545088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 3.6557936668395996, "created_at": "2025-01-14T16:38:15.532960+00:00"} {"global_step": 2578, "acc_step": 0, "speed/wps": 12846.16336032346, "speed/FLOPS": 201766648052184.56, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16802483797073364, "optim/lr": 0.0015462, "optim/total_tokens": 675807232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 3.7224373817443848, "created_at": "2025-01-14T16:38:20.639506+00:00"} {"global_step": 2579, "acc_step": 0, "speed/wps": 12857.451617251669, "speed/FLOPS": 201943945638932.6, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.1639900654554367, "optim/lr": 0.0015467999999999999, "optim/total_tokens": 676069376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.626272201538086, "created_at": "2025-01-14T16:38:25.744111+00:00"} {"global_step": 2580, "acc_step": 0, "speed/wps": 12853.654946237984, "speed/FLOPS": 201884313703490.0, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.19936798512935638, "optim/lr": 0.0015474000000000002, "optim/total_tokens": 676331520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451791, "loss/out": 3.674604892730713, "created_at": "2025-01-14T16:38:30.848531+00:00"} {"global_step": 2581, "acc_step": 0, "speed/wps": 12854.958696344358, "speed/FLOPS": 201904790890450.97, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20143723487854004, "optim/lr": 0.0015480000000000001, "optim/total_tokens": 676593664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481758, "loss/out": 3.593635320663452, "created_at": "2025-01-14T16:38:35.951267+00:00"} {"global_step": 2582, "acc_step": 0, "speed/wps": 12866.988931558253, "speed/FLOPS": 202093742265760.72, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.17802520096302032, "optim/lr": 0.0015486, "optim/total_tokens": 676855808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329565, "loss/out": 3.6850199699401855, "created_at": "2025-01-14T16:38:41.048840+00:00"} {"global_step": 2583, "acc_step": 0, "speed/wps": 12856.606639032127, "speed/FLOPS": 201930674094872.94, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.14828364551067352, "optim/lr": 0.0015492, "optim/total_tokens": 677117952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.6821470260620117, "created_at": "2025-01-14T16:38:46.158075+00:00"} {"global_step": 2584, "acc_step": 0, "speed/wps": 12862.197057149431, "speed/FLOPS": 202018479293444.72, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.16709065437316895, "optim/lr": 0.0015497999999999998, "optim/total_tokens": 677380096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294309, "loss/out": 3.7165560722351074, "created_at": "2025-01-14T16:38:51.261201+00:00"} {"global_step": 2585, "acc_step": 0, "speed/wps": 12862.605117074827, "speed/FLOPS": 202024888435306.16, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.16267040371894836, "optim/lr": 0.0015504000000000002, "optim/total_tokens": 677642240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 3.5925607681274414, "created_at": "2025-01-14T16:38:56.359877+00:00"} {"global_step": 2586, "acc_step": 0, "speed/wps": 12857.075221378884, "speed/FLOPS": 201938033824528.38, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17461688816547394, "optim/lr": 0.001551, "optim/total_tokens": 677904384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351470, "loss/out": 3.6158249378204346, "created_at": "2025-01-14T16:39:01.463137+00:00"} {"global_step": 2587, "acc_step": 0, "speed/wps": 12858.23890286012, "speed/FLOPS": 201956311041257.1, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.15214261412620544, "optim/lr": 0.0015516, "optim/total_tokens": 678166528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443835, "loss/out": 3.6611194610595703, "created_at": "2025-01-14T16:39:06.564364+00:00"} {"global_step": 2588, "acc_step": 0, "speed/wps": 12859.544064837553, "speed/FLOPS": 201976810403593.22, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.15705429017543793, "optim/lr": 0.0015521999999999999, "optim/total_tokens": 678428672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358303, "loss/out": 3.7241148948669434, "created_at": "2025-01-14T16:39:11.670608+00:00"} {"global_step": 2589, "acc_step": 0, "speed/wps": 12863.24740709202, "speed/FLOPS": 202034976482625.16, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.15888440608978271, "optim/lr": 0.0015527999999999998, "optim/total_tokens": 678690816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454995, "loss/out": 3.6872246265411377, "created_at": "2025-01-14T16:39:16.778349+00:00"} {"global_step": 2590, "acc_step": 0, "speed/wps": 12867.79613927895, "speed/FLOPS": 202106420572233.9, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1511824131011963, "optim/lr": 0.0015534000000000001, "optim/total_tokens": 678952960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470641, "loss/out": 3.7372798919677734, "created_at": "2025-01-14T16:39:21.875780+00:00"} {"global_step": 2591, "acc_step": 0, "speed/wps": 12861.20395401113, "speed/FLOPS": 202002881243990.97, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16960109770298004, "optim/lr": 0.001554, "optim/total_tokens": 679215104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.6600091457366943, "created_at": "2025-01-14T16:39:26.976568+00:00"} {"global_step": 2592, "acc_step": 0, "speed/wps": 12855.701106869366, "speed/FLOPS": 201916451467925.2, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20967833697795868, "optim/lr": 0.0015546, "optim/total_tokens": 679477248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454695, "loss/out": 3.6446995735168457, "created_at": "2025-01-14T16:39:32.080092+00:00"} {"global_step": 2593, "acc_step": 0, "speed/wps": 12858.227047082997, "speed/FLOPS": 201956124829985.84, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.23165671527385712, "optim/lr": 0.0015551999999999999, "optim/total_tokens": 679739392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.7337260246276855, "created_at": "2025-01-14T16:39:37.210903+00:00"} {"global_step": 2594, "acc_step": 0, "speed/wps": 12865.63719116153, "speed/FLOPS": 202072511325344.25, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24300150573253632, "optim/lr": 0.0015558, "optim/total_tokens": 680001536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 3.6987626552581787, "created_at": "2025-01-14T16:39:42.310107+00:00"} {"global_step": 2595, "acc_step": 0, "speed/wps": 12854.305387588402, "speed/FLOPS": 201894529778698.38, "speed/curr_iter_time": 1.2865, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.21561157703399658, "optim/lr": 0.0015564, "optim/total_tokens": 680263680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 3.6445484161376953, "created_at": "2025-01-14T16:39:47.412269+00:00"} {"global_step": 2596, "acc_step": 0, "speed/wps": 12857.741195185381, "speed/FLOPS": 201948493858303.3, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21083082258701324, "optim/lr": 0.001557, "optim/total_tokens": 680525824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 3.5935935974121094, "created_at": "2025-01-14T16:39:52.521738+00:00"} {"global_step": 2597, "acc_step": 0, "speed/wps": 12862.780575473464, "speed/FLOPS": 202027644250561.84, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.17214365303516388, "optim/lr": 0.0015576, "optim/total_tokens": 680787968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 3.7800471782684326, "created_at": "2025-01-14T16:39:57.620630+00:00"} {"global_step": 2598, "acc_step": 0, "speed/wps": 12862.378624208608, "speed/FLOPS": 202021331053608.38, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.23932461440563202, "optim/lr": 0.0015582, "optim/total_tokens": 681050112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495864, "loss/out": 3.7160143852233887, "created_at": "2025-01-14T16:40:02.722411+00:00"} {"global_step": 2599, "acc_step": 0, "speed/wps": 12853.7177120704, "speed/FLOPS": 201885299527137.38, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.14446653425693512, "optim/lr": 0.0015588, "optim/total_tokens": 681312256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 3.7455379962921143, "created_at": "2025-01-14T16:40:07.825390+00:00"} {"global_step": 2600, "acc_step": 0, "speed/wps": 12860.157354997593, "speed/FLOPS": 201986442968305.62, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1890851855278015, "optim/lr": 0.0015594, "optim/total_tokens": 681574400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.65677547454834, "created_at": "2025-01-14T16:40:12.925756+00:00"} {"global_step": 2601, "acc_step": 0, "speed/wps": 12834.53207126471, "speed/FLOPS": 201583962674450.8, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1967756152153015, "optim/lr": 0.0015600000000000002, "optim/total_tokens": 681836544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465633, "loss/out": 3.6354212760925293, "created_at": "2025-01-14T16:40:18.036746+00:00"} {"global_step": 2602, "acc_step": 0, "speed/wps": 12861.135123362494, "speed/FLOPS": 202001800164070.9, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1573532670736313, "optim/lr": 0.0015606, "optim/total_tokens": 682098688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.7933385372161865, "created_at": "2025-01-14T16:40:23.138340+00:00"} {"global_step": 2603, "acc_step": 0, "speed/wps": 12863.443147727618, "speed/FLOPS": 202038050858283.03, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1844566911458969, "optim/lr": 0.0015612, "optim/total_tokens": 682360832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.704512596130371, "created_at": "2025-01-14T16:40:28.240315+00:00"} {"global_step": 2604, "acc_step": 0, "speed/wps": 12861.881305977287, "speed/FLOPS": 202013519987398.62, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2020835280418396, "optim/lr": 0.0015618, "optim/total_tokens": 682622976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473839, "loss/out": 3.61936354637146, "created_at": "2025-01-14T16:40:33.339303+00:00"} {"global_step": 2605, "acc_step": 0, "speed/wps": 12857.607802261215, "speed/FLOPS": 201946398739128.22, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15028996765613556, "optim/lr": 0.0015624000000000002, "optim/total_tokens": 682885120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 3.69254994392395, "created_at": "2025-01-14T16:40:38.442676+00:00"} {"global_step": 2606, "acc_step": 0, "speed/wps": 12862.799900030808, "speed/FLOPS": 202027947769289.66, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14275559782981873, "optim/lr": 0.0015630000000000002, "optim/total_tokens": 683147264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 3.6171071529388428, "created_at": "2025-01-14T16:40:43.543449+00:00"} {"global_step": 2607, "acc_step": 0, "speed/wps": 12861.600215868037, "speed/FLOPS": 202009105080974.56, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16001293063163757, "optim/lr": 0.0015636, "optim/total_tokens": 683409408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496326, "loss/out": 3.6273906230926514, "created_at": "2025-01-14T16:40:48.649403+00:00"} {"global_step": 2608, "acc_step": 0, "speed/wps": 12858.590951399074, "speed/FLOPS": 201961840447326.78, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1636616289615631, "optim/lr": 0.0015642, "optim/total_tokens": 683671552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.6867754459381104, "created_at": "2025-01-14T16:40:53.755114+00:00"} {"global_step": 2609, "acc_step": 0, "speed/wps": 12853.728547591114, "speed/FLOPS": 201885469713879.75, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.15454460680484772, "optim/lr": 0.0015647999999999999, "optim/total_tokens": 683933696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460287, "loss/out": 3.581554889678955, "created_at": "2025-01-14T16:40:58.857284+00:00"} {"global_step": 2610, "acc_step": 0, "speed/wps": 12863.825656759134, "speed/FLOPS": 202044058688245.5, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3165341317653656, "optim/lr": 0.0015654000000000002, "optim/total_tokens": 684195840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.6793158054351807, "created_at": "2025-01-14T16:41:03.956561+00:00"} {"global_step": 2611, "acc_step": 0, "speed/wps": 12860.388063688042, "speed/FLOPS": 201990066565316.1, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4075619578361511, "optim/lr": 0.0015660000000000001, "optim/total_tokens": 684457984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466683, "loss/out": 3.7656798362731934, "created_at": "2025-01-14T16:41:09.060328+00:00"} {"global_step": 2612, "acc_step": 0, "speed/wps": 12857.648114218378, "speed/FLOPS": 201947031893809.9, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2700625956058502, "optim/lr": 0.0015666, "optim/total_tokens": 684720128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 3.8084309101104736, "created_at": "2025-01-14T16:41:14.164152+00:00"} {"global_step": 2613, "acc_step": 0, "speed/wps": 12861.562457904898, "speed/FLOPS": 202008512040278.6, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3476683497428894, "optim/lr": 0.0015672, "optim/total_tokens": 684982272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320810, "loss/out": 3.6603291034698486, "created_at": "2025-01-14T16:41:19.272856+00:00"} {"global_step": 2614, "acc_step": 0, "speed/wps": 12860.657520257722, "speed/FLOPS": 201994298751015.03, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.36746078729629517, "optim/lr": 0.0015677999999999998, "optim/total_tokens": 685244416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 3.7922441959381104, "created_at": "2025-01-14T16:41:24.375755+00:00"} {"global_step": 2615, "acc_step": 0, "speed/wps": 12865.894292118739, "speed/FLOPS": 202076549449170.94, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.27605077624320984, "optim/lr": 0.0015684000000000002, "optim/total_tokens": 685506560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311717, "loss/out": 3.877453327178955, "created_at": "2025-01-14T16:41:29.473664+00:00"} {"global_step": 2616, "acc_step": 0, "speed/wps": 12863.553941674412, "speed/FLOPS": 202039791029464.28, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2478710114955902, "optim/lr": 0.001569, "optim/total_tokens": 685768704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.7075352668762207, "created_at": "2025-01-14T16:41:34.575216+00:00"} {"global_step": 2617, "acc_step": 0, "speed/wps": 12862.511460184767, "speed/FLOPS": 202023417425149.72, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20760872960090637, "optim/lr": 0.0015696, "optim/total_tokens": 686030848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 3.658079147338867, "created_at": "2025-01-14T16:41:39.679679+00:00"} {"global_step": 2618, "acc_step": 0, "speed/wps": 12858.049363415925, "speed/FLOPS": 201953334063831.53, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.22574464976787567, "optim/lr": 0.0015702, "optim/total_tokens": 686292992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.694277763366699, "created_at": "2025-01-14T16:41:44.781186+00:00"} {"global_step": 2619, "acc_step": 0, "speed/wps": 12863.834308242365, "speed/FLOPS": 202044194571678.0, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.19296669960021973, "optim/lr": 0.0015707999999999998, "optim/total_tokens": 686555136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 3.795151710510254, "created_at": "2025-01-14T16:41:49.882770+00:00"} {"global_step": 2620, "acc_step": 0, "speed/wps": 12860.980899970475, "speed/FLOPS": 201999377873774.12, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.19277210533618927, "optim/lr": 0.0015714000000000001, "optim/total_tokens": 686817280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 3.6143593788146973, "created_at": "2025-01-14T16:41:54.986218+00:00"} {"global_step": 2621, "acc_step": 0, "speed/wps": 12863.427588240706, "speed/FLOPS": 202037806475160.1, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1669553816318512, "optim/lr": 0.001572, "optim/total_tokens": 687079424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451559, "loss/out": 3.717780590057373, "created_at": "2025-01-14T16:42:00.086575+00:00"} {"global_step": 2622, "acc_step": 0, "speed/wps": 12856.343414558944, "speed/FLOPS": 201926539792813.25, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.21767005324363708, "optim/lr": 0.0015726, "optim/total_tokens": 687341568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 3.6530098915100098, "created_at": "2025-01-14T16:42:05.191339+00:00"} {"global_step": 2623, "acc_step": 0, "speed/wps": 12858.772162890253, "speed/FLOPS": 201964686622806.4, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.23473688960075378, "optim/lr": 0.0015731999999999999, "optim/total_tokens": 687603712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320967, "loss/out": 3.770139217376709, "created_at": "2025-01-14T16:42:10.292831+00:00"} {"global_step": 2624, "acc_step": 0, "speed/wps": 12868.08099469204, "speed/FLOPS": 202110894617928.84, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.23380786180496216, "optim/lr": 0.0015738, "optim/total_tokens": 687865856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.6576805114746094, "created_at": "2025-01-14T16:42:15.393577+00:00"} {"global_step": 2625, "acc_step": 0, "speed/wps": 12865.726920575886, "speed/FLOPS": 202073920649097.84, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1788502186536789, "optim/lr": 0.0015744, "optim/total_tokens": 688128000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301810, "loss/out": 3.565192222595215, "created_at": "2025-01-14T16:42:20.495374+00:00"} {"global_step": 2626, "acc_step": 0, "speed/wps": 12861.798102062423, "speed/FLOPS": 202012213155581.5, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1796605885028839, "optim/lr": 0.001575, "optim/total_tokens": 688390144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 3.7519614696502686, "created_at": "2025-01-14T16:42:25.600893+00:00"} {"global_step": 2627, "acc_step": 0, "speed/wps": 12866.77503569881, "speed/FLOPS": 202090382737364.47, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.19578711688518524, "optim/lr": 0.0015756000000000001, "optim/total_tokens": 688652288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302985, "loss/out": 3.5293092727661133, "created_at": "2025-01-14T16:42:30.700685+00:00"} {"global_step": 2628, "acc_step": 0, "speed/wps": 12863.681057134903, "speed/FLOPS": 202041787552446.22, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.18656852841377258, "optim/lr": 0.0015762, "optim/total_tokens": 688914432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460034, "loss/out": 3.669307231903076, "created_at": "2025-01-14T16:42:35.804205+00:00"} {"global_step": 2629, "acc_step": 0, "speed/wps": 12862.63342248441, "speed/FLOPS": 202025333010659.38, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1911468356847763, "optim/lr": 0.0015768, "optim/total_tokens": 689176576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332731, "loss/out": 3.70530366897583, "created_at": "2025-01-14T16:42:40.902850+00:00"} {"global_step": 2630, "acc_step": 0, "speed/wps": 12863.587234079898, "speed/FLOPS": 202040313932440.97, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1865723878145218, "optim/lr": 0.0015774, "optim/total_tokens": 689438720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.691713333129883, "created_at": "2025-01-14T16:42:46.001209+00:00"} {"global_step": 2631, "acc_step": 0, "speed/wps": 12864.373853154904, "speed/FLOPS": 202052668865941.84, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1620108038187027, "optim/lr": 0.0015780000000000002, "optim/total_tokens": 689700864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.618473529815674, "created_at": "2025-01-14T16:42:51.103600+00:00"} {"global_step": 2632, "acc_step": 0, "speed/wps": 12866.639914304506, "speed/FLOPS": 202088260470190.66, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1691093146800995, "optim/lr": 0.0015786, "optim/total_tokens": 689963008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480426, "loss/out": 3.549614906311035, "created_at": "2025-01-14T16:42:56.201704+00:00"} {"global_step": 2633, "acc_step": 0, "speed/wps": 12864.237183008412, "speed/FLOPS": 202050522273486.72, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.18111106753349304, "optim/lr": 0.0015792, "optim/total_tokens": 690225152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 3.758976459503174, "created_at": "2025-01-14T16:43:01.300251+00:00"} {"global_step": 2634, "acc_step": 0, "speed/wps": 12866.260246705639, "speed/FLOPS": 202082297268832.5, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15302607417106628, "optim/lr": 0.0015798, "optim/total_tokens": 690487296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 3.71342134475708, "created_at": "2025-01-14T16:43:06.397867+00:00"} {"global_step": 2635, "acc_step": 0, "speed/wps": 12860.268748078859, "speed/FLOPS": 201988192549717.03, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.19762033224105835, "optim/lr": 0.0015804000000000003, "optim/total_tokens": 690749440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 3.7683091163635254, "created_at": "2025-01-14T16:43:11.506230+00:00"} {"global_step": 2636, "acc_step": 0, "speed/wps": 12862.60719312979, "speed/FLOPS": 202024921042602.16, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.16998913884162903, "optim/lr": 0.0015810000000000002, "optim/total_tokens": 691011584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.603131055831909, "created_at": "2025-01-14T16:43:16.606020+00:00"} {"global_step": 2637, "acc_step": 0, "speed/wps": 12865.82308011364, "speed/FLOPS": 202075430966775.3, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22120732069015503, "optim/lr": 0.0015816, "optim/total_tokens": 691273728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 3.6076762676239014, "created_at": "2025-01-14T16:43:21.709193+00:00"} {"global_step": 2638, "acc_step": 0, "speed/wps": 12862.423263843746, "speed/FLOPS": 202022032180419.5, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.20069682598114014, "optim/lr": 0.0015822, "optim/total_tokens": 691535872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.7565979957580566, "created_at": "2025-01-14T16:43:26.808801+00:00"} {"global_step": 2639, "acc_step": 0, "speed/wps": 12855.256649295836, "speed/FLOPS": 201909470650984.2, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17656344175338745, "optim/lr": 0.0015827999999999999, "optim/total_tokens": 691798016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 3.7063140869140625, "created_at": "2025-01-14T16:43:31.911520+00:00"} {"global_step": 2640, "acc_step": 0, "speed/wps": 12860.18184927661, "speed/FLOPS": 201986827684616.25, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1735425740480423, "optim/lr": 0.0015834000000000002, "optim/total_tokens": 692060160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 3.600219249725342, "created_at": "2025-01-14T16:43:37.011224+00:00"} {"global_step": 2641, "acc_step": 0, "speed/wps": 12863.020715259012, "speed/FLOPS": 202031415975880.53, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19433066248893738, "optim/lr": 0.0015840000000000001, "optim/total_tokens": 692322304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.660421371459961, "created_at": "2025-01-14T16:43:42.115340+00:00"} {"global_step": 2642, "acc_step": 0, "speed/wps": 12866.236834925345, "speed/FLOPS": 202081929554651.84, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18480578064918518, "optim/lr": 0.0015846, "optim/total_tokens": 692584448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476195, "loss/out": 3.657758951187134, "created_at": "2025-01-14T16:43:47.215143+00:00"} {"global_step": 2643, "acc_step": 0, "speed/wps": 12864.720012207563, "speed/FLOPS": 202058105769536.22, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15674443542957306, "optim/lr": 0.0015852, "optim/total_tokens": 692846592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456330, "loss/out": 3.6773250102996826, "created_at": "2025-01-14T16:43:52.314909+00:00"} {"global_step": 2644, "acc_step": 0, "speed/wps": 12864.51907277718, "speed/FLOPS": 202054949739662.38, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.18038450181484222, "optim/lr": 0.0015857999999999998, "optim/total_tokens": 693108736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358828, "loss/out": 3.656836748123169, "created_at": "2025-01-14T16:43:57.413215+00:00"} {"global_step": 2645, "acc_step": 0, "speed/wps": 12859.174466849647, "speed/FLOPS": 201971005359312.28, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15323863923549652, "optim/lr": 0.0015864000000000002, "optim/total_tokens": 693370880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444641, "loss/out": 3.6725008487701416, "created_at": "2025-01-14T16:44:02.513521+00:00"} {"global_step": 2646, "acc_step": 0, "speed/wps": 12865.539917102215, "speed/FLOPS": 202070983502574.12, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.15260209143161774, "optim/lr": 0.001587, "optim/total_tokens": 693633024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351996, "loss/out": 3.571176290512085, "created_at": "2025-01-14T16:44:07.611102+00:00"} {"global_step": 2647, "acc_step": 0, "speed/wps": 12853.661294873227, "speed/FLOPS": 201884413417530.4, "speed/curr_iter_time": 1.2862, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17317740619182587, "optim/lr": 0.0015876, "optim/total_tokens": 693895168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431320, "loss/out": 3.6380152702331543, "created_at": "2025-01-14T16:44:12.715515+00:00"} {"global_step": 2648, "acc_step": 0, "speed/wps": 12867.57864949872, "speed/FLOPS": 202103004596373.16, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.16698740422725677, "optim/lr": 0.0015882, "optim/total_tokens": 694157312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358828, "loss/out": 3.7636325359344482, "created_at": "2025-01-14T16:44:17.814046+00:00"} {"global_step": 2649, "acc_step": 0, "speed/wps": 12866.628777044181, "speed/FLOPS": 202088085544213.06, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21150891482830048, "optim/lr": 0.0015887999999999998, "optim/total_tokens": 694419456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444881, "loss/out": 3.6498594284057617, "created_at": "2025-01-14T16:44:22.915758+00:00"} {"global_step": 2650, "acc_step": 0, "speed/wps": 12857.994994501338, "speed/FLOPS": 201952480125317.22, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.24968397617340088, "optim/lr": 0.0015894000000000001, "optim/total_tokens": 694681600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 388029, "loss/out": 3.6564135551452637, "created_at": "2025-01-14T16:44:28.016377+00:00"} {"global_step": 2651, "acc_step": 0, "speed/wps": 12862.123260041684, "speed/FLOPS": 202017320208459.7, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.24966642260551453, "optim/lr": 0.00159, "optim/total_tokens": 694943744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461100, "loss/out": 3.8020119667053223, "created_at": "2025-01-14T16:44:33.119768+00:00"} {"global_step": 2652, "acc_step": 0, "speed/wps": 12860.827434783705, "speed/FLOPS": 201996967492132.62, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.253508597612381, "optim/lr": 0.0015906, "optim/total_tokens": 695205888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411602, "loss/out": 3.722822666168213, "created_at": "2025-01-14T16:44:38.220423+00:00"} {"global_step": 2653, "acc_step": 0, "speed/wps": 12862.833893899679, "speed/FLOPS": 202028481689712.75, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2467776983976364, "optim/lr": 0.0015912, "optim/total_tokens": 695468032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 3.627135992050171, "created_at": "2025-01-14T16:44:43.320075+00:00"} {"global_step": 2654, "acc_step": 0, "speed/wps": 12858.067635289977, "speed/FLOPS": 201953621048721.72, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.23211008310317993, "optim/lr": 0.0015918, "optim/total_tokens": 695730176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390782, "loss/out": 3.7990784645080566, "created_at": "2025-01-14T16:44:48.421011+00:00"} {"global_step": 2655, "acc_step": 0, "speed/wps": 12866.376227079943, "speed/FLOPS": 202084118899984.7, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.18506382405757904, "optim/lr": 0.0015924000000000001, "optim/total_tokens": 695992320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465057, "loss/out": 3.6834146976470947, "created_at": "2025-01-14T16:44:53.523682+00:00"} {"global_step": 2656, "acc_step": 0, "speed/wps": 12858.282400075086, "speed/FLOPS": 201956994224789.6, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.16738615930080414, "optim/lr": 0.001593, "optim/total_tokens": 696254464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 278898, "loss/out": 3.6632585525512695, "created_at": "2025-01-14T16:44:58.628179+00:00"} {"global_step": 2657, "acc_step": 0, "speed/wps": 12868.62501660175, "speed/FLOPS": 202119439229585.3, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15515513718128204, "optim/lr": 0.0015936000000000001, "optim/total_tokens": 696516608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 3.6478359699249268, "created_at": "2025-01-14T16:45:03.727118+00:00"} {"global_step": 2658, "acc_step": 0, "speed/wps": 12863.601571171846, "speed/FLOPS": 202040539116171.25, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1828126460313797, "optim/lr": 0.0015942, "optim/total_tokens": 696778752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448605, "loss/out": 3.6000256538391113, "created_at": "2025-01-14T16:45:08.827215+00:00"} {"global_step": 2659, "acc_step": 0, "speed/wps": 12858.16648803933, "speed/FLOPS": 201955173666988.22, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.16775110363960266, "optim/lr": 0.0015948, "optim/total_tokens": 697040896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 3.708733558654785, "created_at": "2025-01-14T16:45:13.930444+00:00"} {"global_step": 2660, "acc_step": 0, "speed/wps": 12860.179343908536, "speed/FLOPS": 201986788334367.9, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1770433485507965, "optim/lr": 0.0015954000000000003, "optim/total_tokens": 697303040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320721, "loss/out": 3.5946507453918457, "created_at": "2025-01-14T16:45:19.030018+00:00"} {"global_step": 2661, "acc_step": 0, "speed/wps": 12863.697934095744, "speed/FLOPS": 202042052628307.62, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18771164119243622, "optim/lr": 0.0015960000000000002, "optim/total_tokens": 697565184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.769103765487671, "created_at": "2025-01-14T16:45:24.128262+00:00"} {"global_step": 2662, "acc_step": 0, "speed/wps": 12849.711890068278, "speed/FLOPS": 201822382588017.78, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.17348070442676544, "optim/lr": 0.0015966, "optim/total_tokens": 697827328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307153, "loss/out": 3.649035692214966, "created_at": "2025-01-14T16:45:29.236343+00:00"} {"global_step": 2663, "acc_step": 0, "speed/wps": 12860.18736061976, "speed/FLOPS": 201986914247833.8, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.14733169972896576, "optim/lr": 0.0015972, "optim/total_tokens": 698089472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 3.576460599899292, "created_at": "2025-01-14T16:45:34.339921+00:00"} {"global_step": 2664, "acc_step": 0, "speed/wps": 12861.237709082448, "speed/FLOPS": 202003411413770.38, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1598701775074005, "optim/lr": 0.0015978, "optim/total_tokens": 698351616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323215, "loss/out": 3.623422861099243, "created_at": "2025-01-14T16:45:39.440306+00:00"} {"global_step": 2665, "acc_step": 0, "speed/wps": 12860.341412210513, "speed/FLOPS": 201989333839755.66, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.15585999190807343, "optim/lr": 0.0015984000000000003, "optim/total_tokens": 698613760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441971, "loss/out": 3.635032892227173, "created_at": "2025-01-14T16:45:44.539845+00:00"} {"global_step": 2666, "acc_step": 0, "speed/wps": 12861.124752521988, "speed/FLOPS": 202001637275769.75, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15128225088119507, "optim/lr": 0.0015990000000000002, "optim/total_tokens": 698875904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361473, "loss/out": 3.6378629207611084, "created_at": "2025-01-14T16:45:49.640642+00:00"} {"global_step": 2667, "acc_step": 0, "speed/wps": 12867.25245802343, "speed/FLOPS": 202097881311019.62, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15816402435302734, "optim/lr": 0.0015996, "optim/total_tokens": 699138048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450225, "loss/out": 3.7357559204101562, "created_at": "2025-01-14T16:45:54.738580+00:00"} {"global_step": 2668, "acc_step": 0, "speed/wps": 12862.691996903546, "speed/FLOPS": 202026253002401.6, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.17512981593608856, "optim/lr": 0.0016002, "optim/total_tokens": 699400192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307153, "loss/out": 3.598276376724243, "created_at": "2025-01-14T16:45:59.838396+00:00"} {"global_step": 2669, "acc_step": 0, "speed/wps": 12858.416666953124, "speed/FLOPS": 201959103070611.22, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.16660699248313904, "optim/lr": 0.0016007999999999999, "optim/total_tokens": 699662336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.5927348136901855, "created_at": "2025-01-14T16:46:04.938782+00:00"} {"global_step": 2670, "acc_step": 0, "speed/wps": 12869.495657963713, "speed/FLOPS": 202133113848561.53, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.16714252531528473, "optim/lr": 0.0016014000000000002, "optim/total_tokens": 699924480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481758, "loss/out": 3.694530963897705, "created_at": "2025-01-14T16:46:10.034754+00:00"} {"global_step": 2671, "acc_step": 0, "speed/wps": 12857.665121011247, "speed/FLOPS": 201947299008861.7, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20293475687503815, "optim/lr": 0.0016020000000000001, "optim/total_tokens": 700186624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 3.6448721885681152, "created_at": "2025-01-14T16:46:15.137823+00:00"} {"global_step": 2672, "acc_step": 0, "speed/wps": 12864.282037795547, "speed/FLOPS": 202051226779555.6, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1828262060880661, "optim/lr": 0.0016026, "optim/total_tokens": 700448768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 3.7685513496398926, "created_at": "2025-01-14T16:46:20.238564+00:00"} {"global_step": 2673, "acc_step": 0, "speed/wps": 12870.083913273027, "speed/FLOPS": 202142353206541.8, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.15497438609600067, "optim/lr": 0.0016032, "optim/total_tokens": 700710912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 3.705413341522217, "created_at": "2025-01-14T16:46:25.338466+00:00"} {"global_step": 2674, "acc_step": 0, "speed/wps": 12864.859776035026, "speed/FLOPS": 202060300952502.16, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1489962786436081, "optim/lr": 0.0016037999999999998, "optim/total_tokens": 700973056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.7050437927246094, "created_at": "2025-01-14T16:46:30.438524+00:00"} {"global_step": 2675, "acc_step": 0, "speed/wps": 12861.909022831362, "speed/FLOPS": 202013955318678.8, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17093297839164734, "optim/lr": 0.0016044000000000002, "optim/total_tokens": 701235200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 3.6605587005615234, "created_at": "2025-01-14T16:46:35.537536+00:00"} {"global_step": 2676, "acc_step": 0, "speed/wps": 12861.78571078235, "speed/FLOPS": 202012018533500.56, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1819811463356018, "optim/lr": 0.001605, "optim/total_tokens": 701497344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 3.7550764083862305, "created_at": "2025-01-14T16:46:40.644846+00:00"} {"global_step": 2677, "acc_step": 0, "speed/wps": 12842.72410337028, "speed/FLOPS": 201712629795700.53, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1627490222454071, "optim/lr": 0.0016056, "optim/total_tokens": 701759488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 3.594670295715332, "created_at": "2025-01-14T16:46:45.754427+00:00"} {"global_step": 2678, "acc_step": 0, "speed/wps": 12850.400781544753, "speed/FLOPS": 201833202575291.9, "speed/curr_iter_time": 1.2872, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1580481380224228, "optim/lr": 0.0016062, "optim/total_tokens": 702021632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290365, "loss/out": 3.5465948581695557, "created_at": "2025-01-14T16:46:50.862231+00:00"} {"global_step": 2679, "acc_step": 0, "speed/wps": 12866.901581949529, "speed/FLOPS": 202092370320124.78, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.16927912831306458, "optim/lr": 0.0016068, "optim/total_tokens": 702283776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461622, "loss/out": 3.666322946548462, "created_at": "2025-01-14T16:46:55.959828+00:00"} {"global_step": 2680, "acc_step": 0, "speed/wps": 12864.85339751706, "speed/FLOPS": 202060200769112.5, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1688755601644516, "optim/lr": 0.0016074000000000001, "optim/total_tokens": 702545920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308499, "loss/out": 3.6219496726989746, "created_at": "2025-01-14T16:47:01.057800+00:00"} {"global_step": 2681, "acc_step": 0, "speed/wps": 12865.657834508456, "speed/FLOPS": 202072835557474.66, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14473329484462738, "optim/lr": 0.001608, "optim/total_tokens": 702808064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 3.642979621887207, "created_at": "2025-01-14T16:47:06.155466+00:00"} {"global_step": 2682, "acc_step": 0, "speed/wps": 12868.535018487917, "speed/FLOPS": 202118025685537.2, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1561579555273056, "optim/lr": 0.0016086, "optim/total_tokens": 703070208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 3.686216354370117, "created_at": "2025-01-14T16:47:11.254961+00:00"} {"global_step": 2683, "acc_step": 0, "speed/wps": 12865.980251913928, "speed/FLOPS": 202077899565875.53, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.16248227655887604, "optim/lr": 0.0016092, "optim/total_tokens": 703332352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488197, "loss/out": 3.5918679237365723, "created_at": "2025-01-14T16:47:16.359023+00:00"} {"global_step": 2684, "acc_step": 0, "speed/wps": 12861.433308634902, "speed/FLOPS": 202006483573523.6, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19865736365318298, "optim/lr": 0.0016098, "optim/total_tokens": 703594496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 3.638019561767578, "created_at": "2025-01-14T16:47:21.461777+00:00"} {"global_step": 2685, "acc_step": 0, "speed/wps": 12867.831321967113, "speed/FLOPS": 202106973164697.6, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1768716424703598, "optim/lr": 0.0016104000000000001, "optim/total_tokens": 703856640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 3.683345317840576, "created_at": "2025-01-14T16:47:26.559249+00:00"} {"global_step": 2686, "acc_step": 0, "speed/wps": 12858.217519715521, "speed/FLOPS": 201955975189587.66, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18543854355812073, "optim/lr": 0.0016110000000000002, "optim/total_tokens": 704118784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.6828253269195557, "created_at": "2025-01-14T16:47:31.663159+00:00"} {"global_step": 2687, "acc_step": 0, "speed/wps": 12866.88685165725, "speed/FLOPS": 202092138960642.97, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17444905638694763, "optim/lr": 0.0016116000000000001, "optim/total_tokens": 704380928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 3.661590576171875, "created_at": "2025-01-14T16:47:36.765218+00:00"} {"global_step": 2688, "acc_step": 0, "speed/wps": 12866.272289032162, "speed/FLOPS": 202082486410118.44, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18251395225524902, "optim/lr": 0.0016122, "optim/total_tokens": 704643072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 3.528245687484741, "created_at": "2025-01-14T16:47:41.870307+00:00"} {"global_step": 2689, "acc_step": 0, "speed/wps": 12867.860005694864, "speed/FLOPS": 202107423682057.5, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.14877720177173615, "optim/lr": 0.0016128, "optim/total_tokens": 704905216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.722548484802246, "created_at": "2025-01-14T16:47:46.966925+00:00"} {"global_step": 2690, "acc_step": 0, "speed/wps": 12867.96679782003, "speed/FLOPS": 202109100999131.0, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16217193007469177, "optim/lr": 0.0016133999999999999, "optim/total_tokens": 705167360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 3.6759371757507324, "created_at": "2025-01-14T16:47:52.063575+00:00"} {"global_step": 2691, "acc_step": 0, "speed/wps": 12848.978978585688, "speed/FLOPS": 201810871206057.78, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1645243763923645, "optim/lr": 0.0016140000000000002, "optim/total_tokens": 705429504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 3.589538335800171, "created_at": "2025-01-14T16:47:57.167610+00:00"} {"global_step": 2692, "acc_step": 0, "speed/wps": 12865.649867429498, "speed/FLOPS": 202072710423551.7, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1464395523071289, "optim/lr": 0.0016146, "optim/total_tokens": 705691648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456840, "loss/out": 3.7390737533569336, "created_at": "2025-01-14T16:48:02.268676+00:00"} {"global_step": 2693, "acc_step": 0, "speed/wps": 12865.230227783657, "speed/FLOPS": 202066119406268.47, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17352394759655, "optim/lr": 0.0016152, "optim/total_tokens": 705953792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.7118093967437744, "created_at": "2025-01-14T16:48:07.366409+00:00"} {"global_step": 2694, "acc_step": 0, "speed/wps": 12861.457662030156, "speed/FLOPS": 202006866077061.12, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18699896335601807, "optim/lr": 0.0016158, "optim/total_tokens": 706215936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 3.7037158012390137, "created_at": "2025-01-14T16:48:12.466340+00:00"} {"global_step": 2695, "acc_step": 0, "speed/wps": 12855.498967148822, "speed/FLOPS": 201913276585844.97, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.26713940501213074, "optim/lr": 0.0016163999999999998, "optim/total_tokens": 706478080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 3.6534767150878906, "created_at": "2025-01-14T16:48:17.572678+00:00"} {"global_step": 2696, "acc_step": 0, "speed/wps": 12867.087180404807, "speed/FLOPS": 202095285398903.94, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2036939561367035, "optim/lr": 0.0016170000000000002, "optim/total_tokens": 706740224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462672, "loss/out": 3.676292657852173, "created_at": "2025-01-14T16:48:22.672988+00:00"} {"global_step": 2697, "acc_step": 0, "speed/wps": 12865.540272286049, "speed/FLOPS": 202070989081224.3, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2089512199163437, "optim/lr": 0.0016176, "optim/total_tokens": 707002368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413950, "loss/out": 3.6128740310668945, "created_at": "2025-01-14T16:48:27.774001+00:00"} {"global_step": 2698, "acc_step": 0, "speed/wps": 12860.983605447193, "speed/FLOPS": 201999420367003.7, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.19888144731521606, "optim/lr": 0.0016182, "optim/total_tokens": 707264512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472504, "loss/out": 3.65885329246521, "created_at": "2025-01-14T16:48:32.878606+00:00"} {"global_step": 2699, "acc_step": 0, "speed/wps": 12862.118325375197, "speed/FLOPS": 202017242702741.56, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.18981333076953888, "optim/lr": 0.0016187999999999999, "optim/total_tokens": 707526656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483949, "loss/out": 3.6416704654693604, "created_at": "2025-01-14T16:48:37.978414+00:00"} {"global_step": 2700, "acc_step": 0, "speed/wps": 12845.369580236153, "speed/FLOPS": 201754180645145.25, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.19833511114120483, "optim/lr": 0.0016193999999999998, "optim/total_tokens": 707788800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 3.670750617980957, "created_at": "2025-01-14T16:48:43.088904+00:00"} {"global_step": 2701, "acc_step": 0, "speed/wps": 12864.262786392082, "speed/FLOPS": 202050924409810.0, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.1622329205274582, "optim/lr": 0.0016200000000000001, "optim/total_tokens": 708050944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.762712240219116, "created_at": "2025-01-14T16:48:48.189223+00:00"} {"global_step": 2702, "acc_step": 0, "speed/wps": 12864.893705529299, "speed/FLOPS": 202060833861833.72, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16689659655094147, "optim/lr": 0.0016206, "optim/total_tokens": 708313088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 3.7183456420898438, "created_at": "2025-01-14T16:48:53.291309+00:00"} {"global_step": 2703, "acc_step": 0, "speed/wps": 12864.376153200456, "speed/FLOPS": 202052704991317.8, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.16077880561351776, "optim/lr": 0.0016212, "optim/total_tokens": 708575232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.769432783126831, "created_at": "2025-01-14T16:48:58.392585+00:00"} {"global_step": 2704, "acc_step": 0, "speed/wps": 12870.926388826741, "speed/FLOPS": 202155585442795.62, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14752046763896942, "optim/lr": 0.0016217999999999998, "optim/total_tokens": 708837376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 3.659846544265747, "created_at": "2025-01-14T16:49:03.488192+00:00"} {"global_step": 2705, "acc_step": 0, "speed/wps": 12865.240675809335, "speed/FLOPS": 202066283506868.8, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1565869152545929, "optim/lr": 0.0016224, "optim/total_tokens": 709099520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 3.690706253051758, "created_at": "2025-01-14T16:49:08.585874+00:00"} {"global_step": 2706, "acc_step": 0, "speed/wps": 12856.631754004848, "speed/FLOPS": 201931068560031.1, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.14746101200580597, "optim/lr": 0.001623, "optim/total_tokens": 709361664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455791, "loss/out": 3.8192543983459473, "created_at": "2025-01-14T16:49:13.690658+00:00"} {"global_step": 2707, "acc_step": 0, "speed/wps": 12866.324095523813, "speed/FLOPS": 202083300102259.4, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16169634461402893, "optim/lr": 0.0016236, "optim/total_tokens": 709623808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 3.6129751205444336, "created_at": "2025-01-14T16:49:18.789041+00:00"} {"global_step": 2708, "acc_step": 0, "speed/wps": 12863.972611271533, "speed/FLOPS": 202046366810799.38, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16215087473392487, "optim/lr": 0.0016242, "optim/total_tokens": 709885952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 3.720371723175049, "created_at": "2025-01-14T16:49:23.889264+00:00"} {"global_step": 2709, "acc_step": 0, "speed/wps": 12867.744170339143, "speed/FLOPS": 202105604328621.44, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15567025542259216, "optim/lr": 0.0016248, "optim/total_tokens": 710148096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296001, "loss/out": 3.685488224029541, "created_at": "2025-01-14T16:49:28.985791+00:00"} {"global_step": 2710, "acc_step": 0, "speed/wps": 12868.495579552193, "speed/FLOPS": 202117406242856.84, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.17176200449466705, "optim/lr": 0.0016254, "optim/total_tokens": 710410240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 3.640350341796875, "created_at": "2025-01-14T16:49:34.082183+00:00"} {"global_step": 2711, "acc_step": 0, "speed/wps": 12867.735332413964, "speed/FLOPS": 202105465516861.94, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2053469568490982, "optim/lr": 0.001626, "optim/total_tokens": 710672384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311352, "loss/out": 3.583585500717163, "created_at": "2025-01-14T16:49:39.181640+00:00"} {"global_step": 2712, "acc_step": 0, "speed/wps": 12853.024647667507, "speed/FLOPS": 201874413998319.66, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.2312096655368805, "optim/lr": 0.0016266000000000002, "optim/total_tokens": 710934528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460578, "loss/out": 3.665163040161133, "created_at": "2025-01-14T16:49:44.287871+00:00"} {"global_step": 2713, "acc_step": 0, "speed/wps": 12864.211201098933, "speed/FLOPS": 202050114191895.47, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1789526790380478, "optim/lr": 0.0016272, "optim/total_tokens": 711196672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336478, "loss/out": 3.7015175819396973, "created_at": "2025-01-14T16:49:49.387753+00:00"} {"global_step": 2714, "acc_step": 0, "speed/wps": 12868.71068812493, "speed/FLOPS": 202120784818581.84, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.15985213220119476, "optim/lr": 0.0016278, "optim/total_tokens": 711458816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446220, "loss/out": 3.627796173095703, "created_at": "2025-01-14T16:49:54.487665+00:00"} {"global_step": 2715, "acc_step": 0, "speed/wps": 12861.973740622658, "speed/FLOPS": 202014971800522.7, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1816384345293045, "optim/lr": 0.0016284, "optim/total_tokens": 711720960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359885, "loss/out": 3.698370933532715, "created_at": "2025-01-14T16:49:59.588082+00:00"} {"global_step": 2716, "acc_step": 0, "speed/wps": 12862.78108098189, "speed/FLOPS": 202027652190266.3, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.16637781262397766, "optim/lr": 0.0016290000000000002, "optim/total_tokens": 711983104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459256, "loss/out": 3.6773548126220703, "created_at": "2025-01-14T16:50:04.691634+00:00"} {"global_step": 2717, "acc_step": 0, "speed/wps": 12860.999937054545, "speed/FLOPS": 201999676877339.12, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2169814556837082, "optim/lr": 0.0016296000000000001, "optim/total_tokens": 712245248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401934, "loss/out": 3.539721965789795, "created_at": "2025-01-14T16:50:09.791566+00:00"} {"global_step": 2718, "acc_step": 0, "speed/wps": 12863.775260160595, "speed/FLOPS": 202043267140413.25, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16428786516189575, "optim/lr": 0.0016302, "optim/total_tokens": 712507392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464820, "loss/out": 3.668492317199707, "created_at": "2025-01-14T16:50:14.906076+00:00"} {"global_step": 2719, "acc_step": 0, "speed/wps": 12860.034800394955, "speed/FLOPS": 201984518079863.53, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.17997898161411285, "optim/lr": 0.0016308, "optim/total_tokens": 712769536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 3.6780734062194824, "created_at": "2025-01-14T16:50:20.009485+00:00"} {"global_step": 2720, "acc_step": 0, "speed/wps": 12864.213329410346, "speed/FLOPS": 202050147619950.8, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17578648030757904, "optim/lr": 0.0016313999999999999, "optim/total_tokens": 713031680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 3.7119007110595703, "created_at": "2025-01-14T16:50:25.107577+00:00"} {"global_step": 2721, "acc_step": 0, "speed/wps": 12861.064088611412, "speed/FLOPS": 202000684465693.22, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15504510700702667, "optim/lr": 0.0016320000000000002, "optim/total_tokens": 713293824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464032, "loss/out": 3.700502395629883, "created_at": "2025-01-14T16:50:30.211056+00:00"} {"global_step": 2722, "acc_step": 0, "speed/wps": 12862.811886505217, "speed/FLOPS": 202028136033341.4, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1902361661195755, "optim/lr": 0.0016326000000000001, "optim/total_tokens": 713555968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454466, "loss/out": 3.6924519538879395, "created_at": "2025-01-14T16:50:35.313685+00:00"} {"global_step": 2723, "acc_step": 0, "speed/wps": 12868.961585827501, "speed/FLOPS": 202124725511770.3, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1578705906867981, "optim/lr": 0.0016332, "optim/total_tokens": 713818112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296475, "loss/out": 3.622070074081421, "created_at": "2025-01-14T16:50:40.410666+00:00"} {"global_step": 2724, "acc_step": 0, "speed/wps": 12861.838388580412, "speed/FLOPS": 202012845910706.0, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1844332367181778, "optim/lr": 0.0016338, "optim/total_tokens": 714080256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463770, "loss/out": 3.682227611541748, "created_at": "2025-01-14T16:50:45.511418+00:00"} {"global_step": 2725, "acc_step": 0, "speed/wps": 12861.66601661863, "speed/FLOPS": 202010138572182.34, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15767835080623627, "optim/lr": 0.0016343999999999998, "optim/total_tokens": 714342400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300577, "loss/out": 3.619389533996582, "created_at": "2025-01-14T16:50:50.616068+00:00"} {"global_step": 2726, "acc_step": 0, "speed/wps": 12863.0631326647, "speed/FLOPS": 202032082199524.47, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16247880458831787, "optim/lr": 0.0016350000000000002, "optim/total_tokens": 714604544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439302, "loss/out": 3.624786376953125, "created_at": "2025-01-14T16:50:55.720355+00:00"} {"global_step": 2727, "acc_step": 0, "speed/wps": 12861.39194098435, "speed/FLOPS": 202005833837726.56, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.14824096858501434, "optim/lr": 0.0016356, "optim/total_tokens": 714866688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 373889, "loss/out": 3.711686134338379, "created_at": "2025-01-14T16:51:00.821017+00:00"} {"global_step": 2728, "acc_step": 0, "speed/wps": 12858.599797914534, "speed/FLOPS": 201961979394008.47, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16394208371639252, "optim/lr": 0.0016362, "optim/total_tokens": 715128832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450990, "loss/out": 3.6257827281951904, "created_at": "2025-01-14T16:51:05.923690+00:00"} {"global_step": 2729, "acc_step": 0, "speed/wps": 12855.291589660827, "speed/FLOPS": 201910019437429.47, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.16249437630176544, "optim/lr": 0.0016367999999999999, "optim/total_tokens": 715390976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413463, "loss/out": 3.6160621643066406, "created_at": "2025-01-14T16:51:11.025512+00:00"} {"global_step": 2730, "acc_step": 0, "speed/wps": 12856.359027378303, "speed/FLOPS": 201926785013595.56, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.001, "optim/grad_norm": 0.1697821021080017, "optim/lr": 0.0016373999999999998, "optim/total_tokens": 715653120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468018, "loss/out": 3.6392061710357666, "created_at": "2025-01-14T16:51:16.129322+00:00"} {"global_step": 2731, "acc_step": 0, "speed/wps": 12864.931532445982, "speed/FLOPS": 202061427985539.88, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17480650544166565, "optim/lr": 0.0016380000000000001, "optim/total_tokens": 715915264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431014, "loss/out": 3.7763543128967285, "created_at": "2025-01-14T16:51:21.230069+00:00"} {"global_step": 2732, "acc_step": 0, "speed/wps": 12845.794964902141, "speed/FLOPS": 201760861895864.38, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1510850191116333, "optim/lr": 0.0016386, "optim/total_tokens": 716177408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.6490252017974854, "created_at": "2025-01-14T16:51:26.335963+00:00"} {"global_step": 2733, "acc_step": 0, "speed/wps": 12852.718969169784, "speed/FLOPS": 201869612897466.62, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2043330818414688, "optim/lr": 0.0016392, "optim/total_tokens": 716439552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.573585033416748, "created_at": "2025-01-14T16:51:31.440462+00:00"} {"global_step": 2734, "acc_step": 0, "speed/wps": 12857.116791480184, "speed/FLOPS": 201938686740092.6, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18594740331172943, "optim/lr": 0.0016397999999999999, "optim/total_tokens": 716701696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 3.6622488498687744, "created_at": "2025-01-14T16:51:36.541399+00:00"} {"global_step": 2735, "acc_step": 0, "speed/wps": 12862.38879017822, "speed/FLOPS": 202021490724130.62, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15774966776371002, "optim/lr": 0.0016404, "optim/total_tokens": 716963840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 386926, "loss/out": 3.6077020168304443, "created_at": "2025-01-14T16:51:41.643133+00:00"} {"global_step": 2736, "acc_step": 0, "speed/wps": 12861.822757748201, "speed/FLOPS": 202012600407007.12, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18147079646587372, "optim/lr": 0.001641, "optim/total_tokens": 717225984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449411, "loss/out": 3.6659317016601562, "created_at": "2025-01-14T16:51:46.745011+00:00"} {"global_step": 2737, "acc_step": 0, "speed/wps": 12859.665435055322, "speed/FLOPS": 201978716689643.4, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1847471445798874, "optim/lr": 0.0016416, "optim/total_tokens": 717488128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368810, "loss/out": 3.6712803840637207, "created_at": "2025-01-14T16:51:51.848867+00:00"} {"global_step": 2738, "acc_step": 0, "speed/wps": 12863.69395764745, "speed/FLOPS": 202041990172722.94, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.18358924984931946, "optim/lr": 0.0016422000000000001, "optim/total_tokens": 717750272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459237, "loss/out": 3.6042118072509766, "created_at": "2025-01-14T16:51:56.947066+00:00"} {"global_step": 2739, "acc_step": 0, "speed/wps": 12869.509379801237, "speed/FLOPS": 202133329368875.62, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15395493805408478, "optim/lr": 0.0016428, "optim/total_tokens": 718012416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367469, "loss/out": 3.6523144245147705, "created_at": "2025-01-14T16:52:02.043568+00:00"} {"global_step": 2740, "acc_step": 0, "speed/wps": 12864.008128153697, "speed/FLOPS": 202046924652239.06, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.20118147134780884, "optim/lr": 0.0016434, "optim/total_tokens": 718274560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445404, "loss/out": 3.6680898666381836, "created_at": "2025-01-14T16:52:07.144913+00:00"} {"global_step": 2741, "acc_step": 0, "speed/wps": 12854.534118172574, "speed/FLOPS": 201898122306831.56, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.14775538444519043, "optim/lr": 0.001644, "optim/total_tokens": 718536704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 408061, "loss/out": 3.6706387996673584, "created_at": "2025-01-14T16:52:12.246676+00:00"} {"global_step": 2742, "acc_step": 0, "speed/wps": 12859.88720729247, "speed/FLOPS": 201982199927374.8, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18455545604228973, "optim/lr": 0.0016446000000000002, "optim/total_tokens": 718798848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455517, "loss/out": 3.6485373973846436, "created_at": "2025-01-14T16:52:17.349396+00:00"} {"global_step": 2743, "acc_step": 0, "speed/wps": 12858.705968798107, "speed/FLOPS": 201963646953629.06, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18839147686958313, "optim/lr": 0.0016452, "optim/total_tokens": 719060992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446459, "loss/out": 3.684648036956787, "created_at": "2025-01-14T16:52:22.449833+00:00"} {"global_step": 2744, "acc_step": 0, "speed/wps": 12863.028279362357, "speed/FLOPS": 202031534780517.75, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.15677012503147125, "optim/lr": 0.0016458, "optim/total_tokens": 719323136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 3.6615524291992188, "created_at": "2025-01-14T16:52:27.551746+00:00"} {"global_step": 2745, "acc_step": 0, "speed/wps": 12848.65670654121, "speed/FLOPS": 201805809480751.03, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.14950649440288544, "optim/lr": 0.0016464, "optim/total_tokens": 719585280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 3.6774795055389404, "created_at": "2025-01-14T16:52:32.659347+00:00"} {"global_step": 2746, "acc_step": 0, "speed/wps": 12864.4930121187, "speed/FLOPS": 202054540421209.84, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.16506396234035492, "optim/lr": 0.0016470000000000002, "optim/total_tokens": 719847424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 3.6086783409118652, "created_at": "2025-01-14T16:52:37.757303+00:00"} {"global_step": 2747, "acc_step": 0, "speed/wps": 12865.14475783172, "speed/FLOPS": 202064776983225.28, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.18981699645519257, "optim/lr": 0.0016476000000000001, "optim/total_tokens": 720109568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 3.6080994606018066, "created_at": "2025-01-14T16:52:42.863129+00:00"} {"global_step": 2748, "acc_step": 0, "speed/wps": 12869.603016236244, "speed/FLOPS": 202134800057757.34, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17208555340766907, "optim/lr": 0.0016482, "optim/total_tokens": 720371712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449942, "loss/out": 3.625934362411499, "created_at": "2025-01-14T16:52:47.961876+00:00"} {"global_step": 2749, "acc_step": 0, "speed/wps": 12864.499665443456, "speed/FLOPS": 202054644920817.75, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18593446910381317, "optim/lr": 0.0016488, "optim/total_tokens": 720633856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296475, "loss/out": 3.762334108352661, "created_at": "2025-01-14T16:52:53.060266+00:00"} {"global_step": 2750, "acc_step": 0, "speed/wps": 12866.469156860987, "speed/FLOPS": 202085578489894.88, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24062518775463104, "optim/lr": 0.0016493999999999999, "optim/total_tokens": 720896000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 3.7768614292144775, "created_at": "2025-01-14T16:52:58.158013+00:00"} {"global_step": 2751, "acc_step": 0, "speed/wps": 12866.259204559667, "speed/FLOPS": 202082280900497.97, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20933498442173004, "optim/lr": 0.0016500000000000002, "optim/total_tokens": 721158144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317794, "loss/out": 3.7949845790863037, "created_at": "2025-01-14T16:53:03.258189+00:00"} {"global_step": 2752, "acc_step": 0, "speed/wps": 12863.23405130977, "speed/FLOPS": 202034766711712.12, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15602612495422363, "optim/lr": 0.0016506000000000001, "optim/total_tokens": 721420288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 3.6553139686584473, "created_at": "2025-01-14T16:53:08.362410+00:00"} {"global_step": 2753, "acc_step": 0, "speed/wps": 12865.988374658804, "speed/FLOPS": 202078027144745.62, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.17634481191635132, "optim/lr": 0.0016512, "optim/total_tokens": 721682432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288153, "loss/out": 3.5428245067596436, "created_at": "2025-01-14T16:53:13.461825+00:00"} {"global_step": 2754, "acc_step": 0, "speed/wps": 12868.912266864789, "speed/FLOPS": 202123950889689.53, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15012507140636444, "optim/lr": 0.0016518, "optim/total_tokens": 721944576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469875, "loss/out": 3.588300943374634, "created_at": "2025-01-14T16:53:18.563258+00:00"} {"global_step": 2755, "acc_step": 0, "speed/wps": 12864.908605304845, "speed/FLOPS": 202061067883283.12, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.15044160187244415, "optim/lr": 0.0016523999999999998, "optim/total_tokens": 722206720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298409, "loss/out": 3.5372848510742188, "created_at": "2025-01-14T16:53:23.661944+00:00"} {"global_step": 2756, "acc_step": 0, "speed/wps": 12864.947443666544, "speed/FLOPS": 202061677893123.1, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1441723108291626, "optim/lr": 0.0016530000000000002, "optim/total_tokens": 722468864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463222, "loss/out": 3.6686158180236816, "created_at": "2025-01-14T16:53:28.764266+00:00"} {"global_step": 2757, "acc_step": 0, "speed/wps": 12865.46638270967, "speed/FLOPS": 202069828543892.38, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.15173405408859253, "optim/lr": 0.0016536, "optim/total_tokens": 722731008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297061, "loss/out": 3.6166348457336426, "created_at": "2025-01-14T16:53:33.862004+00:00"} {"global_step": 2758, "acc_step": 0, "speed/wps": 12867.587383678454, "speed/FLOPS": 202103141778667.88, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1613052636384964, "optim/lr": 0.0016542, "optim/total_tokens": 722993152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467205, "loss/out": 3.6784727573394775, "created_at": "2025-01-14T16:53:38.960773+00:00"} {"global_step": 2759, "acc_step": 0, "speed/wps": 12861.887705057165, "speed/FLOPS": 202013620493741.34, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1608331948518753, "optim/lr": 0.0016547999999999999, "optim/total_tokens": 723255296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295129, "loss/out": 3.753330945968628, "created_at": "2025-01-14T16:53:44.059722+00:00"} {"global_step": 2760, "acc_step": 0, "speed/wps": 12870.894503881043, "speed/FLOPS": 202155084645908.94, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16524362564086914, "optim/lr": 0.0016554, "optim/total_tokens": 723517440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 3.578714370727539, "created_at": "2025-01-14T16:53:49.156588+00:00"} {"global_step": 2761, "acc_step": 0, "speed/wps": 12861.497969922806, "speed/FLOPS": 202007499167904.06, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.19873788952827454, "optim/lr": 0.0016560000000000001, "optim/total_tokens": 723779584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462701, "loss/out": 3.628636360168457, "created_at": "2025-01-14T16:53:54.255646+00:00"} {"global_step": 2762, "acc_step": 0, "speed/wps": 12861.38577041489, "speed/FLOPS": 202005736920453.97, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.18227139115333557, "optim/lr": 0.0016566, "optim/total_tokens": 724041728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478372, "loss/out": 3.643582344055176, "created_at": "2025-01-14T16:53:59.357905+00:00"} {"global_step": 2763, "acc_step": 0, "speed/wps": 12863.28208757935, "speed/FLOPS": 202035521187334.62, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.1862746775150299, "optim/lr": 0.0016572, "optim/total_tokens": 724303872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 3.5340702533721924, "created_at": "2025-01-14T16:54:04.457108+00:00"} {"global_step": 2764, "acc_step": 0, "speed/wps": 12855.302603970384, "speed/FLOPS": 201910192432296.3, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2013085037469864, "optim/lr": 0.0016578, "optim/total_tokens": 724566016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 3.702343463897705, "created_at": "2025-01-14T16:54:09.562990+00:00"} {"global_step": 2765, "acc_step": 0, "speed/wps": 12866.765695635793, "speed/FLOPS": 202090236038840.2, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1865486055612564, "optim/lr": 0.0016584, "optim/total_tokens": 724828160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 3.711488962173462, "created_at": "2025-01-14T16:54:14.660094+00:00"} {"global_step": 2766, "acc_step": 0, "speed/wps": 12861.537800866334, "speed/FLOPS": 202008124767605.6, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1759730577468872, "optim/lr": 0.001659, "optim/total_tokens": 725090304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 3.5890824794769287, "created_at": "2025-01-14T16:54:19.761999+00:00"} {"global_step": 2767, "acc_step": 0, "speed/wps": 12862.849470760184, "speed/FLOPS": 202028726345711.84, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.40516045689582825, "optim/lr": 0.0016596, "optim/total_tokens": 725352448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 3.6344072818756104, "created_at": "2025-01-14T16:54:24.861462+00:00"} {"global_step": 2768, "acc_step": 0, "speed/wps": 12866.37566953354, "speed/FLOPS": 202084110142952.28, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1472737342119217, "optim/lr": 0.0016602000000000001, "optim/total_tokens": 725614592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 3.7060601711273193, "created_at": "2025-01-14T16:54:29.958733+00:00"} {"global_step": 2769, "acc_step": 0, "speed/wps": 12861.371532289213, "speed/FLOPS": 202005513291124.97, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.18067097663879395, "optim/lr": 0.0016608, "optim/total_tokens": 725876736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 3.61903715133667, "created_at": "2025-01-14T16:54:35.060016+00:00"} {"global_step": 2770, "acc_step": 0, "speed/wps": 12860.447682366947, "speed/FLOPS": 201991002958592.94, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1847086250782013, "optim/lr": 0.0016614, "optim/total_tokens": 726138880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 3.7352097034454346, "created_at": "2025-01-14T16:54:40.159720+00:00"} {"global_step": 2771, "acc_step": 0, "speed/wps": 12863.505346076738, "speed/FLOPS": 202039027768825.44, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.15703217685222626, "optim/lr": 0.0016620000000000003, "optim/total_tokens": 726401024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 3.7116360664367676, "created_at": "2025-01-14T16:54:45.257993+00:00"} {"global_step": 2772, "acc_step": 0, "speed/wps": 12858.090460521275, "speed/FLOPS": 201953979550344.4, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.16442137956619263, "optim/lr": 0.0016626000000000002, "optim/total_tokens": 726663168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.608956813812256, "created_at": "2025-01-14T16:54:50.358387+00:00"} {"global_step": 2773, "acc_step": 0, "speed/wps": 12865.513043150395, "speed/FLOPS": 202070561410232.12, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.20074410736560822, "optim/lr": 0.0016632, "optim/total_tokens": 726925312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.67110538482666, "created_at": "2025-01-14T16:54:55.458567+00:00"} {"global_step": 2774, "acc_step": 0, "speed/wps": 12866.354235792565, "speed/FLOPS": 202083773496597.22, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.35168135166168213, "optim/lr": 0.0016638, "optim/total_tokens": 727187456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424894, "loss/out": 3.5250706672668457, "created_at": "2025-01-14T16:55:00.558906+00:00"} {"global_step": 2775, "acc_step": 0, "speed/wps": 12863.046186830545, "speed/FLOPS": 202031816041913.06, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1681298315525055, "optim/lr": 0.0016644, "optim/total_tokens": 727449600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448597, "loss/out": 3.5785632133483887, "created_at": "2025-01-14T16:55:05.665592+00:00"} {"global_step": 2776, "acc_step": 0, "speed/wps": 12861.439380520087, "speed/FLOPS": 202006578940824.06, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.21363957226276398, "optim/lr": 0.0016650000000000002, "optim/total_tokens": 727711744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355618, "loss/out": 3.6767873764038086, "created_at": "2025-01-14T16:55:10.764980+00:00"} {"global_step": 2777, "acc_step": 0, "speed/wps": 12867.086700438405, "speed/FLOPS": 202095277860372.03, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1817246675491333, "optim/lr": 0.0016656000000000002, "optim/total_tokens": 727973888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466155, "loss/out": 3.5495445728302, "created_at": "2025-01-14T16:55:15.865755+00:00"} {"global_step": 2778, "acc_step": 0, "speed/wps": 12857.309965836672, "speed/FLOPS": 201941720808810.62, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0054371356964111, "optim/lr": 0.0016662, "optim/total_tokens": 728236032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 3.6461966037750244, "created_at": "2025-01-14T16:55:20.977947+00:00"} {"global_step": 2779, "acc_step": 0, "speed/wps": 12865.529743798714, "speed/FLOPS": 202070823716863.06, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24095161259174347, "optim/lr": 0.0016668, "optim/total_tokens": 728498176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 3.526562213897705, "created_at": "2025-01-14T16:55:26.077946+00:00"} {"global_step": 2780, "acc_step": 0, "speed/wps": 12858.353653157408, "speed/FLOPS": 201958113352359.47, "speed/curr_iter_time": 1.2857, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.27017679810523987, "optim/lr": 0.0016673999999999999, "optim/total_tokens": 728760320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 3.6814517974853516, "created_at": "2025-01-14T16:55:31.180952+00:00"} {"global_step": 2781, "acc_step": 0, "speed/wps": 12863.048208947996, "speed/FLOPS": 202031847802046.3, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.001, "optim/grad_norm": 0.20345133543014526, "optim/lr": 0.0016680000000000002, "optim/total_tokens": 729022464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 3.7318758964538574, "created_at": "2025-01-14T16:55:36.280713+00:00"} {"global_step": 2782, "acc_step": 0, "speed/wps": 12858.986850169213, "speed/FLOPS": 201968058581533.72, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.21055345237255096, "optim/lr": 0.0016686000000000001, "optim/total_tokens": 729284608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 3.805708169937134, "created_at": "2025-01-14T16:55:41.386194+00:00"} {"global_step": 2783, "acc_step": 0, "speed/wps": 12872.673423369682, "speed/FLOPS": 202183025020969.53, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4571433365345001, "optim/lr": 0.0016692, "optim/total_tokens": 729546752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 3.6806838512420654, "created_at": "2025-01-14T16:55:46.481223+00:00"} {"global_step": 2784, "acc_step": 0, "speed/wps": 12857.454116858537, "speed/FLOPS": 201943984898693.28, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.574499249458313, "optim/lr": 0.0016698, "optim/total_tokens": 729808896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 3.6048173904418945, "created_at": "2025-01-14T16:55:51.582844+00:00"} {"global_step": 2785, "acc_step": 0, "speed/wps": 12866.411736063163, "speed/FLOPS": 202084676617360.62, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.24755921959877014, "optim/lr": 0.0016703999999999998, "optim/total_tokens": 730071040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 3.7752790451049805, "created_at": "2025-01-14T16:55:56.684260+00:00"} {"global_step": 2786, "acc_step": 0, "speed/wps": 12862.821827442878, "speed/FLOPS": 202028292169427.66, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2517790198326111, "optim/lr": 0.0016710000000000002, "optim/total_tokens": 730333184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 3.7866692543029785, "created_at": "2025-01-14T16:56:01.785811+00:00"} {"global_step": 2787, "acc_step": 0, "speed/wps": 12862.601677361228, "speed/FLOPS": 202024834409877.44, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0110750198364258, "optim/lr": 0.0016716, "optim/total_tokens": 730595328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473310, "loss/out": 3.663609027862549, "created_at": "2025-01-14T16:56:06.885268+00:00"} {"global_step": 2788, "acc_step": 0, "speed/wps": 12863.553687714078, "speed/FLOPS": 202039787040668.25, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.28863608837127686, "optim/lr": 0.0016722, "optim/total_tokens": 730857472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322311, "loss/out": 3.7298316955566406, "created_at": "2025-01-14T16:56:11.989342+00:00"} {"global_step": 2789, "acc_step": 0, "speed/wps": 12860.61142387215, "speed/FLOPS": 201993574743936.56, "speed/curr_iter_time": 1.2846, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5445209741592407, "optim/lr": 0.0016728, "optim/total_tokens": 731119616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 3.6503849029541016, "created_at": "2025-01-14T16:56:17.091160+00:00"} {"global_step": 2790, "acc_step": 0, "speed/wps": 12861.703069947376, "speed/FLOPS": 202010720545626.9, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.22366812825202942, "optim/lr": 0.0016734, "optim/total_tokens": 731381760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311191, "loss/out": 3.736788749694824, "created_at": "2025-01-14T16:56:22.190202+00:00"} {"global_step": 2791, "acc_step": 0, "speed/wps": 12869.241503284029, "speed/FLOPS": 202129122000071.97, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2166236788034439, "optim/lr": 0.0016740000000000001, "optim/total_tokens": 731643904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 3.6729841232299805, "created_at": "2025-01-14T16:56:27.287085+00:00"} {"global_step": 2792, "acc_step": 0, "speed/wps": 12866.995662752744, "speed/FLOPS": 202093847988419.84, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.22170433402061462, "optim/lr": 0.0016746, "optim/total_tokens": 731906048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450225, "loss/out": 3.5264947414398193, "created_at": "2025-01-14T16:56:32.386083+00:00"} {"global_step": 2793, "acc_step": 0, "speed/wps": 12860.671220904309, "speed/FLOPS": 201994513938496.34, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.18194161355495453, "optim/lr": 0.0016752, "optim/total_tokens": 732168192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 3.723752975463867, "created_at": "2025-01-14T16:56:37.487029+00:00"} {"global_step": 2794, "acc_step": 0, "speed/wps": 12858.985990136916, "speed/FLOPS": 201968045073544.7, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.20686578750610352, "optim/lr": 0.0016758, "optim/total_tokens": 732430336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447026, "loss/out": 3.696226119995117, "created_at": "2025-01-14T16:56:42.589145+00:00"} {"global_step": 2795, "acc_step": 0, "speed/wps": 12861.647198536653, "speed/FLOPS": 202009843008346.12, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.18723343312740326, "optim/lr": 0.0016764, "optim/total_tokens": 732692480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 3.626617908477783, "created_at": "2025-01-14T16:56:47.691451+00:00"} {"global_step": 2796, "acc_step": 0, "speed/wps": 12867.05493568273, "speed/FLOPS": 202094778951234.8, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2084917575120926, "optim/lr": 0.001677, "optim/total_tokens": 732954624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328073, "loss/out": 3.566226005554199, "created_at": "2025-01-14T16:56:52.788964+00:00"} {"global_step": 2797, "acc_step": 0, "speed/wps": 12858.977645949766, "speed/FLOPS": 201967914016619.4, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.1759902685880661, "optim/lr": 0.0016776000000000002, "optim/total_tokens": 733216768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482522, "loss/out": 3.724661350250244, "created_at": "2025-01-14T16:56:57.889219+00:00"} {"global_step": 2798, "acc_step": 0, "speed/wps": 12862.56996060824, "speed/FLOPS": 202024336254688.34, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.16657376289367676, "optim/lr": 0.0016782000000000001, "optim/total_tokens": 733478912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 3.687180280685425, "created_at": "2025-01-14T16:57:02.989750+00:00"} {"global_step": 2799, "acc_step": 0, "speed/wps": 12860.26065840746, "speed/FLOPS": 201988065490311.34, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.1629074215888977, "optim/lr": 0.0016788, "optim/total_tokens": 733741056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 3.6378772258758545, "created_at": "2025-01-14T16:57:08.096684+00:00"} {"global_step": 2800, "acc_step": 0, "speed/wps": 12868.410920441844, "speed/FLOPS": 202116076555195.0, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.18132399022579193, "optim/lr": 0.0016794, "optim/total_tokens": 734003200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462938, "loss/out": 3.5914130210876465, "created_at": "2025-01-14T16:57:13.193522+00:00"} {"global_step": 2801, "acc_step": 0, "speed/wps": 12855.491806452477, "speed/FLOPS": 201913164117269.03, "speed/curr_iter_time": 1.2864, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.18626992404460907, "optim/lr": 0.0016800000000000003, "optim/total_tokens": 734265344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 3.561976432800293, "created_at": "2025-01-14T16:57:18.294986+00:00"} {"global_step": 2802, "acc_step": 0, "speed/wps": 12863.60104443541, "speed/FLOPS": 202040530843051.72, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.004867434501648, "optim/lr": 0.0016806000000000002, "optim/total_tokens": 734527488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 3.6928935050964355, "created_at": "2025-01-14T16:57:23.393292+00:00"} {"global_step": 2803, "acc_step": 0, "speed/wps": 12864.893145759397, "speed/FLOPS": 202060825069878.22, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.78009295463562, "optim/lr": 0.0016812, "optim/total_tokens": 734789632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401352, "loss/out": 3.731065511703491, "created_at": "2025-01-14T16:57:28.491389+00:00"} {"global_step": 2804, "acc_step": 0, "speed/wps": 12863.901448756918, "speed/FLOPS": 202045249105719.8, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 11.02057933807373, "optim/lr": 0.0016818, "optim/total_tokens": 735051776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467205, "loss/out": 4.152737617492676, "created_at": "2025-01-14T16:57:33.592178+00:00"} {"global_step": 2805, "acc_step": 0, "speed/wps": 12862.940936495002, "speed/FLOPS": 202030162940764.25, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 59.382720947265625, "optim/lr": 0.0016824, "optim/total_tokens": 735313920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 414236, "loss/out": 4.693148136138916, "created_at": "2025-01-14T16:57:38.692268+00:00"} {"global_step": 2806, "acc_step": 0, "speed/wps": 12867.828133586896, "speed/FLOPS": 202106923086805.0, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.221628665924072, "optim/lr": 0.0016830000000000003, "optim/total_tokens": 735576064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471738, "loss/out": 4.389403343200684, "created_at": "2025-01-14T16:57:43.793568+00:00"} {"global_step": 2807, "acc_step": 0, "speed/wps": 12853.61455835842, "speed/FLOPS": 201883679356342.66, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.660479545593262, "optim/lr": 0.0016836000000000002, "optim/total_tokens": 735838208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482522, "loss/out": 4.4399871826171875, "created_at": "2025-01-14T16:57:48.900650+00:00"} {"global_step": 2808, "acc_step": 0, "speed/wps": 12866.06588822612, "speed/FLOPS": 202079244601835.12, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 16.25612449645996, "optim/lr": 0.0016842, "optim/total_tokens": 736100352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 4.286606311798096, "created_at": "2025-01-14T16:57:53.998188+00:00"} {"global_step": 2809, "acc_step": 0, "speed/wps": 12868.836973569532, "speed/FLOPS": 202122768305023.72, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5140990018844604, "optim/lr": 0.0016848, "optim/total_tokens": 736362496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444881, "loss/out": 4.423412322998047, "created_at": "2025-01-14T16:57:59.097787+00:00"} {"global_step": 2810, "acc_step": 0, "speed/wps": 12858.044408380207, "speed/FLOPS": 201953256238186.66, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 82.37698364257812, "optim/lr": 0.0016853999999999999, "optim/total_tokens": 736624640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 4.262578964233398, "created_at": "2025-01-14T16:58:04.198621+00:00"} {"global_step": 2811, "acc_step": 0, "speed/wps": 12861.927613643165, "speed/FLOPS": 202014247312924.62, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 798.4000244140625, "optim/lr": 0.0016860000000000002, "optim/total_tokens": 736886784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 5.183161735534668, "created_at": "2025-01-14T16:58:09.297908+00:00"} {"global_step": 2812, "acc_step": 0, "speed/wps": 12888.732011031476, "speed/FLOPS": 202435247207009.9, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 106.23219299316406, "optim/lr": 0.0016866000000000001, "optim/total_tokens": 737148928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453911, "loss/out": 9.185815811157227, "created_at": "2025-01-14T16:58:14.387099+00:00"} {"global_step": 2813, "acc_step": 0, "speed/wps": 12920.662320090656, "speed/FLOPS": 202936756587626.94, "speed/curr_iter_time": 1.2774, "speed/data_load_time": 0.0007, "optim/grad_norm": 63.20344161987305, "optim/lr": 0.0016872, "optim/total_tokens": 737411072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 8.713871002197266, "created_at": "2025-01-14T16:58:19.462885+00:00"} {"global_step": 2814, "acc_step": 0, "speed/wps": 12939.998144445073, "speed/FLOPS": 203240452279320.06, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0007, "optim/grad_norm": 36.953857421875, "optim/lr": 0.0016878, "optim/total_tokens": 737673216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316527, "loss/out": 9.004961013793945, "created_at": "2025-01-14T16:58:24.533230+00:00"} {"global_step": 2815, "acc_step": 0, "speed/wps": 12942.026714248525, "speed/FLOPS": 203272313755630.3, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 45.087642669677734, "optim/lr": 0.0016883999999999998, "optim/total_tokens": 737935360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302697, "loss/out": 9.35610294342041, "created_at": "2025-01-14T16:58:29.603835+00:00"} {"global_step": 2816, "acc_step": 0, "speed/wps": 12939.420038309641, "speed/FLOPS": 203231372328062.03, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0008, "optim/grad_norm": 64.82530212402344, "optim/lr": 0.001689, "optim/total_tokens": 738197504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 9.060983657836914, "created_at": "2025-01-14T16:58:34.674629+00:00"} {"global_step": 2817, "acc_step": 0, "speed/wps": 12893.92904781998, "speed/FLOPS": 202516873811250.28, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0009, "optim/grad_norm": 1024.5465087890625, "optim/lr": 0.0016896, "optim/total_tokens": 738459648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470119, "loss/out": 8.303787231445312, "created_at": "2025-01-14T16:58:39.760969+00:00"} {"global_step": 2818, "acc_step": 0, "speed/wps": 12852.154773350556, "speed/FLOPS": 201860751426839.34, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0008, "optim/grad_norm": 5598.5556640625, "optim/lr": 0.0016902, "optim/total_tokens": 738721792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458734, "loss/out": 7.025389671325684, "created_at": "2025-01-14T16:58:44.867073+00:00"} {"global_step": 2819, "acc_step": 0, "speed/wps": 12864.559832601934, "speed/FLOPS": 202055589928719.97, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0007, "optim/grad_norm": 280.0191345214844, "optim/lr": 0.0016908, "optim/total_tokens": 738983936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 7.161530017852783, "created_at": "2025-01-14T16:58:49.966149+00:00"} {"global_step": 2820, "acc_step": 0, "speed/wps": 12854.165216668882, "speed/FLOPS": 201892328201792.28, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0006, "optim/grad_norm": 1170.9989013671875, "optim/lr": 0.0016914, "optim/total_tokens": 739246080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487669, "loss/out": 7.654204845428467, "created_at": "2025-01-14T16:58:55.070740+00:00"} {"global_step": 2821, "acc_step": 0, "speed/wps": 12861.80159306482, "speed/FLOPS": 202012267986571.28, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 775.5830688476562, "optim/lr": 0.001692, "optim/total_tokens": 739508224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 8.231619834899902, "created_at": "2025-01-14T16:59:00.169996+00:00"} {"global_step": 2822, "acc_step": 0, "speed/wps": 12866.049894257374, "speed/FLOPS": 202078993394577.97, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 321.3277282714844, "optim/lr": 0.0016926, "optim/total_tokens": 739770368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490054, "loss/out": 8.886079788208008, "created_at": "2025-01-14T16:59:05.267316+00:00"} {"global_step": 2823, "acc_step": 0, "speed/wps": 12907.561953608229, "speed/FLOPS": 202730997330230.4, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 91.68246459960938, "optim/lr": 0.0016932000000000002, "optim/total_tokens": 740032512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 9.695940017700195, "created_at": "2025-01-14T16:59:10.348898+00:00"} {"global_step": 2824, "acc_step": 0, "speed/wps": 12910.197828934395, "speed/FLOPS": 202772397374300.12, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0009, "optim/grad_norm": 71.53424072265625, "optim/lr": 0.0016938, "optim/total_tokens": 740294656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473839, "loss/out": 9.573965072631836, "created_at": "2025-01-14T16:59:15.431330+00:00"} {"global_step": 2825, "acc_step": 0, "speed/wps": 12923.978089503162, "speed/FLOPS": 202988835302594.97, "speed/curr_iter_time": 1.279, "speed/data_load_time": 0.0007, "optim/grad_norm": 98.64205169677734, "optim/lr": 0.0016944, "optim/total_tokens": 740556800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 388627, "loss/out": 9.16939926147461, "created_at": "2025-01-14T16:59:20.505770+00:00"} {"global_step": 2826, "acc_step": 0, "speed/wps": 12934.352382611758, "speed/FLOPS": 203151777831638.12, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.94565200805664, "optim/lr": 0.001695, "optim/total_tokens": 740818944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 8.80105209350586, "created_at": "2025-01-14T16:59:25.578349+00:00"} {"global_step": 2827, "acc_step": 0, "speed/wps": 12941.208705571362, "speed/FLOPS": 203259465805293.75, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.0945563316345215, "optim/lr": 0.0016956000000000002, "optim/total_tokens": 741081088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 8.290185928344727, "created_at": "2025-01-14T16:59:30.647226+00:00"} {"global_step": 2828, "acc_step": 0, "speed/wps": 12917.640996284848, "speed/FLOPS": 202889302545522.75, "speed/curr_iter_time": 1.2785, "speed/data_load_time": 0.0007, "optim/grad_norm": 44.92050552368164, "optim/lr": 0.0016962000000000001, "optim/total_tokens": 741343232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 8.087660789489746, "created_at": "2025-01-14T16:59:35.728047+00:00"} {"global_step": 2829, "acc_step": 0, "speed/wps": 12897.819765355065, "speed/FLOPS": 202577982876544.94, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0009, "optim/grad_norm": 62.7982063293457, "optim/lr": 0.0016968, "optim/total_tokens": 741605376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305035, "loss/out": 8.082213401794434, "created_at": "2025-01-14T16:59:40.813088+00:00"} {"global_step": 2830, "acc_step": 0, "speed/wps": 12884.66672317186, "speed/FLOPS": 202371396274883.72, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 99.39263916015625, "optim/lr": 0.0016974, "optim/total_tokens": 741867520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 8.129680633544922, "created_at": "2025-01-14T16:59:45.909834+00:00"} {"global_step": 2831, "acc_step": 0, "speed/wps": 12864.315305748785, "speed/FLOPS": 202051749298476.2, "speed/curr_iter_time": 1.2821, "speed/data_load_time": 0.0007, "optim/grad_norm": 91.49803161621094, "optim/lr": 0.0016979999999999999, "optim/total_tokens": 742129664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321540, "loss/out": 8.239484786987305, "created_at": "2025-01-14T16:59:51.011421+00:00"} {"global_step": 2832, "acc_step": 0, "speed/wps": 12861.317466720215, "speed/FLOPS": 202004664117069.97, "speed/curr_iter_time": 1.284, "speed/data_load_time": 0.0007, "optim/grad_norm": 216.65292358398438, "optim/lr": 0.0016986000000000002, "optim/total_tokens": 742391808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 8.274347305297852, "created_at": "2025-01-14T16:59:56.110953+00:00"} {"global_step": 2833, "acc_step": 0, "speed/wps": 12851.985942310095, "speed/FLOPS": 201858099703350.75, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 11.679139137268066, "optim/lr": 0.0016992, "optim/total_tokens": 742653952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337166, "loss/out": 8.364116668701172, "created_at": "2025-01-14T17:00:01.213792+00:00"} {"global_step": 2834, "acc_step": 0, "speed/wps": 12853.019396006475, "speed/FLOPS": 201874331513766.6, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0009, "optim/grad_norm": 16.864103317260742, "optim/lr": 0.0016998, "optim/total_tokens": 742916096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441687, "loss/out": 8.284510612487793, "created_at": "2025-01-14T17:00:06.318681+00:00"} {"global_step": 2835, "acc_step": 0, "speed/wps": 12851.865493934823, "speed/FLOPS": 201856207896103.7, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 247.7421112060547, "optim/lr": 0.0017004, "optim/total_tokens": 743178240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369775, "loss/out": 8.192779541015625, "created_at": "2025-01-14T17:00:11.426679+00:00"} {"global_step": 2836, "acc_step": 0, "speed/wps": 12847.864588731121, "speed/FLOPS": 201793368182059.7, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 19.021339416503906, "optim/lr": 0.0017009999999999998, "optim/total_tokens": 743440384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 8.143774032592773, "created_at": "2025-01-14T17:00:16.537201+00:00"} {"global_step": 2837, "acc_step": 0, "speed/wps": 12844.607376262837, "speed/FLOPS": 201742209184366.28, "speed/curr_iter_time": 1.2863, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.045158386230469, "optim/lr": 0.0017016000000000002, "optim/total_tokens": 743702528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 8.140677452087402, "created_at": "2025-01-14T17:00:21.648684+00:00"} {"global_step": 2838, "acc_step": 0, "speed/wps": 12853.05440403089, "speed/FLOPS": 201874881362896.1, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.355013132095337, "optim/lr": 0.0017022, "optim/total_tokens": 743964672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447839, "loss/out": 7.914936542510986, "created_at": "2025-01-14T17:00:26.751575+00:00"} {"global_step": 2839, "acc_step": 0, "speed/wps": 12855.239519709201, "speed/FLOPS": 201909201607287.56, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 32.05109405517578, "optim/lr": 0.0017028, "optim/total_tokens": 744226816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 8.257819175720215, "created_at": "2025-01-14T17:00:31.853642+00:00"} {"global_step": 2840, "acc_step": 0, "speed/wps": 12857.905603146872, "speed/FLOPS": 201951076111258.75, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7251267433166504, "optim/lr": 0.0017033999999999999, "optim/total_tokens": 744488960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 8.47528076171875, "created_at": "2025-01-14T17:00:36.954338+00:00"} {"global_step": 2841, "acc_step": 0, "speed/wps": 12864.85772512443, "speed/FLOPS": 202060268740133.0, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.7416436672210693, "optim/lr": 0.0017039999999999998, "optim/total_tokens": 744751104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 391544, "loss/out": 8.255779266357422, "created_at": "2025-01-14T17:00:42.061849+00:00"} {"global_step": 2842, "acc_step": 0, "speed/wps": 12864.156255632402, "speed/FLOPS": 202049251197840.62, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.001, "optim/grad_norm": 2.6384212970733643, "optim/lr": 0.0017046000000000001, "optim/total_tokens": 745013248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 8.15721321105957, "created_at": "2025-01-14T17:00:47.161903+00:00"} {"global_step": 2843, "acc_step": 0, "speed/wps": 12865.696453928582, "speed/FLOPS": 202073442128536.38, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3287904262542725, "optim/lr": 0.0017052, "optim/total_tokens": 745275392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 8.234865188598633, "created_at": "2025-01-14T17:00:52.260076+00:00"} {"global_step": 2844, "acc_step": 0, "speed/wps": 12874.406008071344, "speed/FLOPS": 202210237644530.03, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.31373131275177, "optim/lr": 0.0017058, "optim/total_tokens": 745537536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464820, "loss/out": 7.959339618682861, "created_at": "2025-01-14T17:00:57.354785+00:00"} {"global_step": 2845, "acc_step": 0, "speed/wps": 12877.324598732233, "speed/FLOPS": 202256078121423.12, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.893930196762085, "optim/lr": 0.0017063999999999998, "optim/total_tokens": 745799680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364255, "loss/out": 7.9161882400512695, "created_at": "2025-01-14T17:01:02.447626+00:00"} {"global_step": 2846, "acc_step": 0, "speed/wps": 12873.270510946762, "speed/FLOPS": 202192403101851.9, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7111443281173706, "optim/lr": 0.001707, "optim/total_tokens": 746061824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448361, "loss/out": 7.74253511428833, "created_at": "2025-01-14T17:01:07.546768+00:00"} {"global_step": 2847, "acc_step": 0, "speed/wps": 12883.642869701083, "speed/FLOPS": 202355315249203.06, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4865812063217163, "optim/lr": 0.0017076, "optim/total_tokens": 746323968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322689, "loss/out": 7.603037357330322, "created_at": "2025-01-14T17:01:12.637061+00:00"} {"global_step": 2848, "acc_step": 0, "speed/wps": 12879.815169926935, "speed/FLOPS": 202295195964439.38, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1530871391296387, "optim/lr": 0.0017082, "optim/total_tokens": 746586112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480426, "loss/out": 7.573335647583008, "created_at": "2025-01-14T17:01:17.732004+00:00"} {"global_step": 2849, "acc_step": 0, "speed/wps": 12876.536593031233, "speed/FLOPS": 202243701409055.6, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6685593128204346, "optim/lr": 0.0017088000000000001, "optim/total_tokens": 746848256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459765, "loss/out": 7.453154563903809, "created_at": "2025-01-14T17:01:22.825309+00:00"} {"global_step": 2850, "acc_step": 0, "speed/wps": 12865.665476985852, "speed/FLOPS": 202072955593084.03, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2941877841949463, "optim/lr": 0.0017094, "optim/total_tokens": 747110400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 7.461277961730957, "created_at": "2025-01-14T17:01:27.923488+00:00"} {"global_step": 2851, "acc_step": 0, "speed/wps": 12883.109011990067, "speed/FLOPS": 202346930280254.2, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0743159055709839, "optim/lr": 0.00171, "optim/total_tokens": 747372544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 7.43869686126709, "created_at": "2025-01-14T17:01:33.014149+00:00"} {"global_step": 2852, "acc_step": 0, "speed/wps": 12879.867333070486, "speed/FLOPS": 202296015258288.53, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8814327120780945, "optim/lr": 0.0017106, "optim/total_tokens": 747634688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481758, "loss/out": 7.425464153289795, "created_at": "2025-01-14T17:01:38.106238+00:00"} {"global_step": 2853, "acc_step": 0, "speed/wps": 12869.53839581749, "speed/FLOPS": 202133785105283.06, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2406654357910156, "optim/lr": 0.0017112000000000002, "optim/total_tokens": 747896832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 7.530595779418945, "created_at": "2025-01-14T17:01:43.204401+00:00"} {"global_step": 2854, "acc_step": 0, "speed/wps": 12872.004604488031, "speed/FLOPS": 202172520301379.75, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7700560092926025, "optim/lr": 0.0017118, "optim/total_tokens": 748158976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318405, "loss/out": 7.423970699310303, "created_at": "2025-01-14T17:01:48.301456+00:00"} {"global_step": 2855, "acc_step": 0, "speed/wps": 12868.417502515367, "speed/FLOPS": 202116179935704.4, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6662287712097168, "optim/lr": 0.0017124, "optim/total_tokens": 748421120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 7.352672576904297, "created_at": "2025-01-14T17:01:53.403664+00:00"} {"global_step": 2856, "acc_step": 0, "speed/wps": 12867.508837663905, "speed/FLOPS": 202101908105575.84, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0402096509933472, "optim/lr": 0.001713, "optim/total_tokens": 748683264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330091, "loss/out": 7.33909797668457, "created_at": "2025-01-14T17:01:58.501174+00:00"} {"global_step": 2857, "acc_step": 0, "speed/wps": 12861.179625082832, "speed/FLOPS": 202002499124740.6, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1578458547592163, "optim/lr": 0.0017136000000000002, "optim/total_tokens": 748945408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 7.406146049499512, "created_at": "2025-01-14T17:02:03.603095+00:00"} {"global_step": 2858, "acc_step": 0, "speed/wps": 12868.153415337047, "speed/FLOPS": 202112032083675.1, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9972140789031982, "optim/lr": 0.0017142000000000001, "optim/total_tokens": 749207552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333311, "loss/out": 7.314810752868652, "created_at": "2025-01-14T17:02:08.700804+00:00"} {"global_step": 2859, "acc_step": 0, "speed/wps": 12871.94789479328, "speed/FLOPS": 202171629597696.4, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.029408574104309, "optim/lr": 0.0017148, "optim/total_tokens": 749469696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 7.269048690795898, "created_at": "2025-01-14T17:02:13.797262+00:00"} {"global_step": 2860, "acc_step": 0, "speed/wps": 12871.566112752058, "speed/FLOPS": 202165633186115.5, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.991295337677002, "optim/lr": 0.0017154, "optim/total_tokens": 749731840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469353, "loss/out": 7.2549591064453125, "created_at": "2025-01-14T17:02:18.916685+00:00"} {"global_step": 2861, "acc_step": 0, "speed/wps": 12857.411884439543, "speed/FLOPS": 201943321580519.56, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6287093758583069, "optim/lr": 0.0017159999999999999, "optim/total_tokens": 749993984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444878, "loss/out": 7.2646918296813965, "created_at": "2025-01-14T17:02:24.017774+00:00"} {"global_step": 2862, "acc_step": 0, "speed/wps": 12864.888286582393, "speed/FLOPS": 202060748749826.16, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3802856206893921, "optim/lr": 0.0017166000000000002, "optim/total_tokens": 750256128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355742, "loss/out": 7.120932579040527, "created_at": "2025-01-14T17:02:29.116085+00:00"} {"global_step": 2863, "acc_step": 0, "speed/wps": 12854.67300325805, "speed/FLOPS": 201900303687947.06, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5956692695617676, "optim/lr": 0.0017172, "optim/total_tokens": 750518272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446220, "loss/out": 7.044703483581543, "created_at": "2025-01-14T17:02:34.219929+00:00"} {"global_step": 2864, "acc_step": 0, "speed/wps": 12860.691373529637, "speed/FLOPS": 201994830463169.72, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.001, "optim/grad_norm": 0.5361446738243103, "optim/lr": 0.0017178, "optim/total_tokens": 750780416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355086, "loss/out": 6.975766181945801, "created_at": "2025-01-14T17:02:39.322317+00:00"} {"global_step": 2865, "acc_step": 0, "speed/wps": 12866.828514290506, "speed/FLOPS": 202091222692134.34, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7161484956741333, "optim/lr": 0.0017184, "optim/total_tokens": 751042560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447263, "loss/out": 7.16135311126709, "created_at": "2025-01-14T17:02:44.419498+00:00"} {"global_step": 2866, "acc_step": 0, "speed/wps": 12867.421057474356, "speed/FLOPS": 202100529397076.34, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0004298686981201, "optim/lr": 0.0017189999999999998, "optim/total_tokens": 751304704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317879, "loss/out": 7.059093475341797, "created_at": "2025-01-14T17:02:49.520790+00:00"} {"global_step": 2867, "acc_step": 0, "speed/wps": 12865.514837879959, "speed/FLOPS": 202070589598926.12, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5551013350486755, "optim/lr": 0.0017196000000000002, "optim/total_tokens": 751566848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 7.105678558349609, "created_at": "2025-01-14T17:02:54.622089+00:00"} {"global_step": 2868, "acc_step": 0, "speed/wps": 12864.894589871761, "speed/FLOPS": 202060847751647.3, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5248175859451294, "optim/lr": 0.0017202, "optim/total_tokens": 751828992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 6.888180732727051, "created_at": "2025-01-14T17:02:59.721717+00:00"} {"global_step": 2869, "acc_step": 0, "speed/wps": 12859.973037483202, "speed/FLOPS": 201983548008463.3, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.44653379917144775, "optim/lr": 0.0017208, "optim/total_tokens": 752091136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 6.8773322105407715, "created_at": "2025-01-14T17:03:04.826272+00:00"} {"global_step": 2870, "acc_step": 0, "speed/wps": 12859.496634582534, "speed/FLOPS": 201976065446262.12, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6277133226394653, "optim/lr": 0.0017213999999999999, "optim/total_tokens": 752353280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 6.865468502044678, "created_at": "2025-01-14T17:03:09.927145+00:00"} {"global_step": 2871, "acc_step": 0, "speed/wps": 12857.693223783732, "speed/FLOPS": 201947740401520.8, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5253242254257202, "optim/lr": 0.001722, "optim/total_tokens": 752615424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 6.778594017028809, "created_at": "2025-01-14T17:03:15.028626+00:00"} {"global_step": 2872, "acc_step": 0, "speed/wps": 12859.939761432794, "speed/FLOPS": 201983025362365.56, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.37544167041778564, "optim/lr": 0.0017226000000000001, "optim/total_tokens": 752877568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 6.73991584777832, "created_at": "2025-01-14T17:03:20.129412+00:00"} {"global_step": 2873, "acc_step": 0, "speed/wps": 12848.293746670925, "speed/FLOPS": 201800108697229.1, "speed/curr_iter_time": 1.286, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5092166066169739, "optim/lr": 0.0017232, "optim/total_tokens": 753139712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 6.7465996742248535, "created_at": "2025-01-14T17:03:25.233955+00:00"} {"global_step": 2874, "acc_step": 0, "speed/wps": 12855.363718466297, "speed/FLOPS": 201911152319434.75, "speed/curr_iter_time": 1.2839, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.5138044357299805, "optim/lr": 0.0017238, "optim/total_tokens": 753401856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458410, "loss/out": 6.707355499267578, "created_at": "2025-01-14T17:03:30.337180+00:00"} {"global_step": 2875, "acc_step": 0, "speed/wps": 12854.378841414056, "speed/FLOPS": 201895683471966.4, "speed/curr_iter_time": 1.2856, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.4622947871685028, "optim/lr": 0.0017244, "optim/total_tokens": 753664000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 6.6747894287109375, "created_at": "2025-01-14T17:03:35.441551+00:00"} {"global_step": 2876, "acc_step": 0, "speed/wps": 12841.154053570617, "speed/FLOPS": 201687970006122.34, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.4221101999282837, "optim/lr": 0.001725, "optim/total_tokens": 753926144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308786, "loss/out": 6.636884689331055, "created_at": "2025-01-14T17:03:40.548700+00:00"} {"global_step": 2877, "acc_step": 0, "speed/wps": 12861.117063725149, "speed/FLOPS": 202001516512649.88, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4242396056652069, "optim/lr": 0.0017256, "optim/total_tokens": 754188288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 6.556241035461426, "created_at": "2025-01-14T17:03:45.648937+00:00"} {"global_step": 2878, "acc_step": 0, "speed/wps": 12857.428061323777, "speed/FLOPS": 201943575660715.8, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6032335758209229, "optim/lr": 0.0017262, "optim/total_tokens": 754450432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303928, "loss/out": 6.469627380371094, "created_at": "2025-01-14T17:03:50.750678+00:00"} {"global_step": 2879, "acc_step": 0, "speed/wps": 12855.131315978822, "speed/FLOPS": 201907502119014.47, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6584526896476746, "optim/lr": 0.0017268000000000001, "optim/total_tokens": 754712576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 6.542499542236328, "created_at": "2025-01-14T17:03:55.852374+00:00"} {"global_step": 2880, "acc_step": 0, "speed/wps": 12854.342236498429, "speed/FLOPS": 201895108541466.06, "speed/curr_iter_time": 1.2853, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5713533759117126, "optim/lr": 0.0017274, "optim/total_tokens": 754974720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464007, "loss/out": 6.6698408126831055, "created_at": "2025-01-14T17:04:00.957067+00:00"} {"global_step": 2881, "acc_step": 0, "speed/wps": 12858.20525515968, "speed/FLOPS": 201955782557884.94, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.44455841183662415, "optim/lr": 0.001728, "optim/total_tokens": 755236864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 6.41524600982666, "created_at": "2025-01-14T17:04:06.066026+00:00"} {"global_step": 2882, "acc_step": 0, "speed/wps": 12854.264990925014, "speed/FLOPS": 201893895293589.03, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5387513637542725, "optim/lr": 0.0017286000000000003, "optim/total_tokens": 755499008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 6.406182289123535, "created_at": "2025-01-14T17:04:11.172944+00:00"} {"global_step": 2883, "acc_step": 0, "speed/wps": 12855.234339039811, "speed/FLOPS": 201909120237755.8, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5554025769233704, "optim/lr": 0.0017292000000000002, "optim/total_tokens": 755761152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462701, "loss/out": 6.4510817527771, "created_at": "2025-01-14T17:04:16.279086+00:00"} {"global_step": 2884, "acc_step": 0, "speed/wps": 12855.848606247742, "speed/FLOPS": 201918768148347.97, "speed/curr_iter_time": 1.2855, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.34550631046295166, "optim/lr": 0.0017298, "optim/total_tokens": 756023296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306907, "loss/out": 6.314034938812256, "created_at": "2025-01-14T17:04:21.381001+00:00"} {"global_step": 2885, "acc_step": 0, "speed/wps": 12852.007203801144, "speed/FLOPS": 201858433644284.03, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.36691638827323914, "optim/lr": 0.0017304, "optim/total_tokens": 756285440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 6.241008758544922, "created_at": "2025-01-14T17:04:26.487732+00:00"} {"global_step": 2886, "acc_step": 0, "speed/wps": 12855.545726762537, "speed/FLOPS": 201914011009832.8, "speed/curr_iter_time": 1.2854, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4223978519439697, "optim/lr": 0.001731, "optim/total_tokens": 756547584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429438, "loss/out": 6.194665908813477, "created_at": "2025-01-14T17:04:31.589192+00:00"} {"global_step": 2887, "acc_step": 0, "speed/wps": 12857.077389613847, "speed/FLOPS": 201938067879637.94, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6203648447990417, "optim/lr": 0.0017316000000000002, "optim/total_tokens": 756809728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 6.031332492828369, "created_at": "2025-01-14T17:04:36.690298+00:00"} {"global_step": 2888, "acc_step": 0, "speed/wps": 12863.173823660187, "speed/FLOPS": 202033820753713.97, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5392648577690125, "optim/lr": 0.0017322000000000001, "optim/total_tokens": 757071872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 6.14822244644165, "created_at": "2025-01-14T17:04:41.788822+00:00"} {"global_step": 2889, "acc_step": 0, "speed/wps": 12847.625518824383, "speed/FLOPS": 201789613260657.88, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.46323099732398987, "optim/lr": 0.0017328, "optim/total_tokens": 757334016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 6.036079406738281, "created_at": "2025-01-14T17:04:46.898355+00:00"} {"global_step": 2890, "acc_step": 0, "speed/wps": 12860.02056761048, "speed/FLOPS": 201984294534425.44, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5124673843383789, "optim/lr": 0.0017334, "optim/total_tokens": 757596160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 5.952236175537109, "created_at": "2025-01-14T17:04:52.001319+00:00"} {"global_step": 2891, "acc_step": 0, "speed/wps": 12861.622266132088, "speed/FLOPS": 202009451410671.72, "speed/curr_iter_time": 1.2844, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5693142414093018, "optim/lr": 0.0017339999999999999, "optim/total_tokens": 757858304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464799, "loss/out": 5.9375410079956055, "created_at": "2025-01-14T17:04:57.101380+00:00"} {"global_step": 2892, "acc_step": 0, "speed/wps": 12856.83843478915, "speed/FLOPS": 201934314765757.72, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6445496678352356, "optim/lr": 0.0017346000000000002, "optim/total_tokens": 758120448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 5.915777206420898, "created_at": "2025-01-14T17:05:02.203187+00:00"} {"global_step": 2893, "acc_step": 0, "speed/wps": 12861.640752680385, "speed/FLOPS": 202009741767315.9, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.9217785000801086, "optim/lr": 0.0017352000000000001, "optim/total_tokens": 758382592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499226, "loss/out": 6.022037982940674, "created_at": "2025-01-14T17:05:07.306826+00:00"} {"global_step": 2894, "acc_step": 0, "speed/wps": 12850.670145958553, "speed/FLOPS": 201837433313556.12, "speed/curr_iter_time": 1.2858, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9280699491500854, "optim/lr": 0.0017358, "optim/total_tokens": 758644736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 384529, "loss/out": 6.019388198852539, "created_at": "2025-01-14T17:05:12.414358+00:00"} {"global_step": 2895, "acc_step": 0, "speed/wps": 12860.993207430658, "speed/FLOPS": 201999571179348.5, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0825823545455933, "optim/lr": 0.0017364, "optim/total_tokens": 758906880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466392, "loss/out": 5.974050045013428, "created_at": "2025-01-14T17:05:17.513638+00:00"} {"global_step": 2896, "acc_step": 0, "speed/wps": 12860.316630586913, "speed/FLOPS": 201988944610303.5, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3019689321517944, "optim/lr": 0.0017369999999999998, "optim/total_tokens": 759169024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426517, "loss/out": 5.895280838012695, "created_at": "2025-01-14T17:05:22.616261+00:00"} {"global_step": 2897, "acc_step": 0, "speed/wps": 12862.058079961886, "speed/FLOPS": 202016296465739.66, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1630547046661377, "optim/lr": 0.0017376000000000002, "optim/total_tokens": 759431168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 5.922107696533203, "created_at": "2025-01-14T17:05:27.716206+00:00"} {"global_step": 2898, "acc_step": 0, "speed/wps": 12856.068706077142, "speed/FLOPS": 201922225118617.56, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.152129054069519, "optim/lr": 0.0017382, "optim/total_tokens": 759693312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 5.877375602722168, "created_at": "2025-01-14T17:05:32.817528+00:00"} {"global_step": 2899, "acc_step": 0, "speed/wps": 12859.682675141963, "speed/FLOPS": 201978987468894.47, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4718838930130005, "optim/lr": 0.0017388, "optim/total_tokens": 759955456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 5.925600528717041, "created_at": "2025-01-14T17:05:37.922574+00:00"} {"global_step": 2900, "acc_step": 0, "speed/wps": 12861.088741483105, "speed/FLOPS": 202001071672919.75, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2428903579711914, "optim/lr": 0.0017393999999999999, "optim/total_tokens": 760217600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 5.8320512771606445, "created_at": "2025-01-14T17:05:43.024380+00:00"} {"global_step": 2901, "acc_step": 0, "speed/wps": 12859.131670909344, "speed/FLOPS": 201970333190262.03, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.519144296646118, "optim/lr": 0.00174, "optim/total_tokens": 760479744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 5.839663505554199, "created_at": "2025-01-14T17:05:48.128327+00:00"} {"global_step": 2902, "acc_step": 0, "speed/wps": 12864.003041535287, "speed/FLOPS": 202046844759907.2, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.307252883911133, "optim/lr": 0.0017406000000000001, "optim/total_tokens": 760741888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 5.939519882202148, "created_at": "2025-01-14T17:05:53.247581+00:00"} {"global_step": 2903, "acc_step": 0, "speed/wps": 12864.497414742158, "speed/FLOPS": 202054609570461.06, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.001, "optim/grad_norm": 1.1391582489013672, "optim/lr": 0.0017412, "optim/total_tokens": 761004032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298409, "loss/out": 5.77297306060791, "created_at": "2025-01-14T17:05:58.347819+00:00"} {"global_step": 2904, "acc_step": 0, "speed/wps": 12860.088476795412, "speed/FLOPS": 201985361141491.8, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0910873413085938, "optim/lr": 0.0017418, "optim/total_tokens": 761266176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448605, "loss/out": 5.895415306091309, "created_at": "2025-01-14T17:06:03.447646+00:00"} {"global_step": 2905, "acc_step": 0, "speed/wps": 12860.045923892601, "speed/FLOPS": 201984692789679.34, "speed/curr_iter_time": 1.2852, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9658827185630798, "optim/lr": 0.0017424, "optim/total_tokens": 761528320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293067, "loss/out": 6.013387680053711, "created_at": "2025-01-14T17:06:08.547581+00:00"} {"global_step": 2906, "acc_step": 0, "speed/wps": 12864.808411740507, "speed/FLOPS": 202059494205674.8, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2125720977783203, "optim/lr": 0.001743, "optim/total_tokens": 761790464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 5.987446308135986, "created_at": "2025-01-14T17:06:13.649475+00:00"} {"global_step": 2907, "acc_step": 0, "speed/wps": 12863.185533372205, "speed/FLOPS": 202034004670832.06, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7526334524154663, "optim/lr": 0.0017436, "optim/total_tokens": 762052608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463248, "loss/out": 5.992156028747559, "created_at": "2025-01-14T17:06:18.755499+00:00"} {"global_step": 2908, "acc_step": 0, "speed/wps": 12868.321000625874, "speed/FLOPS": 202114664240922.44, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5901927947998047, "optim/lr": 0.0017442000000000002, "optim/total_tokens": 762314752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 5.874334335327148, "created_at": "2025-01-14T17:06:23.855215+00:00"} {"global_step": 2909, "acc_step": 0, "speed/wps": 12864.813877659028, "speed/FLOPS": 202059580055436.12, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1495813131332397, "optim/lr": 0.0017448000000000001, "optim/total_tokens": 762576896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 5.918877124786377, "created_at": "2025-01-14T17:06:28.962402+00:00"} {"global_step": 2910, "acc_step": 0, "speed/wps": 12865.026640500177, "speed/FLOPS": 202062921788223.16, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2961748838424683, "optim/lr": 0.0017454, "optim/total_tokens": 762839040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 5.852435111999512, "created_at": "2025-01-14T17:06:34.060295+00:00"} {"global_step": 2911, "acc_step": 0, "speed/wps": 12863.66263761818, "speed/FLOPS": 202041498248625.0, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2013347148895264, "optim/lr": 0.001746, "optim/total_tokens": 763101184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444357, "loss/out": 5.926914691925049, "created_at": "2025-01-14T17:06:39.163714+00:00"} {"global_step": 2912, "acc_step": 0, "speed/wps": 12861.477892343484, "speed/FLOPS": 202007183821931.3, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1452033519744873, "optim/lr": 0.0017466000000000003, "optim/total_tokens": 763363328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 5.864767074584961, "created_at": "2025-01-14T17:06:44.265280+00:00"} {"global_step": 2913, "acc_step": 0, "speed/wps": 12856.106452929736, "speed/FLOPS": 201922817984807.16, "speed/curr_iter_time": 1.2861, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2198315858840942, "optim/lr": 0.0017472000000000002, "optim/total_tokens": 763625472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300225, "loss/out": 5.732958793640137, "created_at": "2025-01-14T17:06:49.370897+00:00"} {"global_step": 2914, "acc_step": 0, "speed/wps": 12870.389356678654, "speed/FLOPS": 202147150614950.4, "speed/curr_iter_time": 1.2831, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4440956115722656, "optim/lr": 0.0017478, "optim/total_tokens": 763887616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459528, "loss/out": 5.723567008972168, "created_at": "2025-01-14T17:06:54.468147+00:00"} {"global_step": 2915, "acc_step": 0, "speed/wps": 12867.018813786512, "speed/FLOPS": 202094211607218.03, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3464899063110352, "optim/lr": 0.0017484, "optim/total_tokens": 764149760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320810, "loss/out": 5.760884761810303, "created_at": "2025-01-14T17:06:59.567970+00:00"} {"global_step": 2916, "acc_step": 0, "speed/wps": 12867.357903631084, "speed/FLOPS": 202099537479185.3, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4331835508346558, "optim/lr": 0.001749, "optim/total_tokens": 764411904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 5.782114505767822, "created_at": "2025-01-14T17:07:04.666537+00:00"} {"global_step": 2917, "acc_step": 0, "speed/wps": 12864.616303177729, "speed/FLOPS": 202056476876711.53, "speed/curr_iter_time": 1.2849, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4457979202270508, "optim/lr": 0.0017496000000000002, "optim/total_tokens": 764674048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306381, "loss/out": 5.64413595199585, "created_at": "2025-01-14T17:07:09.766613+00:00"} {"global_step": 2918, "acc_step": 0, "speed/wps": 12863.90840482802, "speed/FLOPS": 202045358360375.1, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0009, "optim/grad_norm": 6.025032997131348, "optim/lr": 0.0017502000000000001, "optim/total_tokens": 764936192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461077, "loss/out": 5.709605693817139, "created_at": "2025-01-14T17:07:14.867368+00:00"} {"global_step": 2919, "acc_step": 0, "speed/wps": 12867.868657899373, "speed/FLOPS": 202107559576818.7, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.3753819465637207, "optim/lr": 0.0017508, "optim/total_tokens": 765198336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334608, "loss/out": 5.632495880126953, "created_at": "2025-01-14T17:07:19.966513+00:00"} {"global_step": 2920, "acc_step": 0, "speed/wps": 12868.227206355688, "speed/FLOPS": 202113191073022.34, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2557921409606934, "optim/lr": 0.0017514, "optim/total_tokens": 765460480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 5.68716287612915, "created_at": "2025-01-14T17:07:25.065813+00:00"} {"global_step": 2921, "acc_step": 0, "speed/wps": 12872.759271257477, "speed/FLOPS": 202184373380014.72, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3016561269760132, "optim/lr": 0.0017519999999999999, "optim/total_tokens": 765722624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 5.667759895324707, "created_at": "2025-01-14T17:07:30.162997+00:00"} {"global_step": 2922, "acc_step": 0, "speed/wps": 12868.25510587832, "speed/FLOPS": 202113629273363.22, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3822306394577026, "optim/lr": 0.0017526000000000002, "optim/total_tokens": 765984768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 5.71450138092041, "created_at": "2025-01-14T17:07:35.259688+00:00"} {"global_step": 2923, "acc_step": 0, "speed/wps": 12871.359059425813, "speed/FLOPS": 202162381129105.75, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.4216594696044922, "optim/lr": 0.0017532000000000001, "optim/total_tokens": 766246912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 5.566981792449951, "created_at": "2025-01-14T17:07:40.355815+00:00"} {"global_step": 2924, "acc_step": 0, "speed/wps": 12855.49028929711, "speed/FLOPS": 201913140288259.2, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.705140471458435, "optim/lr": 0.0017538, "optim/total_tokens": 766509056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 5.571955680847168, "created_at": "2025-01-14T17:07:45.460723+00:00"} {"global_step": 2925, "acc_step": 0, "speed/wps": 12866.82124685999, "speed/FLOPS": 202091108547151.6, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0014, "optim/grad_norm": 1.3501566648483276, "optim/lr": 0.0017544, "optim/total_tokens": 766771200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 5.599997520446777, "created_at": "2025-01-14T17:07:50.557727+00:00"} {"global_step": 2926, "acc_step": 0, "speed/wps": 12874.589969290884, "speed/FLOPS": 202213127008272.1, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3881595134735107, "optim/lr": 0.0017549999999999998, "optim/total_tokens": 767033344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 5.5915985107421875, "created_at": "2025-01-14T17:07:55.652278+00:00"} {"global_step": 2927, "acc_step": 0, "speed/wps": 12884.557820605014, "speed/FLOPS": 202369685810423.6, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.559980869293213, "optim/lr": 0.0017556000000000002, "optim/total_tokens": 767295488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438536, "loss/out": 5.6960062980651855, "created_at": "2025-01-14T17:08:00.748259+00:00"} {"global_step": 2928, "acc_step": 0, "speed/wps": 12875.502977507162, "speed/FLOPS": 202227467056913.5, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.611521601676941, "optim/lr": 0.0017562, "optim/total_tokens": 767557632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374420, "loss/out": 5.603724479675293, "created_at": "2025-01-14T17:08:05.841893+00:00"} {"global_step": 2929, "acc_step": 0, "speed/wps": 12865.25039234732, "speed/FLOPS": 202066436118449.78, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.952029824256897, "optim/lr": 0.0017568, "optim/total_tokens": 767819776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453945, "loss/out": 5.558348178863525, "created_at": "2025-01-14T17:08:10.939687+00:00"} {"global_step": 2930, "acc_step": 0, "speed/wps": 12854.588124377487, "speed/FLOPS": 201898970548492.03, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4934717416763306, "optim/lr": 0.0017573999999999999, "optim/total_tokens": 768081920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 392297, "loss/out": 5.556066989898682, "created_at": "2025-01-14T17:08:16.041726+00:00"} {"global_step": 2931, "acc_step": 0, "speed/wps": 12873.08125577938, "speed/FLOPS": 202189430589384.3, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.161848783493042, "optim/lr": 0.001758, "optim/total_tokens": 768344064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439586, "loss/out": 5.676789283752441, "created_at": "2025-01-14T17:08:21.136333+00:00"} {"global_step": 2932, "acc_step": 0, "speed/wps": 12871.68854819616, "speed/FLOPS": 202167556203009.06, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5678261518478394, "optim/lr": 0.0017586000000000001, "optim/total_tokens": 768606208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327688, "loss/out": 5.68914270401001, "created_at": "2025-01-14T17:08:26.233917+00:00"} {"global_step": 2933, "acc_step": 0, "speed/wps": 12867.617327510725, "speed/FLOPS": 202103612087700.84, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.263217806816101, "optim/lr": 0.0017592, "optim/total_tokens": 768868352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 5.651252746582031, "created_at": "2025-01-14T17:08:31.334962+00:00"} {"global_step": 2934, "acc_step": 0, "speed/wps": 12865.620788968772, "speed/FLOPS": 202072253706368.0, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4045628309249878, "optim/lr": 0.0017598000000000002, "optim/total_tokens": 769130496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 5.668710231781006, "created_at": "2025-01-14T17:08:36.433457+00:00"} {"global_step": 2935, "acc_step": 0, "speed/wps": 12870.32820045596, "speed/FLOPS": 202146190072435.56, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.542607069015503, "optim/lr": 0.0017604, "optim/total_tokens": 769392640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 5.550506114959717, "created_at": "2025-01-14T17:08:41.530237+00:00"} {"global_step": 2936, "acc_step": 0, "speed/wps": 12874.94151922543, "speed/FLOPS": 202218648583074.34, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9505259990692139, "optim/lr": 0.001761, "optim/total_tokens": 769654784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 5.613884925842285, "created_at": "2025-01-14T17:08:46.626508+00:00"} {"global_step": 2937, "acc_step": 0, "speed/wps": 12870.640722390242, "speed/FLOPS": 202151098658863.6, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.1484103202819824, "optim/lr": 0.0017616, "optim/total_tokens": 769916928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 5.678806781768799, "created_at": "2025-01-14T17:08:51.729632+00:00"} {"global_step": 2938, "acc_step": 0, "speed/wps": 12873.873624496113, "speed/FLOPS": 202201875828909.3, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.1948556900024414, "optim/lr": 0.0017622000000000002, "optim/total_tokens": 770179072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 5.568763732910156, "created_at": "2025-01-14T17:08:56.824040+00:00"} {"global_step": 2939, "acc_step": 0, "speed/wps": 12868.259965245803, "speed/FLOPS": 202113705596407.03, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.167443037033081, "optim/lr": 0.0017628000000000001, "optim/total_tokens": 770441216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476195, "loss/out": 5.538800239562988, "created_at": "2025-01-14T17:09:01.923690+00:00"} {"global_step": 2940, "acc_step": 0, "speed/wps": 12870.92896900941, "speed/FLOPS": 202155625968109.97, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.203624248504639, "optim/lr": 0.0017634, "optim/total_tokens": 770703360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 5.498298645019531, "created_at": "2025-01-14T17:09:07.019868+00:00"} {"global_step": 2941, "acc_step": 0, "speed/wps": 12867.59150841778, "speed/FLOPS": 202103206563366.97, "speed/curr_iter_time": 1.2835, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.531553268432617, "optim/lr": 0.001764, "optim/total_tokens": 770965504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469331, "loss/out": 5.664546966552734, "created_at": "2025-01-14T17:09:12.116815+00:00"} {"global_step": 2942, "acc_step": 0, "speed/wps": 12864.438896784854, "speed/FLOPS": 202053690465529.5, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.207998514175415, "optim/lr": 0.0017645999999999998, "optim/total_tokens": 771227648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 5.684869289398193, "created_at": "2025-01-14T17:09:17.215216+00:00"} {"global_step": 2943, "acc_step": 0, "speed/wps": 12872.379242439247, "speed/FLOPS": 202178404505210.1, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.700455665588379, "optim/lr": 0.0017652000000000002, "optim/total_tokens": 771489792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 378157, "loss/out": 5.749948501586914, "created_at": "2025-01-14T17:09:22.315389+00:00"} {"global_step": 2944, "acc_step": 0, "speed/wps": 12867.554578800746, "speed/FLOPS": 202102626532984.84, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.392214059829712, "optim/lr": 0.0017658, "optim/total_tokens": 771751936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 5.709667205810547, "created_at": "2025-01-14T17:09:27.412079+00:00"} {"global_step": 2945, "acc_step": 0, "speed/wps": 12868.272267815822, "speed/FLOPS": 202113898825174.7, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.420562982559204, "optim/lr": 0.0017664, "optim/total_tokens": 772014080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325381, "loss/out": 5.691958904266357, "created_at": "2025-01-14T17:09:32.508301+00:00"} {"global_step": 2946, "acc_step": 0, "speed/wps": 12872.760665329619, "speed/FLOPS": 202184395275833.38, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.008260250091553, "optim/lr": 0.001767, "optim/total_tokens": 772276224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 5.7772321701049805, "created_at": "2025-01-14T17:09:37.609102+00:00"} {"global_step": 2947, "acc_step": 0, "speed/wps": 12867.091657740177, "speed/FLOPS": 202095355721608.4, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.862178325653076, "optim/lr": 0.0017675999999999998, "optim/total_tokens": 772538368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 5.646448135375977, "created_at": "2025-01-14T17:09:42.708147+00:00"} {"global_step": 2948, "acc_step": 0, "speed/wps": 12869.421951249366, "speed/FLOPS": 202131956183329.06, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.193027496337891, "optim/lr": 0.0017682000000000002, "optim/total_tokens": 772800512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 5.677261829376221, "created_at": "2025-01-14T17:09:47.805381+00:00"} {"global_step": 2949, "acc_step": 0, "speed/wps": 12864.150559805805, "speed/FLOPS": 202049161737057.03, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.798574924468994, "optim/lr": 0.0017688, "optim/total_tokens": 773062656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477242, "loss/out": 5.752246856689453, "created_at": "2025-01-14T17:09:52.920056+00:00"} {"global_step": 2950, "acc_step": 0, "speed/wps": 12872.665819660628, "speed/FLOPS": 202182905594270.25, "speed/curr_iter_time": 1.2847, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.9339699745178223, "optim/lr": 0.0017694, "optim/total_tokens": 773324800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 5.721176624298096, "created_at": "2025-01-14T17:09:58.014912+00:00"} {"global_step": 2951, "acc_step": 0, "speed/wps": 12869.854473630168, "speed/FLOPS": 202138749541667.6, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.8813443183898926, "optim/lr": 0.0017699999999999999, "optim/total_tokens": 773586944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459528, "loss/out": 5.85264778137207, "created_at": "2025-01-14T17:10:03.113025+00:00"} {"global_step": 2952, "acc_step": 0, "speed/wps": 12878.560885188654, "speed/FLOPS": 202275495699056.8, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.105516195297241, "optim/lr": 0.0017705999999999998, "optim/total_tokens": 773849088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 5.795292854309082, "created_at": "2025-01-14T17:10:08.205571+00:00"} {"global_step": 2953, "acc_step": 0, "speed/wps": 12867.688427611689, "speed/FLOPS": 202104728812488.25, "speed/curr_iter_time": 1.2845, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.9638476371765137, "optim/lr": 0.0017712000000000001, "optim/total_tokens": 774111232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 5.702136993408203, "created_at": "2025-01-14T17:10:13.305139+00:00"} {"global_step": 2954, "acc_step": 0, "speed/wps": 12868.987255089804, "speed/FLOPS": 202125128682808.03, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.001, "optim/grad_norm": 2.9485294818878174, "optim/lr": 0.0017718, "optim/total_tokens": 774373376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 5.791065216064453, "created_at": "2025-01-14T17:10:18.401822+00:00"} {"global_step": 2955, "acc_step": 0, "speed/wps": 12873.048477047374, "speed/FLOPS": 202188915754355.44, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.543067693710327, "optim/lr": 0.0017724, "optim/total_tokens": 774635520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 5.840756416320801, "created_at": "2025-01-14T17:10:23.496937+00:00"} {"global_step": 2956, "acc_step": 0, "speed/wps": 12867.90429983378, "speed/FLOPS": 202108119382375.72, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.582166910171509, "optim/lr": 0.001773, "optim/total_tokens": 774897664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 5.771249771118164, "created_at": "2025-01-14T17:10:28.597917+00:00"} {"global_step": 2957, "acc_step": 0, "speed/wps": 12873.240863436811, "speed/FLOPS": 202191937446968.16, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.359656810760498, "optim/lr": 0.0017736, "optim/total_tokens": 775159808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 5.775822162628174, "created_at": "2025-01-14T17:10:33.694056+00:00"} {"global_step": 2958, "acc_step": 0, "speed/wps": 12878.006739563496, "speed/FLOPS": 202266792080537.84, "speed/curr_iter_time": 1.2834, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.3017611503601074, "optim/lr": 0.0017742, "optim/total_tokens": 775421952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 5.839643478393555, "created_at": "2025-01-14T17:10:38.789586+00:00"} {"global_step": 2959, "acc_step": 0, "speed/wps": 12863.064698634145, "speed/FLOPS": 202032106795226.5, "speed/curr_iter_time": 1.2859, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.015767574310303, "optim/lr": 0.0017748, "optim/total_tokens": 775684096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466392, "loss/out": 5.84531307220459, "created_at": "2025-01-14T17:10:43.890866+00:00"} {"global_step": 2960, "acc_step": 0, "speed/wps": 12860.671702742642, "speed/FLOPS": 201994521506429.5, "speed/curr_iter_time": 1.2843, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.593592882156372, "optim/lr": 0.0017754000000000001, "optim/total_tokens": 775946240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310132, "loss/out": 5.891631126403809, "created_at": "2025-01-14T17:10:48.992306+00:00"} {"global_step": 2961, "acc_step": 0, "speed/wps": 12873.728876612444, "speed/FLOPS": 202199602364491.84, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.972315788269043, "optim/lr": 0.001776, "optim/total_tokens": 776208384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 5.835482597351074, "created_at": "2025-01-14T17:10:54.088957+00:00"} {"global_step": 2962, "acc_step": 0, "speed/wps": 12863.692313924126, "speed/FLOPS": 202041964355789.44, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.721273183822632, "optim/lr": 0.0017766, "optim/total_tokens": 776470528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301523, "loss/out": 5.751204490661621, "created_at": "2025-01-14T17:10:59.187386+00:00"} {"global_step": 2963, "acc_step": 0, "speed/wps": 12869.946545087243, "speed/FLOPS": 202140195650420.4, "speed/curr_iter_time": 1.2838, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.8634352684021, "optim/lr": 0.0017772, "optim/total_tokens": 776732672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 5.928203582763672, "created_at": "2025-01-14T17:11:04.283456+00:00"} {"global_step": 2964, "acc_step": 0, "speed/wps": 12872.673039534046, "speed/FLOPS": 202183018992303.44, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.2893500328063965, "optim/lr": 0.0017778000000000002, "optim/total_tokens": 776994816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321779, "loss/out": 5.9933576583862305, "created_at": "2025-01-14T17:11:09.382785+00:00"} {"global_step": 2965, "acc_step": 0, "speed/wps": 12870.528431576573, "speed/FLOPS": 202149334977331.16, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.4030725955963135, "optim/lr": 0.0017784, "optim/total_tokens": 777256960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 5.884539604187012, "created_at": "2025-01-14T17:11:14.478892+00:00"} {"global_step": 2966, "acc_step": 0, "speed/wps": 12866.794040533894, "speed/FLOPS": 202090681234414.84, "speed/curr_iter_time": 1.2848, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.223341703414917, "optim/lr": 0.001779, "optim/total_tokens": 777519104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325239, "loss/out": 5.9169392585754395, "created_at": "2025-01-14T17:11:19.582705+00:00"} {"global_step": 2967, "acc_step": 0, "speed/wps": 12872.646444254317, "speed/FLOPS": 202182601276889.53, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.323321342468262, "optim/lr": 0.0017795999999999999, "optim/total_tokens": 777781248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461077, "loss/out": 5.870033264160156, "created_at": "2025-01-14T17:11:24.685553+00:00"} {"global_step": 2968, "acc_step": 0, "speed/wps": 12855.940348237462, "speed/FLOPS": 201920209082363.22, "speed/curr_iter_time": 1.2867, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.885368824005127, "optim/lr": 0.0017802000000000002, "optim/total_tokens": 778043392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366026, "loss/out": 5.980596542358398, "created_at": "2025-01-14T17:11:29.788131+00:00"} {"global_step": 2969, "acc_step": 0, "speed/wps": 12877.046912176196, "speed/FLOPS": 202251716672479.25, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.130101203918457, "optim/lr": 0.0017808000000000001, "optim/total_tokens": 778305536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448890, "loss/out": 5.862586975097656, "created_at": "2025-01-14T17:11:34.881124+00:00"} {"global_step": 2970, "acc_step": 0, "speed/wps": 12867.460948642096, "speed/FLOPS": 202101155942681.94, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.9435021877288818, "optim/lr": 0.0017814, "optim/total_tokens": 778567680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320571, "loss/out": 5.8465142250061035, "created_at": "2025-01-14T17:11:39.978091+00:00"} {"global_step": 2971, "acc_step": 0, "speed/wps": 12880.645598788995, "speed/FLOPS": 202308238990846.88, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.8685965538024902, "optim/lr": 0.001782, "optim/total_tokens": 778829824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 5.989499092102051, "created_at": "2025-01-14T17:11:45.070252+00:00"} {"global_step": 2972, "acc_step": 0, "speed/wps": 12869.510923803997, "speed/FLOPS": 202133353619560.66, "speed/curr_iter_time": 1.2851, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.5652531385421753, "optim/lr": 0.0017825999999999999, "optim/total_tokens": 779091968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454995, "loss/out": 6.183463096618652, "created_at": "2025-01-14T17:11:50.174095+00:00"} {"global_step": 2973, "acc_step": 0, "speed/wps": 12871.182582942798, "speed/FLOPS": 202159609323435.9, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.9002797603607178, "optim/lr": 0.0017832000000000002, "optim/total_tokens": 779354112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 6.163747787475586, "created_at": "2025-01-14T17:11:55.272594+00:00"} {"global_step": 2974, "acc_step": 0, "speed/wps": 12874.505993153269, "speed/FLOPS": 202211808047635.53, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7287033796310425, "optim/lr": 0.0017838, "optim/total_tokens": 779616256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 6.1725592613220215, "created_at": "2025-01-14T17:12:00.370152+00:00"} {"global_step": 2975, "acc_step": 0, "speed/wps": 12866.175842838755, "speed/FLOPS": 202080971590120.03, "speed/curr_iter_time": 1.2829, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5497514009475708, "optim/lr": 0.0017844, "optim/total_tokens": 779878400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 6.07270622253418, "created_at": "2025-01-14T17:12:05.467690+00:00"} {"global_step": 2976, "acc_step": 0, "speed/wps": 12878.75668438805, "speed/FLOPS": 202278570994539.66, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.3568434715270996, "optim/lr": 0.001785, "optim/total_tokens": 780140544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 6.257919788360596, "created_at": "2025-01-14T17:12:10.561271+00:00"} {"global_step": 2977, "acc_step": 0, "speed/wps": 12865.721451529205, "speed/FLOPS": 202073834750204.47, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.4520214796066284, "optim/lr": 0.0017855999999999998, "optim/total_tokens": 780402688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 6.116578102111816, "created_at": "2025-01-14T17:12:15.659828+00:00"} {"global_step": 2978, "acc_step": 0, "speed/wps": 12874.232507678244, "speed/FLOPS": 202207512582441.84, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6063157320022583, "optim/lr": 0.0017862000000000002, "optim/total_tokens": 780664832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 6.1892313957214355, "created_at": "2025-01-14T17:12:20.754168+00:00"} {"global_step": 2979, "acc_step": 0, "speed/wps": 12876.932407662818, "speed/FLOPS": 202249918221750.97, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7193810939788818, "optim/lr": 0.0017868, "optim/total_tokens": 780926976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 386926, "loss/out": 6.214531898498535, "created_at": "2025-01-14T17:12:25.857486+00:00"} {"global_step": 2980, "acc_step": 0, "speed/wps": 12880.905202714895, "speed/FLOPS": 202312316427236.3, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.6373965740203857, "optim/lr": 0.0017874, "optim/total_tokens": 781189120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454182, "loss/out": 6.213119029998779, "created_at": "2025-01-14T17:12:30.952281+00:00"} {"global_step": 2981, "acc_step": 0, "speed/wps": 12877.475257654865, "speed/FLOPS": 202258444426830.12, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8290470838546753, "optim/lr": 0.0017879999999999999, "optim/total_tokens": 781451264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368286, "loss/out": 6.097745418548584, "created_at": "2025-01-14T17:12:36.045319+00:00"} {"global_step": 2982, "acc_step": 0, "speed/wps": 12875.746864976207, "speed/FLOPS": 202231297644751.34, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.140019416809082, "optim/lr": 0.0017886, "optim/total_tokens": 781713408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447026, "loss/out": 6.103629112243652, "created_at": "2025-01-14T17:12:41.138819+00:00"} {"global_step": 2983, "acc_step": 0, "speed/wps": 12876.318685173186, "speed/FLOPS": 202240278866708.22, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5386686325073242, "optim/lr": 0.0017892000000000001, "optim/total_tokens": 781975552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314047, "loss/out": 6.263134002685547, "created_at": "2025-01-14T17:12:46.231950+00:00"} {"global_step": 2984, "acc_step": 0, "speed/wps": 12875.357992738805, "speed/FLOPS": 202225189871896.5, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2041975259780884, "optim/lr": 0.0017898, "optim/total_tokens": 782237696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 6.073573112487793, "created_at": "2025-01-14T17:12:51.326884+00:00"} {"global_step": 2985, "acc_step": 0, "speed/wps": 12877.321448063185, "speed/FLOPS": 202256028635836.25, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.588083744049072, "optim/lr": 0.0017904, "optim/total_tokens": 782499840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 6.028851509094238, "created_at": "2025-01-14T17:12:56.419745+00:00"} {"global_step": 2986, "acc_step": 0, "speed/wps": 12869.130121808965, "speed/FLOPS": 202127372600951.9, "speed/curr_iter_time": 1.283, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.6201996803283691, "optim/lr": 0.001791, "optim/total_tokens": 782761984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369243, "loss/out": 6.130796909332275, "created_at": "2025-01-14T17:13:01.515710+00:00"} {"global_step": 2987, "acc_step": 0, "speed/wps": 12879.788158509898, "speed/FLOPS": 202294771713017.25, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6469954252243042, "optim/lr": 0.0017916, "optim/total_tokens": 783024128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 5.964956760406494, "created_at": "2025-01-14T17:13:06.607886+00:00"} {"global_step": 2988, "acc_step": 0, "speed/wps": 12878.939761117836, "speed/FLOPS": 202281446466154.72, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2597997188568115, "optim/lr": 0.0017922, "optim/total_tokens": 783286272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444119, "loss/out": 6.007758140563965, "created_at": "2025-01-14T17:13:11.705736+00:00"} {"global_step": 2989, "acc_step": 0, "speed/wps": 12872.78273032924, "speed/FLOPS": 202184741836972.9, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.9270764589309692, "optim/lr": 0.0017928, "optim/total_tokens": 783548416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 6.046921730041504, "created_at": "2025-01-14T17:13:16.805113+00:00"} {"global_step": 2990, "acc_step": 0, "speed/wps": 12878.8086645354, "speed/FLOPS": 202279387414181.97, "speed/curr_iter_time": 1.2837, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3160604238510132, "optim/lr": 0.0017934000000000001, "optim/total_tokens": 783810560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460556, "loss/out": 5.9401750564575195, "created_at": "2025-01-14T17:13:21.897272+00:00"} {"global_step": 2991, "acc_step": 0, "speed/wps": 12878.979313597276, "speed/FLOPS": 202282067692195.12, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 14.193866729736328, "optim/lr": 0.001794, "optim/total_tokens": 784072704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 5.9427995681762695, "created_at": "2025-01-14T17:13:26.989584+00:00"} {"global_step": 2992, "acc_step": 0, "speed/wps": 12881.498889883935, "speed/FLOPS": 202321641100037.75, "speed/curr_iter_time": 1.2841, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5843616724014282, "optim/lr": 0.0017946, "optim/total_tokens": 784334848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 6.219351291656494, "created_at": "2025-01-14T17:13:32.081233+00:00"} {"global_step": 2993, "acc_step": 0, "speed/wps": 12884.135804865797, "speed/FLOPS": 202363057473328.78, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2465488910675049, "optim/lr": 0.0017952000000000003, "optim/total_tokens": 784596992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368286, "loss/out": 6.064493179321289, "created_at": "2025-01-14T17:13:37.175994+00:00"} {"global_step": 2994, "acc_step": 0, "speed/wps": 12874.906116261727, "speed/FLOPS": 202218092530880.66, "speed/curr_iter_time": 1.2825, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4218083620071411, "optim/lr": 0.0017958000000000002, "optim/total_tokens": 784859136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444357, "loss/out": 6.054861068725586, "created_at": "2025-01-14T17:13:42.277459+00:00"} {"global_step": 2995, "acc_step": 0, "speed/wps": 12879.304532345504, "speed/FLOPS": 202287175699533.44, "speed/curr_iter_time": 1.2836, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3818433284759521, "optim/lr": 0.0017964, "optim/total_tokens": 785121280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345083, "loss/out": 6.06024169921875, "created_at": "2025-01-14T17:13:47.371721+00:00"} {"global_step": 2996, "acc_step": 0, "speed/wps": 12884.575004822216, "speed/FLOPS": 202369955712168.34, "speed/curr_iter_time": 1.2826, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.266558051109314, "optim/lr": 0.001797, "optim/total_tokens": 785383424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461385, "loss/out": 6.012670516967773, "created_at": "2025-01-14T17:13:52.464048+00:00"} {"global_step": 2997, "acc_step": 0, "speed/wps": 12883.862013913656, "speed/FLOPS": 202358757210197.22, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4364376068115234, "optim/lr": 0.0017976, "optim/total_tokens": 785645568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324848, "loss/out": 5.956330299377441, "created_at": "2025-01-14T17:13:57.555452+00:00"} {"global_step": 2998, "acc_step": 0, "speed/wps": 12874.576768963761, "speed/FLOPS": 202212919678995.47, "speed/curr_iter_time": 1.2832, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4260225296020508, "optim/lr": 0.0017982000000000002, "optim/total_tokens": 785907712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 5.992611408233643, "created_at": "2025-01-14T17:14:02.649650+00:00"} {"global_step": 2999, "acc_step": 0, "speed/wps": 12872.407258867153, "speed/FLOPS": 202178844541708.97, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.434693455696106, "optim/lr": 0.0017988000000000001, "optim/total_tokens": 786169856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 5.952284812927246, "created_at": "2025-01-14T17:14:07.748820+00:00"} {"global_step": 3000, "acc_step": 0, "speed/wps": 12870.267200174472, "speed/FLOPS": 202145231979191.6, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.031731128692627, "optim/lr": 0.0017994, "optim/total_tokens": 786432000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 384529, "loss/out": 5.9633612632751465, "created_at": "2025-01-14T17:14:12.845072+00:00"} {"global_step": 3001, "acc_step": 0, "speed/wps": 12222.269391088555, "speed/FLOPS": 191967535945195.9, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.2759, "optim/grad_norm": 1.6777311563491821, "optim/lr": 0.0018, "optim/total_tokens": 786694144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 5.881398677825928, "created_at": "2025-01-14T17:14:18.216654+00:00"} {"global_step": 3002, "acc_step": 0, "speed/wps": 12893.670416494502, "speed/FLOPS": 202512811650886.97, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.5237996578216553, "optim/lr": 0.0018005999999999999, "optim/total_tokens": 786956288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467205, "loss/out": 5.883975982666016, "created_at": "2025-01-14T17:14:23.304800+00:00"} {"global_step": 3003, "acc_step": 0, "speed/wps": 12889.918246249272, "speed/FLOPS": 202453878661163.38, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5854976177215576, "optim/lr": 0.0018012000000000002, "optim/total_tokens": 787218432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326344, "loss/out": 5.888504505157471, "created_at": "2025-01-14T17:14:28.393126+00:00"} {"global_step": 3004, "acc_step": 0, "speed/wps": 12889.656446733792, "speed/FLOPS": 202449766740022.53, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6487922668457031, "optim/lr": 0.0018018, "optim/total_tokens": 787480576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 5.86959171295166, "created_at": "2025-01-14T17:14:33.484623+00:00"} {"global_step": 3005, "acc_step": 0, "speed/wps": 12890.405007558851, "speed/FLOPS": 202461523916410.47, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.001, "optim/grad_norm": 1.7766928672790527, "optim/lr": 0.0018024, "optim/total_tokens": 787742720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 5.8018107414245605, "created_at": "2025-01-14T17:14:38.572485+00:00"} {"global_step": 3006, "acc_step": 0, "speed/wps": 12893.732555485269, "speed/FLOPS": 202513787629124.88, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0975399017333984, "optim/lr": 0.001803, "optim/total_tokens": 788004864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 5.838774681091309, "created_at": "2025-01-14T17:14:43.659360+00:00"} {"global_step": 3007, "acc_step": 0, "speed/wps": 12886.72916783447, "speed/FLOPS": 202403789802405.97, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.9561781883239746, "optim/lr": 0.0018035999999999998, "optim/total_tokens": 788267008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 5.7204060554504395, "created_at": "2025-01-14T17:14:48.754919+00:00"} {"global_step": 3008, "acc_step": 0, "speed/wps": 12891.799128003664, "speed/FLOPS": 202483420493718.5, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.3573155403137207, "optim/lr": 0.0018042000000000002, "optim/total_tokens": 788529152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440909, "loss/out": 5.7723894119262695, "created_at": "2025-01-14T17:14:53.843941+00:00"} {"global_step": 3009, "acc_step": 0, "speed/wps": 12897.275506096685, "speed/FLOPS": 202569434537001.47, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.4940996170043945, "optim/lr": 0.0018048, "optim/total_tokens": 788791296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452615, "loss/out": 5.7251176834106445, "created_at": "2025-01-14T17:14:58.929828+00:00"} {"global_step": 3010, "acc_step": 0, "speed/wps": 12891.808702839275, "speed/FLOPS": 202483570879668.97, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.9499974250793457, "optim/lr": 0.0018054, "optim/total_tokens": 789053440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328460, "loss/out": 5.826386451721191, "created_at": "2025-01-14T17:15:04.017144+00:00"} {"global_step": 3011, "acc_step": 0, "speed/wps": 12886.542162184489, "speed/FLOPS": 202400852621700.3, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.7055869102478027, "optim/lr": 0.0018059999999999999, "optim/total_tokens": 789315584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 5.8589677810668945, "created_at": "2025-01-14T17:15:09.112139+00:00"} {"global_step": 3012, "acc_step": 0, "speed/wps": 12888.45700492716, "speed/FLOPS": 202430927858243.62, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0009, "optim/grad_norm": 2.6328370571136475, "optim/lr": 0.0018066, "optim/total_tokens": 789577728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 5.831650257110596, "created_at": "2025-01-14T17:15:14.202817+00:00"} {"global_step": 3013, "acc_step": 0, "speed/wps": 12892.457037793063, "speed/FLOPS": 202493753871023.8, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5669119358062744, "optim/lr": 0.0018072000000000001, "optim/total_tokens": 789839872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328985, "loss/out": 5.868463516235352, "created_at": "2025-01-14T17:15:19.290886+00:00"} {"global_step": 3014, "acc_step": 0, "speed/wps": 12890.215023939534, "speed/FLOPS": 202458539962604.25, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.553833484649658, "optim/lr": 0.0018078, "optim/total_tokens": 790102016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 5.809045791625977, "created_at": "2025-01-14T17:15:24.382468+00:00"} {"global_step": 3015, "acc_step": 0, "speed/wps": 12891.377294324715, "speed/FLOPS": 202476795016130.38, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.3755569458007812, "optim/lr": 0.0018084, "optim/total_tokens": 790364160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480663, "loss/out": 6.005258560180664, "created_at": "2025-01-14T17:15:29.473138+00:00"} {"global_step": 3016, "acc_step": 0, "speed/wps": 12894.684482409933, "speed/FLOPS": 202528738949561.4, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.8765814304351807, "optim/lr": 0.001809, "optim/total_tokens": 790626304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465057, "loss/out": 5.860687732696533, "created_at": "2025-01-14T17:15:34.563364+00:00"} {"global_step": 3017, "acc_step": 0, "speed/wps": 12891.253239462221, "speed/FLOPS": 202474846564045.7, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3161675930023193, "optim/lr": 0.0018096, "optim/total_tokens": 790888448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453131, "loss/out": 5.831653594970703, "created_at": "2025-01-14T17:15:39.650732+00:00"} {"global_step": 3018, "acc_step": 0, "speed/wps": 12893.977871195902, "speed/FLOPS": 202517640649459.5, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.082038640975952, "optim/lr": 0.0018102, "optim/total_tokens": 791150592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309558, "loss/out": 5.838061332702637, "created_at": "2025-01-14T17:15:44.739610+00:00"} {"global_step": 3019, "acc_step": 0, "speed/wps": 12890.962225651872, "speed/FLOPS": 202470275792259.25, "speed/curr_iter_time": 1.2833, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.9274301528930664, "optim/lr": 0.0018108000000000002, "optim/total_tokens": 791412736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475620, "loss/out": 5.91648006439209, "created_at": "2025-01-14T17:15:49.827959+00:00"} {"global_step": 3020, "acc_step": 0, "speed/wps": 12892.449292591205, "speed/FLOPS": 202493632221985.6, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.132287502288818, "optim/lr": 0.0018114000000000001, "optim/total_tokens": 791674880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444357, "loss/out": 5.884151458740234, "created_at": "2025-01-14T17:15:54.919905+00:00"} {"global_step": 3021, "acc_step": 0, "speed/wps": 12895.650017958873, "speed/FLOPS": 202543904012142.38, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.743453502655029, "optim/lr": 0.001812, "optim/total_tokens": 791937024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 5.796868801116943, "created_at": "2025-01-14T17:16:00.007362+00:00"} {"global_step": 3022, "acc_step": 0, "speed/wps": 12895.297364597578, "speed/FLOPS": 202538365106506.75, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.050280570983887, "optim/lr": 0.0018126, "optim/total_tokens": 792199168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 5.959926605224609, "created_at": "2025-01-14T17:16:05.096120+00:00"} {"global_step": 3023, "acc_step": 0, "speed/wps": 12890.973962396502, "speed/FLOPS": 202470460133961.62, "speed/curr_iter_time": 1.2827, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.333007335662842, "optim/lr": 0.0018132000000000003, "optim/total_tokens": 792461312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362676, "loss/out": 5.96086311340332, "created_at": "2025-01-14T17:16:10.185502+00:00"} {"global_step": 3024, "acc_step": 0, "speed/wps": 12891.611568958782, "speed/FLOPS": 202480474621185.5, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.268041610717773, "optim/lr": 0.0018138000000000002, "optim/total_tokens": 792723456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453409, "loss/out": 5.879972457885742, "created_at": "2025-01-14T17:16:15.273509+00:00"} {"global_step": 3025, "acc_step": 0, "speed/wps": 12877.890294888586, "speed/FLOPS": 202264963156906.62, "speed/curr_iter_time": 1.2842, "speed/data_load_time": 0.0011, "optim/grad_norm": 2.4428868293762207, "optim/lr": 0.0018144, "optim/total_tokens": 792985600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457380, "loss/out": 5.94272518157959, "created_at": "2025-01-14T17:16:20.367691+00:00"} {"global_step": 3026, "acc_step": 0, "speed/wps": 12892.748045871656, "speed/FLOPS": 202498324552785.12, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1201677322387695, "optim/lr": 0.001815, "optim/total_tokens": 793247744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368810, "loss/out": 5.838842391967773, "created_at": "2025-01-14T17:16:25.455044+00:00"} {"global_step": 3027, "acc_step": 0, "speed/wps": 12887.925191201528, "speed/FLOPS": 202422574992893.94, "speed/curr_iter_time": 1.2828, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.227447032928467, "optim/lr": 0.0018156, "optim/total_tokens": 793509888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 5.910170078277588, "created_at": "2025-01-14T17:16:30.544041+00:00"} {"global_step": 3028, "acc_step": 0, "speed/wps": 12894.140563323497, "speed/FLOPS": 202520195952885.78, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.091474533081055, "optim/lr": 0.0018162000000000002, "optim/total_tokens": 793772032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443543, "loss/out": 5.941120624542236, "created_at": "2025-01-14T17:16:35.630388+00:00"} {"global_step": 3029, "acc_step": 0, "speed/wps": 12897.73742438527, "speed/FLOPS": 202576689598465.97, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.8675355911254883, "optim/lr": 0.0018168000000000001, "optim/total_tokens": 794034176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470624, "loss/out": 5.7991814613342285, "created_at": "2025-01-14T17:16:40.716215+00:00"} {"global_step": 3030, "acc_step": 0, "speed/wps": 12888.951429920635, "speed/FLOPS": 202438693482177.28, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.8308117389678955, "optim/lr": 0.0018174, "optim/total_tokens": 794296320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464298, "loss/out": 6.036623477935791, "created_at": "2025-01-14T17:16:45.809137+00:00"} {"global_step": 3031, "acc_step": 0, "speed/wps": 12885.618961212645, "speed/FLOPS": 202386352481826.44, "speed/curr_iter_time": 1.2824, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.6094419956207275, "optim/lr": 0.001818, "optim/total_tokens": 794558464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365072, "loss/out": 6.042702674865723, "created_at": "2025-01-14T17:16:50.906171+00:00"} {"global_step": 3032, "acc_step": 0, "speed/wps": 12888.650258188458, "speed/FLOPS": 202433963166270.12, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.2926297187805176, "optim/lr": 0.0018185999999999999, "optim/total_tokens": 794820608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435852, "loss/out": 6.086207866668701, "created_at": "2025-01-14T17:16:55.995664+00:00"} {"global_step": 3033, "acc_step": 0, "speed/wps": 12891.791274948566, "speed/FLOPS": 202483297150696.94, "speed/curr_iter_time": 1.2823, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.334197044372559, "optim/lr": 0.0018192000000000002, "optim/total_tokens": 795082752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318319, "loss/out": 5.990421295166016, "created_at": "2025-01-14T17:17:01.086975+00:00"} {"global_step": 3034, "acc_step": 0, "speed/wps": 12896.972368631368, "speed/FLOPS": 202564673346554.16, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.650984764099121, "optim/lr": 0.0018198, "optim/total_tokens": 795344896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 6.049737930297852, "created_at": "2025-01-14T17:17:06.172371+00:00"} {"global_step": 3035, "acc_step": 0, "speed/wps": 12894.383269522603, "speed/FLOPS": 202524007987256.0, "speed/curr_iter_time": 1.2818, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6581135988235474, "optim/lr": 0.0018204, "optim/total_tokens": 795607040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 5.955948829650879, "created_at": "2025-01-14T17:17:11.259141+00:00"} {"global_step": 3036, "acc_step": 0, "speed/wps": 12891.350189518522, "speed/FLOPS": 202476369297903.22, "speed/curr_iter_time": 1.2811, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.017946481704712, "optim/lr": 0.001821, "optim/total_tokens": 795869184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 5.954556465148926, "created_at": "2025-01-14T17:17:16.349506+00:00"} {"global_step": 3037, "acc_step": 0, "speed/wps": 12891.484536197258, "speed/FLOPS": 202478479397104.44, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0012, "optim/grad_norm": 3.8019165992736816, "optim/lr": 0.0018215999999999998, "optim/total_tokens": 796131328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463248, "loss/out": 6.0815653800964355, "created_at": "2025-01-14T17:17:21.436889+00:00"} {"global_step": 3038, "acc_step": 0, "speed/wps": 12896.101035619597, "speed/FLOPS": 202550987864267.25, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.440563440322876, "optim/lr": 0.0018222000000000002, "optim/total_tokens": 796393472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328746, "loss/out": 6.022475242614746, "created_at": "2025-01-14T17:17:26.522576+00:00"} {"global_step": 3039, "acc_step": 0, "speed/wps": 12901.16188360094, "speed/FLOPS": 202630475436148.16, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.1727664470672607, "optim/lr": 0.0018228, "optim/total_tokens": 796655616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421165, "loss/out": 5.96496057510376, "created_at": "2025-01-14T17:17:31.606142+00:00"} {"global_step": 3040, "acc_step": 0, "speed/wps": 12896.139526051124, "speed/FLOPS": 202551592409385.78, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6441006660461426, "optim/lr": 0.0018234, "optim/total_tokens": 796917760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 5.936215877532959, "created_at": "2025-01-14T17:17:36.692822+00:00"} {"global_step": 3041, "acc_step": 0, "speed/wps": 12896.145604752586, "speed/FLOPS": 202551687883745.25, "speed/curr_iter_time": 1.2815, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.4147706031799316, "optim/lr": 0.001824, "optim/total_tokens": 797179904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446505, "loss/out": 6.021082878112793, "created_at": "2025-01-14T17:17:41.783853+00:00"} {"global_step": 3042, "acc_step": 0, "speed/wps": 12886.611701636524, "speed/FLOPS": 202401944834351.75, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.1673672199249268, "optim/lr": 0.0018246, "optim/total_tokens": 797442048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 5.973382949829102, "created_at": "2025-01-14T17:17:46.877066+00:00"} {"global_step": 3043, "acc_step": 0, "speed/wps": 12891.926192945832, "speed/FLOPS": 202485416223240.75, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5474493503570557, "optim/lr": 0.0018252000000000001, "optim/total_tokens": 797704192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 5.983297348022461, "created_at": "2025-01-14T17:17:51.967696+00:00"} {"global_step": 3044, "acc_step": 0, "speed/wps": 12897.110624003397, "speed/FLOPS": 202566844837154.66, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.6630022525787354, "optim/lr": 0.0018258, "optim/total_tokens": 797966336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440105, "loss/out": 5.998872756958008, "created_at": "2025-01-14T17:17:57.056645+00:00"} {"global_step": 3045, "acc_step": 0, "speed/wps": 12898.897741575403, "speed/FLOPS": 202594913974379.9, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.590498447418213, "optim/lr": 0.0018264000000000002, "optim/total_tokens": 798228480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437717, "loss/out": 5.989227294921875, "created_at": "2025-01-14T17:18:02.141644+00:00"} {"global_step": 3046, "acc_step": 0, "speed/wps": 12893.899106621655, "speed/FLOPS": 202516403543587.06, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.60720443725586, "optim/lr": 0.001827, "optim/total_tokens": 798490624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327402, "loss/out": 6.001278877258301, "created_at": "2025-01-14T17:18:07.254953+00:00"} {"global_step": 3047, "acc_step": 0, "speed/wps": 12887.269006531778, "speed/FLOPS": 202412268710963.28, "speed/curr_iter_time": 1.285, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.1979522705078125, "optim/lr": 0.0018276, "optim/total_tokens": 798752768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 6.15160608291626, "created_at": "2025-01-14T17:18:12.346939+00:00"} {"global_step": 3048, "acc_step": 0, "speed/wps": 12894.961349349787, "speed/FLOPS": 202533087525306.84, "speed/curr_iter_time": 1.2809, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.0906589031219482, "optim/lr": 0.0018282, "optim/total_tokens": 799014912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355742, "loss/out": 5.991233825683594, "created_at": "2025-01-14T17:18:17.437454+00:00"} {"global_step": 3049, "acc_step": 0, "speed/wps": 12894.634611660633, "speed/FLOPS": 202527955660914.03, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9382636547088623, "optim/lr": 0.0018288000000000002, "optim/total_tokens": 799277056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464799, "loss/out": 5.979921340942383, "created_at": "2025-01-14T17:18:22.525687+00:00"} {"global_step": 3050, "acc_step": 0, "speed/wps": 12894.198385463083, "speed/FLOPS": 202521104129043.9, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6755180358886719, "optim/lr": 0.0018294000000000001, "optim/total_tokens": 799539200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445691, "loss/out": 5.960273742675781, "created_at": "2025-01-14T17:18:27.614885+00:00"} {"global_step": 3051, "acc_step": 0, "speed/wps": 12893.050498155677, "speed/FLOPS": 202503074981518.66, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.738349676132202, "optim/lr": 0.00183, "optim/total_tokens": 799801344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 6.077316761016846, "created_at": "2025-01-14T17:18:32.702973+00:00"} {"global_step": 3052, "acc_step": 0, "speed/wps": 12896.726300784696, "speed/FLOPS": 202560808512889.5, "speed/curr_iter_time": 1.2816, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4041929244995117, "optim/lr": 0.0018306, "optim/total_tokens": 800063488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453945, "loss/out": 5.943408489227295, "created_at": "2025-01-14T17:18:37.788805+00:00"} {"global_step": 3053, "acc_step": 0, "speed/wps": 12899.889164099868, "speed/FLOPS": 202610485627483.94, "speed/curr_iter_time": 1.2812, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6902492046356201, "optim/lr": 0.0018312000000000003, "optim/total_tokens": 800325632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 6.080635070800781, "created_at": "2025-01-14T17:18:42.874317+00:00"} {"global_step": 3054, "acc_step": 0, "speed/wps": 12899.267195805664, "speed/FLOPS": 202600716760749.47, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4666454792022705, "optim/lr": 0.0018318000000000002, "optim/total_tokens": 800587776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440909, "loss/out": 5.979854583740234, "created_at": "2025-01-14T17:18:47.959712+00:00"} {"global_step": 3055, "acc_step": 0, "speed/wps": 12894.737297716734, "speed/FLOPS": 202529568486530.53, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3606877326965332, "optim/lr": 0.0018324, "optim/total_tokens": 800849920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443306, "loss/out": 6.111673831939697, "created_at": "2025-01-14T17:18:53.046165+00:00"} {"global_step": 3056, "acc_step": 0, "speed/wps": 12894.624741995385, "speed/FLOPS": 202527800644259.0, "speed/curr_iter_time": 1.2822, "speed/data_load_time": 0.001, "optim/grad_norm": 1.2062183618545532, "optim/lr": 0.001833, "optim/total_tokens": 801112064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 341043, "loss/out": 6.011229991912842, "created_at": "2025-01-14T17:18:58.132339+00:00"} {"global_step": 3057, "acc_step": 0, "speed/wps": 12895.752542942644, "speed/FLOPS": 202545514307896.6, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2262017726898193, "optim/lr": 0.0018336, "optim/total_tokens": 801374208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 5.991436958312988, "created_at": "2025-01-14T17:19:03.217856+00:00"} {"global_step": 3058, "acc_step": 0, "speed/wps": 12890.971874812996, "speed/FLOPS": 202470427345594.06, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5140165090560913, "optim/lr": 0.0018342000000000002, "optim/total_tokens": 801636352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 6.031024932861328, "created_at": "2025-01-14T17:19:08.305465+00:00"} {"global_step": 3059, "acc_step": 0, "speed/wps": 12900.047604456731, "speed/FLOPS": 202612974151008.34, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3123677968978882, "optim/lr": 0.0018348000000000001, "optim/total_tokens": 801898496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315181, "loss/out": 5.894944667816162, "created_at": "2025-01-14T17:19:13.391625+00:00"} {"global_step": 3060, "acc_step": 0, "speed/wps": 12897.46762211194, "speed/FLOPS": 202572451983016.16, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2539218664169312, "optim/lr": 0.0018354, "optim/total_tokens": 802160640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472483, "loss/out": 5.934938907623291, "created_at": "2025-01-14T17:19:18.493094+00:00"} {"global_step": 3061, "acc_step": 0, "speed/wps": 12908.617959558054, "speed/FLOPS": 202747583354776.66, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.34700608253479, "optim/lr": 0.001836, "optim/total_tokens": 802422784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338641, "loss/out": 5.977286338806152, "created_at": "2025-01-14T17:19:23.573921+00:00"} {"global_step": 3062, "acc_step": 0, "speed/wps": 12900.569963285663, "speed/FLOPS": 202621178514212.97, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3311043977737427, "optim/lr": 0.0018365999999999999, "optim/total_tokens": 802684928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 6.033289909362793, "created_at": "2025-01-14T17:19:28.657813+00:00"} {"global_step": 3063, "acc_step": 0, "speed/wps": 12895.937906428293, "speed/FLOPS": 202548425696154.84, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4541397094726562, "optim/lr": 0.0018372000000000002, "optim/total_tokens": 802947072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 5.993467807769775, "created_at": "2025-01-14T17:19:33.745756+00:00"} {"global_step": 3064, "acc_step": 0, "speed/wps": 12905.59009242751, "speed/FLOPS": 202700026540766.03, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6681979894638062, "optim/lr": 0.0018378000000000001, "optim/total_tokens": 803209216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309558, "loss/out": 5.834477424621582, "created_at": "2025-01-14T17:19:38.827690+00:00"} {"global_step": 3065, "acc_step": 0, "speed/wps": 12896.527747253302, "speed/FLOPS": 202557689956837.8, "speed/curr_iter_time": 1.2819, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.266681432723999, "optim/lr": 0.0018384, "optim/total_tokens": 803471360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 5.977397918701172, "created_at": "2025-01-14T17:19:43.913198+00:00"} {"global_step": 3066, "acc_step": 0, "speed/wps": 12902.0114017126, "speed/FLOPS": 202643818285451.97, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.160325050354004, "optim/lr": 0.001839, "optim/total_tokens": 803733504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464820, "loss/out": 5.778197288513184, "created_at": "2025-01-14T17:19:48.996410+00:00"} {"global_step": 3067, "acc_step": 0, "speed/wps": 12902.18157330212, "speed/FLOPS": 202646491064105.66, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3268407583236694, "optim/lr": 0.0018396, "optim/total_tokens": 803995648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461077, "loss/out": 5.898347854614258, "created_at": "2025-01-14T17:19:54.079915+00:00"} {"global_step": 3068, "acc_step": 0, "speed/wps": 12889.987831587438, "speed/FLOPS": 202454971594519.56, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2106719017028809, "optim/lr": 0.0018402, "optim/total_tokens": 804257792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474863, "loss/out": 5.72377872467041, "created_at": "2025-01-14T17:19:59.167713+00:00"} {"global_step": 3069, "acc_step": 0, "speed/wps": 12902.203282709124, "speed/FLOPS": 202646832040174.5, "speed/curr_iter_time": 1.282, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5823040008544922, "optim/lr": 0.0018408, "optim/total_tokens": 804519936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420425, "loss/out": 5.708720684051514, "created_at": "2025-01-14T17:20:04.254523+00:00"} {"global_step": 3070, "acc_step": 0, "speed/wps": 12905.432332447874, "speed/FLOPS": 202697548703500.9, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3322120904922485, "optim/lr": 0.0018414, "optim/total_tokens": 804782080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325239, "loss/out": 5.6741461753845215, "created_at": "2025-01-14T17:20:09.337674+00:00"} {"global_step": 3071, "acc_step": 0, "speed/wps": 12898.779358305723, "speed/FLOPS": 202593054602453.9, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2069429159164429, "optim/lr": 0.001842, "optim/total_tokens": 805044224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 5.66584587097168, "created_at": "2025-01-14T17:20:14.422555+00:00"} {"global_step": 3072, "acc_step": 0, "speed/wps": 12900.802533422131, "speed/FLOPS": 202624831347791.06, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.385695219039917, "optim/lr": 0.0018426, "optim/total_tokens": 805306368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 5.780911922454834, "created_at": "2025-01-14T17:20:19.506635+00:00"} {"global_step": 3073, "acc_step": 0, "speed/wps": 12902.251026155862, "speed/FLOPS": 202647581916611.9, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7385809421539307, "optim/lr": 0.0018432, "optim/total_tokens": 805568512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362390, "loss/out": 5.667013168334961, "created_at": "2025-01-14T17:20:24.589927+00:00"} {"global_step": 3074, "acc_step": 0, "speed/wps": 12912.00595144113, "speed/FLOPS": 202800796422889.06, "speed/curr_iter_time": 1.2786, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.025550127029419, "optim/lr": 0.0018438, "optim/total_tokens": 805830656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 5.625520706176758, "created_at": "2025-01-14T17:20:29.669554+00:00"} {"global_step": 3075, "acc_step": 0, "speed/wps": 12906.140121587034, "speed/FLOPS": 202708665504535.6, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.001772165298462, "optim/lr": 0.0018444000000000002, "optim/total_tokens": 806092800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473815, "loss/out": 5.918671607971191, "created_at": "2025-01-14T17:20:34.757265+00:00"} {"global_step": 3076, "acc_step": 0, "speed/wps": 12907.456179503228, "speed/FLOPS": 202729336002562.53, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4300591945648193, "optim/lr": 0.001845, "optim/total_tokens": 806354944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439054, "loss/out": 5.790043354034424, "created_at": "2025-01-14T17:20:39.842282+00:00"} {"global_step": 3077, "acc_step": 0, "speed/wps": 12901.185720678275, "speed/FLOPS": 202630849830203.28, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.2991284132003784, "optim/lr": 0.0018456, "optim/total_tokens": 806617088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441450, "loss/out": 5.749507427215576, "created_at": "2025-01-14T17:20:44.927925+00:00"} {"global_step": 3078, "acc_step": 0, "speed/wps": 12906.57837530787, "speed/FLOPS": 202715548881445.06, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.001, "optim/grad_norm": 1.7305241823196411, "optim/lr": 0.0018461999999999999, "optim/total_tokens": 806879232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457889, "loss/out": 5.6836838722229, "created_at": "2025-01-14T17:20:50.009371+00:00"} {"global_step": 3079, "acc_step": 0, "speed/wps": 12901.593616097616, "speed/FLOPS": 202637256388271.06, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3553037643432617, "optim/lr": 0.0018468000000000002, "optim/total_tokens": 807141376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318166, "loss/out": 5.721927165985107, "created_at": "2025-01-14T17:20:55.096923+00:00"} {"global_step": 3080, "acc_step": 0, "speed/wps": 12900.353989329327, "speed/FLOPS": 202617786346449.4, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.691884756088257, "optim/lr": 0.0018474000000000001, "optim/total_tokens": 807403520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 5.633557319641113, "created_at": "2025-01-14T17:21:00.181767+00:00"} {"global_step": 3081, "acc_step": 0, "speed/wps": 12913.53876603001, "speed/FLOPS": 202824871382315.4, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6069947481155396, "optim/lr": 0.001848, "optim/total_tokens": 807665664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 5.574007987976074, "created_at": "2025-01-14T17:21:05.266080+00:00"} {"global_step": 3082, "acc_step": 0, "speed/wps": 12908.622851825343, "speed/FLOPS": 202747660194557.1, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7307162284851074, "optim/lr": 0.0018486, "optim/total_tokens": 807927808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379737, "loss/out": 5.706705093383789, "created_at": "2025-01-14T17:21:10.348076+00:00"} {"global_step": 3083, "acc_step": 0, "speed/wps": 12906.815977278164, "speed/FLOPS": 202719280746887.84, "speed/curr_iter_time": 1.2788, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.5123393535614014, "optim/lr": 0.0018491999999999998, "optim/total_tokens": 808189952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 5.787197589874268, "created_at": "2025-01-14T17:21:15.429357+00:00"} {"global_step": 3084, "acc_step": 0, "speed/wps": 12904.969207273156, "speed/FLOPS": 202690274686231.28, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.3562965393066406, "optim/lr": 0.0018498000000000002, "optim/total_tokens": 808452096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462963, "loss/out": 5.767632961273193, "created_at": "2025-01-14T17:21:20.514106+00:00"} {"global_step": 3085, "acc_step": 0, "speed/wps": 12903.681004659767, "speed/FLOPS": 202670041694012.12, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9982749223709106, "optim/lr": 0.0018504, "optim/total_tokens": 808714240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437710, "loss/out": 5.754398345947266, "created_at": "2025-01-14T17:21:25.600583+00:00"} {"global_step": 3086, "acc_step": 0, "speed/wps": 12904.540940561337, "speed/FLOPS": 202683548169022.03, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1001009941101074, "optim/lr": 0.001851, "optim/total_tokens": 808976384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427880, "loss/out": 5.752654075622559, "created_at": "2025-01-14T17:21:30.683135+00:00"} {"global_step": 3087, "acc_step": 0, "speed/wps": 12896.487649659975, "speed/FLOPS": 202557060169034.78, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4777770042419434, "optim/lr": 0.0018516, "optim/total_tokens": 809238528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328460, "loss/out": 5.777958393096924, "created_at": "2025-01-14T17:21:35.768496+00:00"} {"global_step": 3088, "acc_step": 0, "speed/wps": 12903.816472146029, "speed/FLOPS": 202672169397035.7, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.8886853456497192, "optim/lr": 0.0018521999999999998, "optim/total_tokens": 809500672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 5.781364440917969, "created_at": "2025-01-14T17:21:40.851775+00:00"} {"global_step": 3089, "acc_step": 0, "speed/wps": 12905.600253444176, "speed/FLOPS": 202700186133495.44, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5749081373214722, "optim/lr": 0.0018528000000000001, "optim/total_tokens": 809762816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 5.797494888305664, "created_at": "2025-01-14T17:21:45.942616+00:00"} {"global_step": 3090, "acc_step": 0, "speed/wps": 12890.933793063343, "speed/FLOPS": 202469829219385.94, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9613929986953735, "optim/lr": 0.0018534, "optim/total_tokens": 810024960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411072, "loss/out": 5.803464889526367, "created_at": "2025-01-14T17:21:51.030034+00:00"} {"global_step": 3091, "acc_step": 0, "speed/wps": 12904.923325181211, "speed/FLOPS": 202689554044928.7, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2032508850097656, "optim/lr": 0.001854, "optim/total_tokens": 810287104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 5.692924499511719, "created_at": "2025-01-14T17:21:56.112707+00:00"} {"global_step": 3092, "acc_step": 0, "speed/wps": 12904.033681317982, "speed/FLOPS": 202675580965557.88, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3698831796646118, "optim/lr": 0.0018545999999999999, "optim/total_tokens": 810549248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448077, "loss/out": 6.0042829513549805, "created_at": "2025-01-14T17:22:01.194941+00:00"} {"global_step": 3093, "acc_step": 0, "speed/wps": 12900.288224909142, "speed/FLOPS": 202616753425860.6, "speed/curr_iter_time": 1.2805, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6379947662353516, "optim/lr": 0.0018552, "optim/total_tokens": 810811392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331148, "loss/out": 5.701817989349365, "created_at": "2025-01-14T17:22:06.282535+00:00"} {"global_step": 3094, "acc_step": 0, "speed/wps": 12903.012932292095, "speed/FLOPS": 202659548699450.2, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1104509830474854, "optim/lr": 0.0018558000000000001, "optim/total_tokens": 811073536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 5.870242118835449, "created_at": "2025-01-14T17:22:11.365662+00:00"} {"global_step": 3095, "acc_step": 0, "speed/wps": 12910.38969059102, "speed/FLOPS": 202775410825262.97, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2075138092041016, "optim/lr": 0.0018564, "optim/total_tokens": 811335680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455808, "loss/out": 5.787432670593262, "created_at": "2025-01-14T17:22:16.447310+00:00"} {"global_step": 3096, "acc_step": 0, "speed/wps": 12902.908323407131, "speed/FLOPS": 202657905673164.16, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2466334104537964, "optim/lr": 0.001857, "optim/total_tokens": 811597824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301350, "loss/out": 5.853689670562744, "created_at": "2025-01-14T17:22:21.532964+00:00"} {"global_step": 3097, "acc_step": 0, "speed/wps": 12905.65361975457, "speed/FLOPS": 202701024324733.2, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2976852655410767, "optim/lr": 0.0018576, "optim/total_tokens": 811859968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 5.783078670501709, "created_at": "2025-01-14T17:22:26.617627+00:00"} {"global_step": 3098, "acc_step": 0, "speed/wps": 12907.793473689284, "speed/FLOPS": 202734633671245.3, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2156720161437988, "optim/lr": 0.0018582, "optim/total_tokens": 812122112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 5.830655574798584, "created_at": "2025-01-14T17:22:31.698871+00:00"} {"global_step": 3099, "acc_step": 0, "speed/wps": 12905.415400236496, "speed/FLOPS": 202697282759853.9, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.658281922340393, "optim/lr": 0.0018588, "optim/total_tokens": 812384256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360123, "loss/out": 5.672766208648682, "created_at": "2025-01-14T17:22:36.785584+00:00"} {"global_step": 3100, "acc_step": 0, "speed/wps": 12910.936713079822, "speed/FLOPS": 202784002565137.25, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3992165327072144, "optim/lr": 0.0018594, "optim/total_tokens": 812646400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461889, "loss/out": 5.6703667640686035, "created_at": "2025-01-14T17:22:41.865680+00:00"} {"global_step": 3101, "acc_step": 0, "speed/wps": 12905.298506706175, "speed/FLOPS": 202695446786331.66, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.655245065689087, "optim/lr": 0.00186, "optim/total_tokens": 812908544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463248, "loss/out": 5.841541290283203, "created_at": "2025-01-14T17:22:46.948732+00:00"} {"global_step": 3102, "acc_step": 0, "speed/wps": 12910.468173557512, "speed/FLOPS": 202776643508098.62, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3404736518859863, "optim/lr": 0.0018606, "optim/total_tokens": 813170688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435567, "loss/out": 5.863924026489258, "created_at": "2025-01-14T17:22:52.032604+00:00"} {"global_step": 3103, "acc_step": 0, "speed/wps": 12905.814341477546, "speed/FLOPS": 202703548680247.3, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3096296787261963, "optim/lr": 0.0018612, "optim/total_tokens": 813432832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437432, "loss/out": 5.770165920257568, "created_at": "2025-01-14T17:22:57.117130+00:00"} {"global_step": 3104, "acc_step": 0, "speed/wps": 12915.35034671406, "speed/FLOPS": 202853324746331.16, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0807106494903564, "optim/lr": 0.0018618000000000003, "optim/total_tokens": 813694976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443545, "loss/out": 5.681186676025391, "created_at": "2025-01-14T17:23:02.198043+00:00"} {"global_step": 3105, "acc_step": 0, "speed/wps": 12910.208470922531, "speed/FLOPS": 202772564521347.4, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0097243785858154, "optim/lr": 0.0018624000000000002, "optim/total_tokens": 813957120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364548, "loss/out": 5.733222007751465, "created_at": "2025-01-14T17:23:07.279910+00:00"} {"global_step": 3106, "acc_step": 0, "speed/wps": 12909.176985476035, "speed/FLOPS": 202756363625000.66, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2779655456542969, "optim/lr": 0.001863, "optim/total_tokens": 814219264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 5.936955451965332, "created_at": "2025-01-14T17:23:12.365297+00:00"} {"global_step": 3107, "acc_step": 0, "speed/wps": 12907.859845021218, "speed/FLOPS": 202735676124217.1, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.001, "optim/grad_norm": 1.3043290376663208, "optim/lr": 0.0018636, "optim/total_tokens": 814481408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 5.869905471801758, "created_at": "2025-01-14T17:23:17.447254+00:00"} {"global_step": 3108, "acc_step": 0, "speed/wps": 12908.819999548357, "speed/FLOPS": 202750756670456.4, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1987546682357788, "optim/lr": 0.0018641999999999999, "optim/total_tokens": 814743552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307513, "loss/out": 5.859486103057861, "created_at": "2025-01-14T17:23:22.532177+00:00"} {"global_step": 3109, "acc_step": 0, "speed/wps": 12911.570534407469, "speed/FLOPS": 202793957600049.72, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1591140031814575, "optim/lr": 0.0018648000000000002, "optim/total_tokens": 815005696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 5.734842300415039, "created_at": "2025-01-14T17:23:27.612411+00:00"} {"global_step": 3110, "acc_step": 0, "speed/wps": 12917.515863961633, "speed/FLOPS": 202887337170436.8, "speed/curr_iter_time": 1.2785, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1780762672424316, "optim/lr": 0.0018654000000000001, "optim/total_tokens": 815267840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 5.839147567749023, "created_at": "2025-01-14T17:23:32.691646+00:00"} {"global_step": 3111, "acc_step": 0, "speed/wps": 12912.730031786941, "speed/FLOPS": 202812169099710.72, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2928829193115234, "optim/lr": 0.001866, "optim/total_tokens": 815529984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 339173, "loss/out": 5.698648929595947, "created_at": "2025-01-14T17:23:37.770696+00:00"} {"global_step": 3112, "acc_step": 0, "speed/wps": 12908.63061884008, "speed/FLOPS": 202747782186196.53, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0208789110183716, "optim/lr": 0.0018666, "optim/total_tokens": 815792128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 5.808970928192139, "created_at": "2025-01-14T17:23:42.855485+00:00"} {"global_step": 3113, "acc_step": 0, "speed/wps": 12918.404562285768, "speed/FLOPS": 202901295398816.84, "speed/curr_iter_time": 1.2788, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1194318532943726, "optim/lr": 0.0018671999999999998, "optim/total_tokens": 816054272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 5.66521692276001, "created_at": "2025-01-14T17:23:47.933619+00:00"} {"global_step": 3114, "acc_step": 0, "speed/wps": 12890.148824495165, "speed/FLOPS": 202457500209362.84, "speed/curr_iter_time": 1.2813, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3587590456008911, "optim/lr": 0.0018678000000000002, "optim/total_tokens": 816316416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364786, "loss/out": 5.758714199066162, "created_at": "2025-01-14T17:23:53.021539+00:00"} {"global_step": 3115, "acc_step": 0, "speed/wps": 12901.40187859754, "speed/FLOPS": 202634244887357.44, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.1254327297210693, "optim/lr": 0.0018684, "optim/total_tokens": 816578560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 5.721127510070801, "created_at": "2025-01-14T17:23:58.107891+00:00"} {"global_step": 3116, "acc_step": 0, "speed/wps": 12915.578839370102, "speed/FLOPS": 202856913537476.1, "speed/curr_iter_time": 1.2788, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1800824403762817, "optim/lr": 0.001869, "optim/total_tokens": 816840704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462938, "loss/out": 5.802474498748779, "created_at": "2025-01-14T17:24:03.191264+00:00"} {"global_step": 3117, "acc_step": 0, "speed/wps": 12909.750554434973, "speed/FLOPS": 202765372313667.12, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.450637936592102, "optim/lr": 0.0018696, "optim/total_tokens": 817102848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374860, "loss/out": 5.623378753662109, "created_at": "2025-01-14T17:24:08.276462+00:00"} {"global_step": 3118, "acc_step": 0, "speed/wps": 12904.202178808107, "speed/FLOPS": 202678227450180.12, "speed/curr_iter_time": 1.2817, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4149607419967651, "optim/lr": 0.0018701999999999998, "optim/total_tokens": 817364992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 5.75334358215332, "created_at": "2025-01-14T17:24:13.364359+00:00"} {"global_step": 3119, "acc_step": 0, "speed/wps": 12909.449333552451, "speed/FLOPS": 202760641225786.25, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2167444229125977, "optim/lr": 0.0018708000000000002, "optim/total_tokens": 817627136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465083, "loss/out": 5.78800106048584, "created_at": "2025-01-14T17:24:18.446701+00:00"} {"global_step": 3120, "acc_step": 0, "speed/wps": 12907.677552794608, "speed/FLOPS": 202732812974302.44, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.173654794692993, "optim/lr": 0.0018714, "optim/total_tokens": 817889280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297001, "loss/out": 5.6429901123046875, "created_at": "2025-01-14T17:24:23.527832+00:00"} {"global_step": 3121, "acc_step": 0, "speed/wps": 12916.286314709314, "speed/FLOPS": 202868025409851.4, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4436595439910889, "optim/lr": 0.001872, "optim/total_tokens": 818151424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 5.754889488220215, "created_at": "2025-01-14T17:24:28.605546+00:00"} {"global_step": 3122, "acc_step": 0, "speed/wps": 12912.741504906333, "speed/FLOPS": 202812349300816.47, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6838020086288452, "optim/lr": 0.0018725999999999999, "optim/total_tokens": 818413568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453131, "loss/out": 5.7695207595825195, "created_at": "2025-01-14T17:24:33.686092+00:00"} {"global_step": 3123, "acc_step": 0, "speed/wps": 12912.411922673138, "speed/FLOPS": 202807172758950.72, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8155863285064697, "optim/lr": 0.0018732, "optim/total_tokens": 818675712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324181, "loss/out": 5.711758136749268, "created_at": "2025-01-14T17:24:38.766530+00:00"} {"global_step": 3124, "acc_step": 0, "speed/wps": 12907.088334642784, "speed/FLOPS": 202723558493557.44, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2333120107650757, "optim/lr": 0.0018738000000000001, "optim/total_tokens": 818937856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 5.602194786071777, "created_at": "2025-01-14T17:24:43.849626+00:00"} {"global_step": 3125, "acc_step": 0, "speed/wps": 12909.776216150127, "speed/FLOPS": 202765775366166.5, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.3086342811584473, "optim/lr": 0.0018744, "optim/total_tokens": 819200000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451797, "loss/out": 5.755989074707031, "created_at": "2025-01-14T17:24:48.932621+00:00"} {"global_step": 3126, "acc_step": 0, "speed/wps": 12905.586908976942, "speed/FLOPS": 202699976540300.3, "speed/curr_iter_time": 1.2802, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7248644828796387, "optim/lr": 0.001875, "optim/total_tokens": 819462144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310371, "loss/out": 5.735156059265137, "created_at": "2025-01-14T17:24:54.018487+00:00"} {"global_step": 3127, "acc_step": 0, "speed/wps": 12907.32197559805, "speed/FLOPS": 202727228145816.22, "speed/curr_iter_time": 1.2808, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7070584297180176, "optim/lr": 0.0018756, "optim/total_tokens": 819724288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 5.764949321746826, "created_at": "2025-01-14T17:24:59.107466+00:00"} {"global_step": 3128, "acc_step": 0, "speed/wps": 12904.166692832103, "speed/FLOPS": 202677670094164.12, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6385048627853394, "optim/lr": 0.0018762, "optim/total_tokens": 819986432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453646, "loss/out": 5.682889938354492, "created_at": "2025-01-14T17:25:04.189585+00:00"} {"global_step": 3129, "acc_step": 0, "speed/wps": 12907.646200527997, "speed/FLOPS": 202732320543872.78, "speed/curr_iter_time": 1.2785, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.4925974607467651, "optim/lr": 0.0018768, "optim/total_tokens": 820248576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413426, "loss/out": 5.756728649139404, "created_at": "2025-01-14T17:25:09.270561+00:00"} {"global_step": 3130, "acc_step": 0, "speed/wps": 12904.427295444013, "speed/FLOPS": 202681763216289.7, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.590579867362976, "optim/lr": 0.0018774000000000002, "optim/total_tokens": 820510720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465083, "loss/out": 5.703034400939941, "created_at": "2025-01-14T17:25:14.353930+00:00"} {"global_step": 3131, "acc_step": 0, "speed/wps": 12914.488668609363, "speed/FLOPS": 202839790907627.1, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2822171449661255, "optim/lr": 0.0018780000000000001, "optim/total_tokens": 820772864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 5.555891990661621, "created_at": "2025-01-14T17:25:19.433123+00:00"} {"global_step": 3132, "acc_step": 0, "speed/wps": 12909.782149030034, "speed/FLOPS": 202765868550198.1, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2554001808166504, "optim/lr": 0.0018786, "optim/total_tokens": 821035008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 284247, "loss/out": 5.8251848220825195, "created_at": "2025-01-14T17:25:24.520650+00:00"} {"global_step": 3133, "acc_step": 0, "speed/wps": 12909.337766549605, "speed/FLOPS": 202758888912697.4, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6114894151687622, "optim/lr": 0.0018792, "optim/total_tokens": 821297152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 5.686062812805176, "created_at": "2025-01-14T17:25:29.601036+00:00"} {"global_step": 3134, "acc_step": 0, "speed/wps": 12908.574220580003, "speed/FLOPS": 202746896374022.72, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3283381462097168, "optim/lr": 0.0018798000000000003, "optim/total_tokens": 821559296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441157, "loss/out": 5.83712100982666, "created_at": "2025-01-14T17:25:34.687170+00:00"} {"global_step": 3135, "acc_step": 0, "speed/wps": 12907.873805060048, "speed/FLOPS": 202735895385810.03, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3421149253845215, "optim/lr": 0.0018804000000000002, "optim/total_tokens": 821821440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311191, "loss/out": 5.765957832336426, "created_at": "2025-01-14T17:25:39.773295+00:00"} {"global_step": 3136, "acc_step": 0, "speed/wps": 12911.444742623693, "speed/FLOPS": 202791981867229.5, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3156887292861938, "optim/lr": 0.001881, "optim/total_tokens": 822083584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 5.692075729370117, "created_at": "2025-01-14T17:25:44.858481+00:00"} {"global_step": 3137, "acc_step": 0, "speed/wps": 12908.815718091506, "speed/FLOPS": 202750689424293.22, "speed/curr_iter_time": 1.28, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.036489248275757, "optim/lr": 0.0018816, "optim/total_tokens": 822345728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 5.782217979431152, "created_at": "2025-01-14T17:25:49.942533+00:00"} {"global_step": 3138, "acc_step": 0, "speed/wps": 12911.768359726719, "speed/FLOPS": 202797064718529.03, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4566007852554321, "optim/lr": 0.0018821999999999999, "optim/total_tokens": 822607872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 396624, "loss/out": 5.6656599044799805, "created_at": "2025-01-14T17:25:55.031491+00:00"} {"global_step": 3139, "acc_step": 0, "speed/wps": 12913.718517467008, "speed/FLOPS": 202827694625637.84, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3399986028671265, "optim/lr": 0.0018828000000000002, "optim/total_tokens": 822870016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431855, "loss/out": 5.686077117919922, "created_at": "2025-01-14T17:26:00.111758+00:00"} {"global_step": 3140, "acc_step": 0, "speed/wps": 12911.257300877334, "speed/FLOPS": 202789037837031.0, "speed/curr_iter_time": 1.2789, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2164623737335205, "optim/lr": 0.0018834000000000001, "optim/total_tokens": 823132160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428118, "loss/out": 5.6281843185424805, "created_at": "2025-01-14T17:26:05.192149+00:00"} {"global_step": 3141, "acc_step": 0, "speed/wps": 12917.50877391394, "speed/FLOPS": 202887225811495.06, "speed/curr_iter_time": 1.279, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.208614706993103, "optim/lr": 0.001884, "optim/total_tokens": 823394304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465083, "loss/out": 5.627344131469727, "created_at": "2025-01-14T17:26:10.273065+00:00"} {"global_step": 3142, "acc_step": 0, "speed/wps": 12912.32482954531, "speed/FLOPS": 202805804841699.66, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3725250959396362, "optim/lr": 0.0018846, "optim/total_tokens": 823656448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450227, "loss/out": 5.633447170257568, "created_at": "2025-01-14T17:26:15.353344+00:00"} {"global_step": 3143, "acc_step": 0, "speed/wps": 12907.29986779369, "speed/FLOPS": 202726880912369.44, "speed/curr_iter_time": 1.281, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2420525550842285, "optim/lr": 0.0018851999999999999, "optim/total_tokens": 823918592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440390, "loss/out": 5.604262351989746, "created_at": "2025-01-14T17:26:20.439505+00:00"} {"global_step": 3144, "acc_step": 0, "speed/wps": 12909.18358563659, "speed/FLOPS": 202756467289591.75, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3010437488555908, "optim/lr": 0.0018858000000000002, "optim/total_tokens": 824180736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444067, "loss/out": 5.691556930541992, "created_at": "2025-01-14T17:26:25.519811+00:00"} {"global_step": 3145, "acc_step": 0, "speed/wps": 12911.453889440701, "speed/FLOPS": 202792125530559.66, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2204440832138062, "optim/lr": 0.0018864, "optim/total_tokens": 824442880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 5.740575790405273, "created_at": "2025-01-14T17:26:30.604136+00:00"} {"global_step": 3146, "acc_step": 0, "speed/wps": 12900.783808773818, "speed/FLOPS": 202624537251458.94, "speed/curr_iter_time": 1.2814, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.327665090560913, "optim/lr": 0.001887, "optim/total_tokens": 824705024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 5.652858734130859, "created_at": "2025-01-14T17:26:35.690809+00:00"} {"global_step": 3147, "acc_step": 0, "speed/wps": 12917.541120244934, "speed/FLOPS": 202887733855071.8, "speed/curr_iter_time": 1.2782, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1581846475601196, "optim/lr": 0.0018876, "optim/total_tokens": 824967168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 342920, "loss/out": 5.54427433013916, "created_at": "2025-01-14T17:26:40.768959+00:00"} {"global_step": 3148, "acc_step": 0, "speed/wps": 12911.699929703667, "speed/FLOPS": 202795989930984.3, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.199580430984497, "optim/lr": 0.0018881999999999998, "optim/total_tokens": 825229312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 5.517121315002441, "created_at": "2025-01-14T17:26:45.849160+00:00"} {"global_step": 3149, "acc_step": 0, "speed/wps": 12905.366028680068, "speed/FLOPS": 202696507311716.6, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3339684009552002, "optim/lr": 0.0018888000000000002, "optim/total_tokens": 825491456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 5.523194313049316, "created_at": "2025-01-14T17:26:50.932351+00:00"} {"global_step": 3150, "acc_step": 0, "speed/wps": 12908.704318329468, "speed/FLOPS": 202748939737951.1, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5513031482696533, "optim/lr": 0.0018894, "optim/total_tokens": 825753600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369627, "loss/out": 5.641201972961426, "created_at": "2025-01-14T17:26:56.013370+00:00"} {"global_step": 3151, "acc_step": 0, "speed/wps": 12910.197795774433, "speed/FLOPS": 202772396853477.34, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1946160793304443, "optim/lr": 0.00189, "optim/total_tokens": 826015744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 5.609348773956299, "created_at": "2025-01-14T17:27:01.096507+00:00"} {"global_step": 3152, "acc_step": 0, "speed/wps": 12913.365789915804, "speed/FLOPS": 202822154554747.72, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.247411847114563, "optim/lr": 0.0018906, "optim/total_tokens": 826277888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460271, "loss/out": 5.6464056968688965, "created_at": "2025-01-14T17:27:06.175240+00:00"} {"global_step": 3153, "acc_step": 0, "speed/wps": 12911.245043902805, "speed/FLOPS": 202788845324403.16, "speed/curr_iter_time": 1.2785, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1725506782531738, "optim/lr": 0.0018912, "optim/total_tokens": 826540032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365072, "loss/out": 5.613376617431641, "created_at": "2025-01-14T17:27:11.256554+00:00"} {"global_step": 3154, "acc_step": 0, "speed/wps": 12906.945121047598, "speed/FLOPS": 202721309127248.94, "speed/curr_iter_time": 1.2781, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1165213584899902, "optim/lr": 0.0018918000000000001, "optim/total_tokens": 826802176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453375, "loss/out": 5.586912155151367, "created_at": "2025-01-14T17:27:16.339706+00:00"} {"global_step": 3155, "acc_step": 0, "speed/wps": 12920.771546771992, "speed/FLOPS": 202938472142750.5, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.240181565284729, "optim/lr": 0.0018924, "optim/total_tokens": 827064320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454458, "loss/out": 5.5939178466796875, "created_at": "2025-01-14T17:27:21.427772+00:00"} {"global_step": 3156, "acc_step": 0, "speed/wps": 12912.428541534477, "speed/FLOPS": 202807433781004.72, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2266961336135864, "optim/lr": 0.0018930000000000002, "optim/total_tokens": 827326464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 5.786433219909668, "created_at": "2025-01-14T17:27:26.507796+00:00"} {"global_step": 3157, "acc_step": 0, "speed/wps": 12903.22482164439, "speed/FLOPS": 202662876712894.44, "speed/curr_iter_time": 1.2788, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.119911551475525, "optim/lr": 0.0018936, "optim/total_tokens": 827588608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438773, "loss/out": 5.674159049987793, "created_at": "2025-01-14T17:27:31.591152+00:00"} {"global_step": 3158, "acc_step": 0, "speed/wps": 12918.017849635598, "speed/FLOPS": 202895221545247.03, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3021762371063232, "optim/lr": 0.0018942, "optim/total_tokens": 827850752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 5.659463882446289, "created_at": "2025-01-14T17:27:36.669464+00:00"} {"global_step": 3159, "acc_step": 0, "speed/wps": 12916.447346419875, "speed/FLOPS": 202870554634146.53, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9842883944511414, "optim/lr": 0.0018948, "optim/total_tokens": 828112896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315105, "loss/out": 5.650949001312256, "created_at": "2025-01-14T17:27:41.748116+00:00"} {"global_step": 3160, "acc_step": 0, "speed/wps": 12911.878054443758, "speed/FLOPS": 202798787624796.47, "speed/curr_iter_time": 1.279, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2149629592895508, "optim/lr": 0.0018954000000000002, "optim/total_tokens": 828375040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 5.602099895477295, "created_at": "2025-01-14T17:27:46.828533+00:00"} {"global_step": 3161, "acc_step": 0, "speed/wps": 12913.492083611163, "speed/FLOPS": 202824138170779.06, "speed/curr_iter_time": 1.2787, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1370398998260498, "optim/lr": 0.0018960000000000001, "optim/total_tokens": 828637184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446218, "loss/out": 5.659566879272461, "created_at": "2025-01-14T17:27:51.910849+00:00"} {"global_step": 3162, "acc_step": 0, "speed/wps": 12916.048209838315, "speed/FLOPS": 202864285645662.0, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1670103073120117, "optim/lr": 0.0018966, "optim/total_tokens": 828899328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319377, "loss/out": 5.588042259216309, "created_at": "2025-01-14T17:27:56.991591+00:00"} {"global_step": 3163, "acc_step": 0, "speed/wps": 12914.633642430581, "speed/FLOPS": 202842067920704.22, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1517250537872314, "optim/lr": 0.0018972, "optim/total_tokens": 829161472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 5.435736656188965, "created_at": "2025-01-14T17:28:02.069680+00:00"} {"global_step": 3164, "acc_step": 0, "speed/wps": 12911.914246219978, "speed/FLOPS": 202799356066374.72, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7872878313064575, "optim/lr": 0.0018978000000000003, "optim/total_tokens": 829423616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469875, "loss/out": 5.633922576904297, "created_at": "2025-01-14T17:28:07.149882+00:00"} {"global_step": 3165, "acc_step": 0, "speed/wps": 12918.245542622697, "speed/FLOPS": 202898797776492.38, "speed/curr_iter_time": 1.2786, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1655213832855225, "optim/lr": 0.0018984000000000002, "optim/total_tokens": 829685760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334369, "loss/out": 5.557546615600586, "created_at": "2025-01-14T17:28:12.226669+00:00"} {"global_step": 3166, "acc_step": 0, "speed/wps": 12909.577788734654, "speed/FLOPS": 202762658790938.66, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.221311330795288, "optim/lr": 0.001899, "optim/total_tokens": 829947904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 5.531800746917725, "created_at": "2025-01-14T17:28:17.315316+00:00"} {"global_step": 3167, "acc_step": 0, "speed/wps": 12917.800885581592, "speed/FLOPS": 202891813826639.0, "speed/curr_iter_time": 1.2784, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.197287678718567, "optim/lr": 0.0018996, "optim/total_tokens": 830210048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 5.6379899978637695, "created_at": "2025-01-14T17:28:22.393382+00:00"} {"global_step": 3168, "acc_step": 0, "speed/wps": 12913.978055614945, "speed/FLOPS": 202831771028893.97, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0307577848434448, "optim/lr": 0.0019002, "optim/total_tokens": 830472192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299938, "loss/out": 5.527363300323486, "created_at": "2025-01-14T17:28:27.472187+00:00"} {"global_step": 3169, "acc_step": 0, "speed/wps": 12919.37546564678, "speed/FLOPS": 202916544770262.56, "speed/curr_iter_time": 1.2778, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9987980127334595, "optim/lr": 0.0019008000000000002, "optim/total_tokens": 830734336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 5.542022228240967, "created_at": "2025-01-14T17:28:32.548586+00:00"} {"global_step": 3170, "acc_step": 0, "speed/wps": 12911.665750500752, "speed/FLOPS": 202795453099635.34, "speed/curr_iter_time": 1.2788, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0995022058486938, "optim/lr": 0.0019014000000000001, "optim/total_tokens": 830996480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435053, "loss/out": 5.5290937423706055, "created_at": "2025-01-14T17:28:37.627886+00:00"} {"global_step": 3171, "acc_step": 0, "speed/wps": 12911.721057549876, "speed/FLOPS": 202796321772841.72, "speed/curr_iter_time": 1.2806, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7627828121185303, "optim/lr": 0.001902, "optim/total_tokens": 831258624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444641, "loss/out": 5.577866077423096, "created_at": "2025-01-14T17:28:42.710866+00:00"} {"global_step": 3172, "acc_step": 0, "speed/wps": 12921.592855807367, "speed/FLOPS": 202951371929747.94, "speed/curr_iter_time": 1.2789, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8566667437553406, "optim/lr": 0.0019026, "optim/total_tokens": 831520768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303511, "loss/out": 5.5300822257995605, "created_at": "2025-01-14T17:28:47.789526+00:00"} {"global_step": 3173, "acc_step": 0, "speed/wps": 12916.12104812468, "speed/FLOPS": 202865429671039.66, "speed/curr_iter_time": 1.2789, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0853067636489868, "optim/lr": 0.0019031999999999999, "optim/total_tokens": 831782912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 5.5778093338012695, "created_at": "2025-01-14T17:28:52.881933+00:00"} {"global_step": 3174, "acc_step": 0, "speed/wps": 12915.91982973572, "speed/FLOPS": 202862269259738.0, "speed/curr_iter_time": 1.2789, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.011407494544983, "optim/lr": 0.0019038000000000002, "optim/total_tokens": 832045056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465633, "loss/out": 5.5013275146484375, "created_at": "2025-01-14T17:28:57.959590+00:00"} {"global_step": 3175, "acc_step": 0, "speed/wps": 12925.33922356678, "speed/FLOPS": 203010213783458.3, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.037782073020935, "optim/lr": 0.0019044, "optim/total_tokens": 832307200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376816, "loss/out": 5.637210845947266, "created_at": "2025-01-14T17:29:03.035087+00:00"} {"global_step": 3176, "acc_step": 0, "speed/wps": 12922.46505043257, "speed/FLOPS": 202965070944857.75, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6191263794898987, "optim/lr": 0.001905, "optim/total_tokens": 832569344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 5.573636531829834, "created_at": "2025-01-14T17:29:08.110384+00:00"} {"global_step": 3177, "acc_step": 0, "speed/wps": 12916.495553345649, "speed/FLOPS": 202871311790159.06, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8688368797302246, "optim/lr": 0.0019056, "optim/total_tokens": 832831488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 5.683143615722656, "created_at": "2025-01-14T17:29:13.190279+00:00"} {"global_step": 3178, "acc_step": 0, "speed/wps": 12918.266231819087, "speed/FLOPS": 202899122728751.62, "speed/curr_iter_time": 1.2786, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6970898509025574, "optim/lr": 0.0019062, "optim/total_tokens": 833093632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361126, "loss/out": 5.500273704528809, "created_at": "2025-01-14T17:29:18.271650+00:00"} {"global_step": 3179, "acc_step": 0, "speed/wps": 12919.016893205211, "speed/FLOPS": 202910912897337.53, "speed/curr_iter_time": 1.2787, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8142473697662354, "optim/lr": 0.0019068000000000002, "optim/total_tokens": 833355776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 5.529608249664307, "created_at": "2025-01-14T17:29:23.352606+00:00"} {"global_step": 3180, "acc_step": 0, "speed/wps": 12923.955535330866, "speed/FLOPS": 202988481058326.4, "speed/curr_iter_time": 1.2774, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7566826343536377, "optim/lr": 0.0019074, "optim/total_tokens": 833617920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 5.585428237915039, "created_at": "2025-01-14T17:29:28.429077+00:00"} {"global_step": 3181, "acc_step": 0, "speed/wps": 12921.67496065838, "speed/FLOPS": 202952661499257.0, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8622145056724548, "optim/lr": 0.001908, "optim/total_tokens": 833880064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356143, "loss/out": 5.651668548583984, "created_at": "2025-01-14T17:29:33.512604+00:00"} {"global_step": 3182, "acc_step": 0, "speed/wps": 12918.215310523208, "speed/FLOPS": 202898322939826.66, "speed/curr_iter_time": 1.2787, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1035702228546143, "optim/lr": 0.0019086, "optim/total_tokens": 834142208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 5.3848371505737305, "created_at": "2025-01-14T17:29:38.589947+00:00"} {"global_step": 3183, "acc_step": 0, "speed/wps": 12924.265967125199, "speed/FLOPS": 202993356816234.38, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6416677832603455, "optim/lr": 0.0019092, "optim/total_tokens": 834404352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 5.479206085205078, "created_at": "2025-01-14T17:29:43.669204+00:00"} {"global_step": 3184, "acc_step": 0, "speed/wps": 12917.128087721223, "speed/FLOPS": 202881246611720.78, "speed/curr_iter_time": 1.2804, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7111358642578125, "optim/lr": 0.0019098000000000001, "optim/total_tokens": 834666496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318319, "loss/out": 5.396781921386719, "created_at": "2025-01-14T17:29:48.748590+00:00"} {"global_step": 3185, "acc_step": 0, "speed/wps": 12919.047333032919, "speed/FLOPS": 202911390996660.4, "speed/curr_iter_time": 1.2784, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6750398874282837, "optim/lr": 0.0019104, "optim/total_tokens": 834928640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 5.480099201202393, "created_at": "2025-01-14T17:29:53.830690+00:00"} {"global_step": 3186, "acc_step": 0, "speed/wps": 12918.311540062597, "speed/FLOPS": 202899834356976.0, "speed/curr_iter_time": 1.279, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7608880400657654, "optim/lr": 0.0019110000000000002, "optim/total_tokens": 835190784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438251, "loss/out": 5.4920783042907715, "created_at": "2025-01-14T17:29:58.913726+00:00"} {"global_step": 3187, "acc_step": 0, "speed/wps": 12921.168704761347, "speed/FLOPS": 202944710054725.6, "speed/curr_iter_time": 1.2781, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5951789021492004, "optim/lr": 0.0019116, "optim/total_tokens": 835452928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440628, "loss/out": 5.501750946044922, "created_at": "2025-01-14T17:30:03.989463+00:00"} {"global_step": 3188, "acc_step": 0, "speed/wps": 12917.064421297446, "speed/FLOPS": 202880246643048.53, "speed/curr_iter_time": 1.2784, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5793558359146118, "optim/lr": 0.0019122, "optim/total_tokens": 835715072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472260, "loss/out": 5.418158054351807, "created_at": "2025-01-14T17:30:09.070671+00:00"} {"global_step": 3189, "acc_step": 0, "speed/wps": 12913.07780526252, "speed/FLOPS": 202817631360035.5, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.0201165676116943, "optim/lr": 0.0019127999999999999, "optim/total_tokens": 835977216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431842, "loss/out": 5.494873046875, "created_at": "2025-01-14T17:30:14.154808+00:00"} {"global_step": 3190, "acc_step": 0, "speed/wps": 12921.54835262854, "speed/FLOPS": 202950672946170.72, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7952086925506592, "optim/lr": 0.0019134000000000002, "optim/total_tokens": 836239360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334467, "loss/out": 5.440647125244141, "created_at": "2025-01-14T17:30:19.237633+00:00"} {"global_step": 3191, "acc_step": 0, "speed/wps": 12919.992509587237, "speed/FLOPS": 202926236293254.66, "speed/curr_iter_time": 1.2781, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8590878248214722, "optim/lr": 0.0019140000000000001, "optim/total_tokens": 836501504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 5.383932590484619, "created_at": "2025-01-14T17:30:24.322264+00:00"} {"global_step": 3192, "acc_step": 0, "speed/wps": 12923.551640983389, "speed/FLOPS": 202982137342590.28, "speed/curr_iter_time": 1.2785, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9472472667694092, "optim/lr": 0.0019146, "optim/total_tokens": 836763648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459237, "loss/out": 5.435033798217773, "created_at": "2025-01-14T17:30:29.400783+00:00"} {"global_step": 3193, "acc_step": 0, "speed/wps": 12922.90157992435, "speed/FLOPS": 202971927240380.56, "speed/curr_iter_time": 1.2779, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.549272894859314, "optim/lr": 0.0019152, "optim/total_tokens": 837025792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319663, "loss/out": 5.443286895751953, "created_at": "2025-01-14T17:30:34.477880+00:00"} {"global_step": 3194, "acc_step": 0, "speed/wps": 12924.509431321969, "speed/FLOPS": 202997180755999.5, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6794357895851135, "optim/lr": 0.0019157999999999998, "optim/total_tokens": 837287936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 5.4954729080200195, "created_at": "2025-01-14T17:30:39.556424+00:00"} {"global_step": 3195, "acc_step": 0, "speed/wps": 12917.193820032151, "speed/FLOPS": 202882279027989.56, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6224571466445923, "optim/lr": 0.0019164000000000002, "optim/total_tokens": 837550080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446218, "loss/out": 5.556247711181641, "created_at": "2025-01-14T17:30:44.634382+00:00"} {"global_step": 3196, "acc_step": 0, "speed/wps": 12920.098002956864, "speed/FLOPS": 202927893211587.72, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6698605418205261, "optim/lr": 0.001917, "optim/total_tokens": 837812224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306979, "loss/out": 5.403022766113281, "created_at": "2025-01-14T17:30:49.712354+00:00"} {"global_step": 3197, "acc_step": 0, "speed/wps": 12922.667699728705, "speed/FLOPS": 202968253830522.88, "speed/curr_iter_time": 1.2786, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6932411789894104, "optim/lr": 0.0019176, "optim/total_tokens": 838074368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418530, "loss/out": 5.435299396514893, "created_at": "2025-01-14T17:30:54.792725+00:00"} {"global_step": 3198, "acc_step": 0, "speed/wps": 12923.636974956557, "speed/FLOPS": 202983477629900.03, "speed/curr_iter_time": 1.2774, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.572796106338501, "optim/lr": 0.0019182, "optim/total_tokens": 838336512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 5.410768032073975, "created_at": "2025-01-14T17:30:59.868884+00:00"} {"global_step": 3199, "acc_step": 0, "speed/wps": 12926.071215751805, "speed/FLOPS": 203021710726586.78, "speed/curr_iter_time": 1.2781, "speed/data_load_time": 0.001, "optim/grad_norm": 0.5460453033447266, "optim/lr": 0.0019187999999999998, "optim/total_tokens": 838598656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454466, "loss/out": 5.3596296310424805, "created_at": "2025-01-14T17:31:04.944300+00:00"} {"global_step": 3200, "acc_step": 0, "speed/wps": 12924.853497640466, "speed/FLOPS": 203002584790328.2, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7380021214485168, "optim/lr": 0.0019194000000000001, "optim/total_tokens": 838860800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331435, "loss/out": 5.543121814727783, "created_at": "2025-01-14T17:31:10.019047+00:00"} {"global_step": 3201, "acc_step": 0, "speed/wps": 12917.888844641662, "speed/FLOPS": 202893195344545.84, "speed/curr_iter_time": 1.2787, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.49645838141441345, "optim/lr": 0.00192, "optim/total_tokens": 839122944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 5.420345306396484, "created_at": "2025-01-14T17:31:15.097633+00:00"} {"global_step": 3202, "acc_step": 0, "speed/wps": 12925.069677836374, "speed/FLOPS": 203005980197367.66, "speed/curr_iter_time": 1.2785, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6444792747497559, "optim/lr": 0.0019206, "optim/total_tokens": 839385088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446742, "loss/out": 5.449588775634766, "created_at": "2025-01-14T17:31:20.172494+00:00"} {"global_step": 3203, "acc_step": 0, "speed/wps": 12924.062294499116, "speed/FLOPS": 202990157857766.22, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8632155656814575, "optim/lr": 0.0019211999999999999, "optim/total_tokens": 839647232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454182, "loss/out": 5.405955791473389, "created_at": "2025-01-14T17:31:25.251779+00:00"} {"global_step": 3204, "acc_step": 0, "speed/wps": 12919.99067353251, "speed/FLOPS": 202926207455492.16, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7138339877128601, "optim/lr": 0.0019218, "optim/total_tokens": 839909376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441147, "loss/out": 5.452323913574219, "created_at": "2025-01-14T17:31:30.330491+00:00"} {"global_step": 3205, "acc_step": 0, "speed/wps": 12921.261129899493, "speed/FLOPS": 202946161718525.62, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9779121279716492, "optim/lr": 0.0019224, "optim/total_tokens": 840171520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442256, "loss/out": 5.363803386688232, "created_at": "2025-01-14T17:31:35.412261+00:00"} {"global_step": 3206, "acc_step": 0, "speed/wps": 12912.583110367163, "speed/FLOPS": 202809861496921.9, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5302391648292542, "optim/lr": 0.001923, "optim/total_tokens": 840433664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435866, "loss/out": 5.397038459777832, "created_at": "2025-01-14T17:31:40.491229+00:00"} {"global_step": 3207, "acc_step": 0, "speed/wps": 12919.355894820199, "speed/FLOPS": 202916237383537.97, "speed/curr_iter_time": 1.2782, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8132738471031189, "optim/lr": 0.0019236, "optim/total_tokens": 840695808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 5.338045597076416, "created_at": "2025-01-14T17:31:45.567726+00:00"} {"global_step": 3208, "acc_step": 0, "speed/wps": 12923.879109980633, "speed/FLOPS": 202987280693181.56, "speed/curr_iter_time": 1.2786, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9707282781600952, "optim/lr": 0.0019242, "optim/total_tokens": 840957952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463770, "loss/out": 5.30914306640625, "created_at": "2025-01-14T17:31:50.647148+00:00"} {"global_step": 3209, "acc_step": 0, "speed/wps": 12925.044755329958, "speed/FLOPS": 203005588755157.25, "speed/curr_iter_time": 1.2779, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7434163689613342, "optim/lr": 0.0019248, "optim/total_tokens": 841220096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 5.267434120178223, "created_at": "2025-01-14T17:31:55.721183+00:00"} {"global_step": 3210, "acc_step": 0, "speed/wps": 12930.072759412109, "speed/FLOPS": 203084560468469.1, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6999577879905701, "optim/lr": 0.0019254, "optim/total_tokens": 841482240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330329, "loss/out": 5.405435085296631, "created_at": "2025-01-14T17:32:00.798649+00:00"} {"global_step": 3211, "acc_step": 0, "speed/wps": 12930.935603607722, "speed/FLOPS": 203098112622233.38, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6286808252334595, "optim/lr": 0.001926, "optim/total_tokens": 841744384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 394401, "loss/out": 5.294094085693359, "created_at": "2025-01-14T17:32:05.873287+00:00"} {"global_step": 3212, "acc_step": 0, "speed/wps": 12926.85374556949, "speed/FLOPS": 203034001432682.2, "speed/curr_iter_time": 1.2784, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6354995369911194, "optim/lr": 0.0019266, "optim/total_tokens": 842006528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 5.32600212097168, "created_at": "2025-01-14T17:32:10.948384+00:00"} {"global_step": 3213, "acc_step": 0, "speed/wps": 12923.490287044391, "speed/FLOPS": 202981173694669.16, "speed/curr_iter_time": 1.2779, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6295320987701416, "optim/lr": 0.0019272, "optim/total_tokens": 842268672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 5.251625061035156, "created_at": "2025-01-14T17:32:16.024649+00:00"} {"global_step": 3214, "acc_step": 0, "speed/wps": 12935.095056695283, "speed/FLOPS": 203163442548661.0, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7472245097160339, "optim/lr": 0.0019278, "optim/total_tokens": 842530816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445404, "loss/out": 5.4960432052612305, "created_at": "2025-01-14T17:32:21.094924+00:00"} {"global_step": 3215, "acc_step": 0, "speed/wps": 12923.143257821335, "speed/FLOPS": 202975723123852.94, "speed/curr_iter_time": 1.2783, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5165261030197144, "optim/lr": 0.0019284000000000003, "optim/total_tokens": 842792960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 5.253489971160889, "created_at": "2025-01-14T17:32:26.173641+00:00"} {"global_step": 3216, "acc_step": 0, "speed/wps": 12931.235658473153, "speed/FLOPS": 203102825396213.56, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.026527762413025, "optim/lr": 0.0019290000000000002, "optim/total_tokens": 843055104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442256, "loss/out": 5.378108024597168, "created_at": "2025-01-14T17:32:31.246076+00:00"} {"global_step": 3217, "acc_step": 0, "speed/wps": 12926.266035566527, "speed/FLOPS": 203024770639487.47, "speed/curr_iter_time": 1.2778, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5820590853691101, "optim/lr": 0.0019296, "optim/total_tokens": 843317248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 5.356489181518555, "created_at": "2025-01-14T17:32:36.322697+00:00"} {"global_step": 3218, "acc_step": 0, "speed/wps": 12934.109795555085, "speed/FLOPS": 203147967668563.5, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.49282196164131165, "optim/lr": 0.0019302, "optim/total_tokens": 843579392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357486, "loss/out": 5.473814964294434, "created_at": "2025-01-14T17:32:41.394431+00:00"} {"global_step": 3219, "acc_step": 0, "speed/wps": 12928.663883597013, "speed/FLOPS": 203062432137794.75, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5468186140060425, "optim/lr": 0.0019307999999999999, "optim/total_tokens": 843841536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 5.3704705238342285, "created_at": "2025-01-14T17:32:46.468208+00:00"} {"global_step": 3220, "acc_step": 0, "speed/wps": 12927.210618219588, "speed/FLOPS": 203039606608046.38, "speed/curr_iter_time": 1.2786, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6343674063682556, "optim/lr": 0.0019314000000000002, "optim/total_tokens": 844103680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435329, "loss/out": 5.32733154296875, "created_at": "2025-01-14T17:32:51.541663+00:00"} {"global_step": 3221, "acc_step": 0, "speed/wps": 12927.908167625477, "speed/FLOPS": 203050562579999.22, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.7640881538391113, "optim/lr": 0.0019320000000000001, "optim/total_tokens": 844365824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453131, "loss/out": 5.387073516845703, "created_at": "2025-01-14T17:32:56.616856+00:00"} {"global_step": 3222, "acc_step": 0, "speed/wps": 12925.397249885618, "speed/FLOPS": 203011125166532.2, "speed/curr_iter_time": 1.2789, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.652970016002655, "optim/lr": 0.0019326, "optim/total_tokens": 844627968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 397678, "loss/out": 5.3011651039123535, "created_at": "2025-01-14T17:33:01.690897+00:00"} {"global_step": 3223, "acc_step": 0, "speed/wps": 12922.962294614676, "speed/FLOPS": 202972880848022.84, "speed/curr_iter_time": 1.2787, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.569558322429657, "optim/lr": 0.0019332, "optim/total_tokens": 844890112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 5.372299671173096, "created_at": "2025-01-14T17:33:06.769434+00:00"} {"global_step": 3224, "acc_step": 0, "speed/wps": 12931.92692858592, "speed/FLOPS": 203113682743239.25, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6217756867408752, "optim/lr": 0.0019337999999999998, "optim/total_tokens": 845152256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445407, "loss/out": 5.413517475128174, "created_at": "2025-01-14T17:33:11.848374+00:00"} {"global_step": 3225, "acc_step": 0, "speed/wps": 12930.84421858678, "speed/FLOPS": 203096677294903.28, "speed/curr_iter_time": 1.2782, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5984990000724792, "optim/lr": 0.0019344000000000002, "optim/total_tokens": 845414400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 5.299997329711914, "created_at": "2025-01-14T17:33:16.924956+00:00"} {"global_step": 3226, "acc_step": 0, "speed/wps": 12923.160284049984, "speed/FLOPS": 202975990544170.34, "speed/curr_iter_time": 1.2783, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5453152060508728, "optim/lr": 0.001935, "optim/total_tokens": 845676544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360941, "loss/out": 5.237826347351074, "created_at": "2025-01-14T17:33:22.002923+00:00"} {"global_step": 3227, "acc_step": 0, "speed/wps": 12928.482483088754, "speed/FLOPS": 203059582993542.12, "speed/curr_iter_time": 1.2783, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5122250914573669, "optim/lr": 0.0019356, "optim/total_tokens": 845938688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 5.318000793457031, "created_at": "2025-01-14T17:33:27.080858+00:00"} {"global_step": 3228, "acc_step": 0, "speed/wps": 12927.987654637227, "speed/FLOPS": 203051811032745.34, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7658484578132629, "optim/lr": 0.0019362, "optim/total_tokens": 846200832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 5.421780586242676, "created_at": "2025-01-14T17:33:32.153695+00:00"} {"global_step": 3229, "acc_step": 0, "speed/wps": 12929.614684712495, "speed/FLOPS": 203077365775851.03, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5878950953483582, "optim/lr": 0.0019367999999999998, "optim/total_tokens": 846462976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325239, "loss/out": 5.252462387084961, "created_at": "2025-01-14T17:33:37.226564+00:00"} {"global_step": 3230, "acc_step": 0, "speed/wps": 12929.36341669276, "speed/FLOPS": 203073419266322.66, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6495856046676636, "optim/lr": 0.0019374000000000001, "optim/total_tokens": 846725120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 5.214836120605469, "created_at": "2025-01-14T17:33:42.299549+00:00"} {"global_step": 3231, "acc_step": 0, "speed/wps": 12921.524670399709, "speed/FLOPS": 202950300984224.12, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6203017830848694, "optim/lr": 0.001938, "optim/total_tokens": 846987264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 5.401820659637451, "created_at": "2025-01-14T17:33:47.376688+00:00"} {"global_step": 3232, "acc_step": 0, "speed/wps": 12928.848788184541, "speed/FLOPS": 203065336318427.4, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.609815776348114, "optim/lr": 0.0019386, "optim/total_tokens": 847249408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 5.349650859832764, "created_at": "2025-01-14T17:33:52.450892+00:00"} {"global_step": 3233, "acc_step": 0, "speed/wps": 12921.65377412337, "speed/FLOPS": 202952328735611.3, "speed/curr_iter_time": 1.279, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.573039174079895, "optim/lr": 0.0019391999999999999, "optim/total_tokens": 847511552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336478, "loss/out": 5.311561584472656, "created_at": "2025-01-14T17:33:57.531212+00:00"} {"global_step": 3234, "acc_step": 0, "speed/wps": 12934.424780821779, "speed/FLOPS": 203152914945012.06, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6687129735946655, "optim/lr": 0.0019398, "optim/total_tokens": 847773696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 5.183974266052246, "created_at": "2025-01-14T17:34:02.601632+00:00"} {"global_step": 3235, "acc_step": 0, "speed/wps": 12930.193342917693, "speed/FLOPS": 203086454398123.38, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7170401811599731, "optim/lr": 0.0019404000000000001, "optim/total_tokens": 848035840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428403, "loss/out": 5.21079683303833, "created_at": "2025-01-14T17:34:07.674372+00:00"} {"global_step": 3236, "acc_step": 0, "speed/wps": 12931.926160960798, "speed/FLOPS": 203113670686631.88, "speed/curr_iter_time": 1.2786, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.7290608882904053, "optim/lr": 0.001941, "optim/total_tokens": 848297984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440921, "loss/out": 5.328701972961426, "created_at": "2025-01-14T17:34:12.747003+00:00"} {"global_step": 3237, "acc_step": 0, "speed/wps": 12925.926134615935, "speed/FLOPS": 203019432027985.6, "speed/curr_iter_time": 1.2787, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.839794397354126, "optim/lr": 0.0019416000000000001, "optim/total_tokens": 848560128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349477, "loss/out": 5.227504730224609, "created_at": "2025-01-14T17:34:17.822343+00:00"} {"global_step": 3238, "acc_step": 0, "speed/wps": 12922.071794520269, "speed/FLOPS": 202958894320364.84, "speed/curr_iter_time": 1.2788, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6876269578933716, "optim/lr": 0.0019422, "optim/total_tokens": 848822272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 5.1900410652160645, "created_at": "2025-01-14T17:34:22.898849+00:00"} {"global_step": 3239, "acc_step": 0, "speed/wps": 12926.571007600038, "speed/FLOPS": 203029560644349.25, "speed/curr_iter_time": 1.2788, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7892016172409058, "optim/lr": 0.0019428, "optim/total_tokens": 849084416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 5.261841773986816, "created_at": "2025-01-14T17:34:27.973047+00:00"} {"global_step": 3240, "acc_step": 0, "speed/wps": 12920.847297176786, "speed/FLOPS": 202939661906949.75, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.001, "optim/grad_norm": 0.7259877324104309, "optim/lr": 0.0019434, "optim/total_tokens": 849346560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 5.193641662597656, "created_at": "2025-01-14T17:34:33.054833+00:00"} {"global_step": 3241, "acc_step": 0, "speed/wps": 12925.882840982575, "speed/FLOPS": 203018752041981.97, "speed/curr_iter_time": 1.2783, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0260398387908936, "optim/lr": 0.0019440000000000002, "optim/total_tokens": 849608704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349068, "loss/out": 5.160087585449219, "created_at": "2025-01-14T17:34:38.131228+00:00"} {"global_step": 3242, "acc_step": 0, "speed/wps": 12928.881141448874, "speed/FLOPS": 203065844470901.78, "speed/curr_iter_time": 1.2771, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7233453989028931, "optim/lr": 0.0019446, "optim/total_tokens": 849870848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 5.2412567138671875, "created_at": "2025-01-14T17:34:43.211818+00:00"} {"global_step": 3243, "acc_step": 0, "speed/wps": 12924.500235138175, "speed/FLOPS": 202997036317296.16, "speed/curr_iter_time": 1.2779, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7246950268745422, "optim/lr": 0.0019452, "optim/total_tokens": 850132992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480426, "loss/out": 5.323065757751465, "created_at": "2025-01-14T17:34:48.286019+00:00"} {"global_step": 3244, "acc_step": 0, "speed/wps": 12929.233158755462, "speed/FLOPS": 203071373386424.22, "speed/curr_iter_time": 1.2779, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8546085357666016, "optim/lr": 0.0019458, "optim/total_tokens": 850395136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298060, "loss/out": 5.211219310760498, "created_at": "2025-01-14T17:34:53.364014+00:00"} {"global_step": 3245, "acc_step": 0, "speed/wps": 12936.970243246105, "speed/FLOPS": 203192894930218.03, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0483167171478271, "optim/lr": 0.0019464000000000003, "optim/total_tokens": 850657280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 5.318543434143066, "created_at": "2025-01-14T17:34:58.434576+00:00"} {"global_step": 3246, "acc_step": 0, "speed/wps": 12923.80130649408, "speed/FLOPS": 202986058682512.03, "speed/curr_iter_time": 1.2786, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8366401195526123, "optim/lr": 0.0019470000000000002, "optim/total_tokens": 850919424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428118, "loss/out": 5.195566177368164, "created_at": "2025-01-14T17:35:03.509424+00:00"} {"global_step": 3247, "acc_step": 0, "speed/wps": 12932.317794696957, "speed/FLOPS": 203119821832619.38, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1750807762145996, "optim/lr": 0.0019476, "optim/total_tokens": 851181568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458173, "loss/out": 5.232716083526611, "created_at": "2025-01-14T17:35:08.585480+00:00"} {"global_step": 3248, "acc_step": 0, "speed/wps": 12931.489097618703, "speed/FLOPS": 203106806006256.28, "speed/curr_iter_time": 1.2784, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9470084309577942, "optim/lr": 0.0019482, "optim/total_tokens": 851443712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 398779, "loss/out": 5.309889793395996, "created_at": "2025-01-14T17:35:13.660715+00:00"} {"global_step": 3249, "acc_step": 0, "speed/wps": 12932.074827644225, "speed/FLOPS": 203116005701184.5, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1399105787277222, "optim/lr": 0.0019487999999999999, "optim/total_tokens": 851705856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 5.250151634216309, "created_at": "2025-01-14T17:35:18.782733+00:00"} {"global_step": 3250, "acc_step": 0, "speed/wps": 12925.080173426277, "speed/FLOPS": 203006145045029.5, "speed/curr_iter_time": 1.279, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.078290343284607, "optim/lr": 0.0019494000000000002, "optim/total_tokens": 851968000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361998, "loss/out": 5.20955753326416, "created_at": "2025-01-14T17:35:23.859693+00:00"} {"global_step": 3251, "acc_step": 0, "speed/wps": 12926.855585950037, "speed/FLOPS": 203034030338387.62, "speed/curr_iter_time": 1.2781, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9916976690292358, "optim/lr": 0.0019500000000000001, "optim/total_tokens": 852230144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 5.221110820770264, "created_at": "2025-01-14T17:35:28.938635+00:00"} {"global_step": 3252, "acc_step": 0, "speed/wps": 12932.445961469299, "speed/FLOPS": 203121834867898.62, "speed/curr_iter_time": 1.2778, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2120875120162964, "optim/lr": 0.0019506, "optim/total_tokens": 852492288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 5.276838302612305, "created_at": "2025-01-14T17:35:34.015015+00:00"} {"global_step": 3253, "acc_step": 0, "speed/wps": 12928.080122725278, "speed/FLOPS": 203053263371132.66, "speed/curr_iter_time": 1.2784, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4334660768508911, "optim/lr": 0.0019512, "optim/total_tokens": 852754432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480663, "loss/out": 5.324244976043701, "created_at": "2025-01-14T17:35:39.093377+00:00"} {"global_step": 3254, "acc_step": 0, "speed/wps": 12929.096078604205, "speed/FLOPS": 203069220354280.88, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.005354642868042, "optim/lr": 0.0019517999999999998, "optim/total_tokens": 853016576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358828, "loss/out": 5.190905570983887, "created_at": "2025-01-14T17:35:44.183654+00:00"} {"global_step": 3255, "acc_step": 0, "speed/wps": 12922.508297443112, "speed/FLOPS": 202965750198585.94, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0524193048477173, "optim/lr": 0.0019524000000000002, "optim/total_tokens": 853278720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452093, "loss/out": 5.2862043380737305, "created_at": "2025-01-14T17:35:49.265614+00:00"} {"global_step": 3256, "acc_step": 0, "speed/wps": 12936.814526350085, "speed/FLOPS": 203190449182388.56, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0274202823638916, "optim/lr": 0.001953, "optim/total_tokens": 853540864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442784, "loss/out": 5.257110595703125, "created_at": "2025-01-14T17:35:54.335087+00:00"} {"global_step": 3257, "acc_step": 0, "speed/wps": 12932.200541651684, "speed/FLOPS": 203117980212420.8, "speed/curr_iter_time": 1.2788, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4040688276290894, "optim/lr": 0.0019536, "optim/total_tokens": 853803008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367900, "loss/out": 5.206341743469238, "created_at": "2025-01-14T17:35:59.412364+00:00"} {"global_step": 3258, "acc_step": 0, "speed/wps": 12937.57164314614, "speed/FLOPS": 203202340742058.12, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0902132987976074, "optim/lr": 0.0019542, "optim/total_tokens": 854065152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 5.311150074005127, "created_at": "2025-01-14T17:36:04.483240+00:00"} {"global_step": 3259, "acc_step": 0, "speed/wps": 12926.53115988982, "speed/FLOPS": 203028934781303.8, "speed/curr_iter_time": 1.2781, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.341797113418579, "optim/lr": 0.0019548, "optim/total_tokens": 854327296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 5.275897026062012, "created_at": "2025-01-14T17:36:09.556807+00:00"} {"global_step": 3260, "acc_step": 0, "speed/wps": 12918.159134402831, "speed/FLOPS": 202897440616662.06, "speed/curr_iter_time": 1.2807, "speed/data_load_time": 0.0011, "optim/grad_norm": 2.585902452468872, "optim/lr": 0.0019554000000000004, "optim/total_tokens": 854589440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322066, "loss/out": 5.257162094116211, "created_at": "2025-01-14T17:36:14.633496+00:00"} {"global_step": 3261, "acc_step": 0, "speed/wps": 12923.797451829088, "speed/FLOPS": 202985998139701.47, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5477536916732788, "optim/lr": 0.0019560000000000003, "optim/total_tokens": 854851584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 5.307499885559082, "created_at": "2025-01-14T17:36:19.708446+00:00"} {"global_step": 3262, "acc_step": 0, "speed/wps": 12926.995703377559, "speed/FLOPS": 203036231075128.38, "speed/curr_iter_time": 1.2771, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.7275151014328003, "optim/lr": 0.0019566, "optim/total_tokens": 855113728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 5.315034866333008, "created_at": "2025-01-14T17:36:24.783060+00:00"} {"global_step": 3263, "acc_step": 0, "speed/wps": 12932.499478729187, "speed/FLOPS": 203122675430005.6, "speed/curr_iter_time": 1.2778, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3607580661773682, "optim/lr": 0.0019572, "optim/total_tokens": 855375872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451275, "loss/out": 5.3432230949401855, "created_at": "2025-01-14T17:36:29.856724+00:00"} {"global_step": 3264, "acc_step": 0, "speed/wps": 12936.769956313221, "speed/FLOPS": 203189749148713.97, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.335967779159546, "optim/lr": 0.0019578, "optim/total_tokens": 855638016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 5.331192493438721, "created_at": "2025-01-14T17:36:34.926774+00:00"} {"global_step": 3265, "acc_step": 0, "speed/wps": 12921.54728015222, "speed/FLOPS": 202950656101456.34, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1608357429504395, "optim/lr": 0.0019584000000000003, "optim/total_tokens": 855900160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328985, "loss/out": 5.423135280609131, "created_at": "2025-01-14T17:36:40.006873+00:00"} {"global_step": 3266, "acc_step": 0, "speed/wps": 12934.884394746461, "speed/FLOPS": 203160133813274.9, "speed/curr_iter_time": 1.2779, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2056574821472168, "optim/lr": 0.0019590000000000002, "optim/total_tokens": 856162304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455247, "loss/out": 5.2894206047058105, "created_at": "2025-01-14T17:36:45.080744+00:00"} {"global_step": 3267, "acc_step": 0, "speed/wps": 12924.563692396112, "speed/FLOPS": 202998033000731.47, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2476557493209839, "optim/lr": 0.0019596, "optim/total_tokens": 856424448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482522, "loss/out": 5.375344276428223, "created_at": "2025-01-14T17:36:50.156054+00:00"} {"global_step": 3268, "acc_step": 0, "speed/wps": 12931.967840997944, "speed/FLOPS": 203114325328889.62, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.446799635887146, "optim/lr": 0.0019602, "optim/total_tokens": 856686592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 5.319807052612305, "created_at": "2025-01-14T17:36:55.231393+00:00"} {"global_step": 3269, "acc_step": 0, "speed/wps": 12924.647964369995, "speed/FLOPS": 202999356607883.12, "speed/curr_iter_time": 1.2797, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0888760089874268, "optim/lr": 0.0019608, "optim/total_tokens": 856948736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299645, "loss/out": 5.259738922119141, "created_at": "2025-01-14T17:37:00.306162+00:00"} {"global_step": 3270, "acc_step": 0, "speed/wps": 12926.380728152857, "speed/FLOPS": 203026572044159.94, "speed/curr_iter_time": 1.2774, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1779319047927856, "optim/lr": 0.0019614000000000003, "optim/total_tokens": 857210880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 5.332303047180176, "created_at": "2025-01-14T17:37:05.380661+00:00"} {"global_step": 3271, "acc_step": 0, "speed/wps": 12928.68527365318, "speed/FLOPS": 203062768098019.88, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2371947765350342, "optim/lr": 0.001962, "optim/total_tokens": 857473024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431605, "loss/out": 5.260204792022705, "created_at": "2025-01-14T17:37:10.453671+00:00"} {"global_step": 3272, "acc_step": 0, "speed/wps": 12927.226624494917, "speed/FLOPS": 203039858008595.38, "speed/curr_iter_time": 1.2786, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2916687726974487, "optim/lr": 0.0019626, "optim/total_tokens": 857735168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445933, "loss/out": 5.201406955718994, "created_at": "2025-01-14T17:37:15.529593+00:00"} {"global_step": 3273, "acc_step": 0, "speed/wps": 12930.813723142172, "speed/FLOPS": 203096198322040.56, "speed/curr_iter_time": 1.2779, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2870869636535645, "optim/lr": 0.0019632, "optim/total_tokens": 857997312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477322, "loss/out": 5.242978096008301, "created_at": "2025-01-14T17:37:20.606249+00:00"} {"global_step": 3274, "acc_step": 0, "speed/wps": 12931.901516202648, "speed/FLOPS": 203113283606839.66, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2199536561965942, "optim/lr": 0.0019638, "optim/total_tokens": 858259456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453131, "loss/out": 5.331545352935791, "created_at": "2025-01-14T17:37:25.734598+00:00"} {"global_step": 3275, "acc_step": 0, "speed/wps": 12933.934941384045, "speed/FLOPS": 203145221343532.5, "speed/curr_iter_time": 1.2782, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6004903316497803, "optim/lr": 0.0019644000000000003, "optim/total_tokens": 858521600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 5.2081451416015625, "created_at": "2025-01-14T17:37:30.808082+00:00"} {"global_step": 3276, "acc_step": 0, "speed/wps": 12921.708780007573, "speed/FLOPS": 202953192678608.72, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.176895022392273, "optim/lr": 0.001965, "optim/total_tokens": 858783744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444878, "loss/out": 5.270001411437988, "created_at": "2025-01-14T17:37:35.884063+00:00"} {"global_step": 3277, "acc_step": 0, "speed/wps": 12921.667251494528, "speed/FLOPS": 202952540416245.2, "speed/curr_iter_time": 1.2785, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0843420028686523, "optim/lr": 0.0019656, "optim/total_tokens": 859045888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456851, "loss/out": 5.242465019226074, "created_at": "2025-01-14T17:37:40.960829+00:00"} {"global_step": 3278, "acc_step": 0, "speed/wps": 12924.339178230875, "speed/FLOPS": 202994506697251.62, "speed/curr_iter_time": 1.2781, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1908951997756958, "optim/lr": 0.0019662, "optim/total_tokens": 859308032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 343445, "loss/out": 5.190645694732666, "created_at": "2025-01-14T17:37:46.053539+00:00"} {"global_step": 3279, "acc_step": 0, "speed/wps": 12925.522729634557, "speed/FLOPS": 203013095998416.5, "speed/curr_iter_time": 1.2785, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1890970468521118, "optim/lr": 0.0019668, "optim/total_tokens": 859570176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 5.31889533996582, "created_at": "2025-01-14T17:37:51.131822+00:00"} {"global_step": 3280, "acc_step": 0, "speed/wps": 12927.965117344458, "speed/FLOPS": 203051457053592.94, "speed/curr_iter_time": 1.2782, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1607102155685425, "optim/lr": 0.0019674000000000002, "optim/total_tokens": 859832320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 5.247354984283447, "created_at": "2025-01-14T17:37:56.205344+00:00"} {"global_step": 3281, "acc_step": 0, "speed/wps": 12917.22680025362, "speed/FLOPS": 202882797027687.3, "speed/curr_iter_time": 1.2783, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1679240465164185, "optim/lr": 0.001968, "optim/total_tokens": 860094464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366844, "loss/out": 5.2686686515808105, "created_at": "2025-01-14T17:38:01.282787+00:00"} {"global_step": 3282, "acc_step": 0, "speed/wps": 12923.523529651757, "speed/FLOPS": 202981695815497.97, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1744260787963867, "optim/lr": 0.0019686, "optim/total_tokens": 860356608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 5.284965515136719, "created_at": "2025-01-14T17:38:06.362278+00:00"} {"global_step": 3283, "acc_step": 0, "speed/wps": 12923.154054068447, "speed/FLOPS": 202975892693749.44, "speed/curr_iter_time": 1.2783, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1936196088790894, "optim/lr": 0.0019692, "optim/total_tokens": 860618752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 5.273787498474121, "created_at": "2025-01-14T17:38:11.443376+00:00"} {"global_step": 3284, "acc_step": 0, "speed/wps": 12915.839621026254, "speed/FLOPS": 202861009471737.53, "speed/curr_iter_time": 1.2799, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2865453958511353, "optim/lr": 0.0019698, "optim/total_tokens": 860880896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360518, "loss/out": 5.218639373779297, "created_at": "2025-01-14T17:38:16.521059+00:00"} {"global_step": 3285, "acc_step": 0, "speed/wps": 12928.539414114744, "speed/FLOPS": 203060477173536.72, "speed/curr_iter_time": 1.2778, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.370972990989685, "optim/lr": 0.0019704, "optim/total_tokens": 861143040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 5.231152534484863, "created_at": "2025-01-14T17:38:21.609700+00:00"} {"global_step": 3286, "acc_step": 0, "speed/wps": 12923.99973465161, "speed/FLOPS": 202989175269394.9, "speed/curr_iter_time": 1.2784, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7844922542572021, "optim/lr": 0.001971, "optim/total_tokens": 861405184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 5.128504753112793, "created_at": "2025-01-14T17:38:26.684230+00:00"} {"global_step": 3287, "acc_step": 0, "speed/wps": 12928.356739221927, "speed/FLOPS": 203057608013323.4, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4750374555587769, "optim/lr": 0.0019716, "optim/total_tokens": 861667328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 373078, "loss/out": 5.248903751373291, "created_at": "2025-01-14T17:38:31.758177+00:00"} {"global_step": 3288, "acc_step": 0, "speed/wps": 12919.7411030428, "speed/FLOPS": 202922287608005.47, "speed/curr_iter_time": 1.2789, "speed/data_load_time": 0.001, "optim/grad_norm": 1.2408607006072998, "optim/lr": 0.0019722, "optim/total_tokens": 861929472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 5.266629695892334, "created_at": "2025-01-14T17:38:36.834357+00:00"} {"global_step": 3289, "acc_step": 0, "speed/wps": 12929.766852382425, "speed/FLOPS": 203079755778207.6, "speed/curr_iter_time": 1.2774, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3219547271728516, "optim/lr": 0.0019728, "optim/total_tokens": 862191616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 5.29046630859375, "created_at": "2025-01-14T17:38:41.909885+00:00"} {"global_step": 3290, "acc_step": 0, "speed/wps": 12920.15713275462, "speed/FLOPS": 202928821926314.34, "speed/curr_iter_time": 1.2784, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3332411050796509, "optim/lr": 0.0019734, "optim/total_tokens": 862453760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320435, "loss/out": 5.2693023681640625, "created_at": "2025-01-14T17:38:46.993722+00:00"} {"global_step": 3291, "acc_step": 0, "speed/wps": 12926.384175942958, "speed/FLOPS": 203026626196441.22, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.241757869720459, "optim/lr": 0.001974, "optim/total_tokens": 862715904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448083, "loss/out": 5.243555545806885, "created_at": "2025-01-14T17:38:52.069793+00:00"} {"global_step": 3292, "acc_step": 0, "speed/wps": 12930.718711531395, "speed/FLOPS": 203094706034133.72, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1453397274017334, "optim/lr": 0.0019746, "optim/total_tokens": 862978048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 5.176972389221191, "created_at": "2025-01-14T17:38:57.142970+00:00"} {"global_step": 3293, "acc_step": 0, "speed/wps": 12923.282865201836, "speed/FLOPS": 202977915849604.34, "speed/curr_iter_time": 1.2779, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1405490636825562, "optim/lr": 0.0019752, "optim/total_tokens": 863240192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 5.249266624450684, "created_at": "2025-01-14T17:39:02.222834+00:00"} {"global_step": 3294, "acc_step": 0, "speed/wps": 12922.973346871373, "speed/FLOPS": 202973054438901.7, "speed/curr_iter_time": 1.2783, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.0149999856948853, "optim/lr": 0.0019757999999999998, "optim/total_tokens": 863502336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333025, "loss/out": 5.101114749908447, "created_at": "2025-01-14T17:39:07.297730+00:00"} {"global_step": 3295, "acc_step": 0, "speed/wps": 12926.63668407737, "speed/FLOPS": 203030592183674.7, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9940075874328613, "optim/lr": 0.0019764, "optim/total_tokens": 863764480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 5.193168640136719, "created_at": "2025-01-14T17:39:12.372499+00:00"} {"global_step": 3296, "acc_step": 0, "speed/wps": 12926.628741030227, "speed/FLOPS": 203030467427204.44, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1263314485549927, "optim/lr": 0.001977, "optim/total_tokens": 864026624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445404, "loss/out": 5.306044578552246, "created_at": "2025-01-14T17:39:17.450627+00:00"} {"global_step": 3297, "acc_step": 0, "speed/wps": 12934.83372752938, "speed/FLOPS": 203159338015006.6, "speed/curr_iter_time": 1.2778, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0824635028839111, "optim/lr": 0.0019776, "optim/total_tokens": 864288768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449942, "loss/out": 5.262749671936035, "created_at": "2025-01-14T17:39:22.524439+00:00"} {"global_step": 3298, "acc_step": 0, "speed/wps": 12931.904275365212, "speed/FLOPS": 203113326943279.22, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1101261377334595, "optim/lr": 0.0019782, "optim/total_tokens": 864550912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440339, "loss/out": 5.303983688354492, "created_at": "2025-01-14T17:39:27.606271+00:00"} {"global_step": 3299, "acc_step": 0, "speed/wps": 12932.124684211432, "speed/FLOPS": 203116788767082.6, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9501560926437378, "optim/lr": 0.0019787999999999997, "optim/total_tokens": 864813056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457665, "loss/out": 5.236217021942139, "created_at": "2025-01-14T17:39:32.679864+00:00"} {"global_step": 3300, "acc_step": 0, "speed/wps": 12927.314664467456, "speed/FLOPS": 203041240797343.7, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1074070930480957, "optim/lr": 0.0019794, "optim/total_tokens": 865075200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460034, "loss/out": 5.192895889282227, "created_at": "2025-01-14T17:39:37.758729+00:00"} {"global_step": 3301, "acc_step": 0, "speed/wps": 12927.838490523916, "speed/FLOPS": 203049468205372.84, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0809077024459839, "optim/lr": 0.00198, "optim/total_tokens": 865337344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 394516, "loss/out": 5.110304832458496, "created_at": "2025-01-14T17:39:42.832970+00:00"} {"global_step": 3302, "acc_step": 0, "speed/wps": 12927.962255358534, "speed/FLOPS": 203051412102171.22, "speed/curr_iter_time": 1.2781, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8193038702011108, "optim/lr": 0.0019806, "optim/total_tokens": 865599488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 5.121192455291748, "created_at": "2025-01-14T17:39:47.909665+00:00"} {"global_step": 3303, "acc_step": 0, "speed/wps": 12932.347244356743, "speed/FLOPS": 203120284379994.4, "speed/curr_iter_time": 1.2779, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9898632764816284, "optim/lr": 0.0019812, "optim/total_tokens": 865861632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 5.217376708984375, "created_at": "2025-01-14T17:39:52.985312+00:00"} {"global_step": 3304, "acc_step": 0, "speed/wps": 12937.558693931882, "speed/FLOPS": 203202137356854.12, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7189374566078186, "optim/lr": 0.0019818, "optim/total_tokens": 866123776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453131, "loss/out": 5.233364105224609, "created_at": "2025-01-14T17:39:58.055494+00:00"} {"global_step": 3305, "acc_step": 0, "speed/wps": 12926.369463485078, "speed/FLOPS": 203026395117073.2, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.815024733543396, "optim/lr": 0.0019824, "optim/total_tokens": 866385920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311717, "loss/out": 5.226624488830566, "created_at": "2025-01-14T17:40:03.130263+00:00"} {"global_step": 3306, "acc_step": 0, "speed/wps": 12928.514421218579, "speed/FLOPS": 203060084625758.62, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0919382572174072, "optim/lr": 0.001983, "optim/total_tokens": 866648064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494531, "loss/out": 5.168020248413086, "created_at": "2025-01-14T17:40:08.213758+00:00"} {"global_step": 3307, "acc_step": 0, "speed/wps": 12927.673032683315, "speed/FLOPS": 203046869462623.28, "speed/curr_iter_time": 1.2779, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.8519387245178223, "optim/lr": 0.0019836, "optim/total_tokens": 866910208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465870, "loss/out": 5.185018539428711, "created_at": "2025-01-14T17:40:13.287274+00:00"} {"global_step": 3308, "acc_step": 0, "speed/wps": 12929.070190368206, "speed/FLOPS": 203068813743960.34, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1642262935638428, "optim/lr": 0.0019842, "optim/total_tokens": 867172352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345608, "loss/out": 5.0465312004089355, "created_at": "2025-01-14T17:40:18.359767+00:00"} {"global_step": 3309, "acc_step": 0, "speed/wps": 12922.039603463827, "speed/FLOPS": 202958388715588.25, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2166645526885986, "optim/lr": 0.0019848, "optim/total_tokens": 867434496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 5.194482803344727, "created_at": "2025-01-14T17:40:23.438171+00:00"} {"global_step": 3310, "acc_step": 0, "speed/wps": 12924.185118265344, "speed/FLOPS": 202992086973792.38, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3383499383926392, "optim/lr": 0.0019854, "optim/total_tokens": 867696640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441157, "loss/out": 5.181426525115967, "created_at": "2025-01-14T17:40:28.513146+00:00"} {"global_step": 3311, "acc_step": 0, "speed/wps": 12930.974841655894, "speed/FLOPS": 203098728909698.62, "speed/curr_iter_time": 1.2785, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9343685507774353, "optim/lr": 0.0019860000000000004, "optim/total_tokens": 867958784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462938, "loss/out": 5.237000942230225, "created_at": "2025-01-14T17:40:33.589006+00:00"} {"global_step": 3312, "acc_step": 0, "speed/wps": 12923.697592122104, "speed/FLOPS": 202984429705781.66, "speed/curr_iter_time": 1.2784, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7400788068771362, "optim/lr": 0.0019866000000000003, "optim/total_tokens": 868220928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 406438, "loss/out": 5.231205940246582, "created_at": "2025-01-14T17:40:38.663989+00:00"} {"global_step": 3313, "acc_step": 0, "speed/wps": 12920.928967874424, "speed/FLOPS": 202940944657483.47, "speed/curr_iter_time": 1.2783, "speed/data_load_time": 0.001, "optim/grad_norm": 0.9525064826011658, "optim/lr": 0.0019872, "optim/total_tokens": 868483072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 5.198507785797119, "created_at": "2025-01-14T17:40:43.744857+00:00"} {"global_step": 3314, "acc_step": 0, "speed/wps": 12926.909206557853, "speed/FLOPS": 203034872523715.8, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8751453161239624, "optim/lr": 0.0019878, "optim/total_tokens": 868745216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462701, "loss/out": 5.118577003479004, "created_at": "2025-01-14T17:40:48.823862+00:00"} {"global_step": 3315, "acc_step": 0, "speed/wps": 12925.924501067851, "speed/FLOPS": 203019406370868.2, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8316752910614014, "optim/lr": 0.0019884, "optim/total_tokens": 869007360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 385060, "loss/out": 5.090610504150391, "created_at": "2025-01-14T17:40:53.897567+00:00"} {"global_step": 3316, "acc_step": 0, "speed/wps": 12925.613479136557, "speed/FLOPS": 203014521344048.28, "speed/curr_iter_time": 1.2781, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.8459827899932861, "optim/lr": 0.0019890000000000003, "optim/total_tokens": 869269504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 5.109765529632568, "created_at": "2025-01-14T17:40:58.971386+00:00"} {"global_step": 3317, "acc_step": 0, "speed/wps": 12927.547663177334, "speed/FLOPS": 203044900362256.1, "speed/curr_iter_time": 1.2781, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1120842695236206, "optim/lr": 0.0019896000000000002, "optim/total_tokens": 869531648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 5.183379173278809, "created_at": "2025-01-14T17:41:04.046823+00:00"} {"global_step": 3318, "acc_step": 0, "speed/wps": 12921.913657437486, "speed/FLOPS": 202956410560176.6, "speed/curr_iter_time": 1.2789, "speed/data_load_time": 0.0011, "optim/grad_norm": 1.0437519550323486, "optim/lr": 0.0019902, "optim/total_tokens": 869793792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459743, "loss/out": 5.128094673156738, "created_at": "2025-01-14T17:41:09.122226+00:00"} {"global_step": 3319, "acc_step": 0, "speed/wps": 12928.867209571805, "speed/FLOPS": 203065625651628.03, "speed/curr_iter_time": 1.2781, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9949962496757507, "optim/lr": 0.0019908, "optim/total_tokens": 870055936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309558, "loss/out": 5.178269386291504, "created_at": "2025-01-14T17:41:14.195627+00:00"} {"global_step": 3320, "acc_step": 0, "speed/wps": 12925.662390804691, "speed/FLOPS": 203015289569008.03, "speed/curr_iter_time": 1.2788, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.036366581916809, "optim/lr": 0.0019914, "optim/total_tokens": 870318080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 5.1209869384765625, "created_at": "2025-01-14T17:41:19.269877+00:00"} {"global_step": 3321, "acc_step": 0, "speed/wps": 12925.679143510075, "speed/FLOPS": 203015552693266.72, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2650316953659058, "optim/lr": 0.0019920000000000003, "optim/total_tokens": 870580224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 5.208123207092285, "created_at": "2025-01-14T17:41:24.346547+00:00"} {"global_step": 3322, "acc_step": 0, "speed/wps": 12926.489519639454, "speed/FLOPS": 203028280763952.1, "speed/curr_iter_time": 1.2779, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9385865330696106, "optim/lr": 0.0019926, "optim/total_tokens": 870842368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437432, "loss/out": 5.229007720947266, "created_at": "2025-01-14T17:41:29.431013+00:00"} {"global_step": 3323, "acc_step": 0, "speed/wps": 12924.474087567842, "speed/FLOPS": 202996625633773.53, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0743693113327026, "optim/lr": 0.0019932, "optim/total_tokens": 871104512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 5.185500144958496, "created_at": "2025-01-14T17:41:34.505210+00:00"} {"global_step": 3324, "acc_step": 0, "speed/wps": 12925.811780124308, "speed/FLOPS": 203017635933555.03, "speed/curr_iter_time": 1.2781, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1451945304870605, "optim/lr": 0.0019938, "optim/total_tokens": 871366656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451797, "loss/out": 5.168974876403809, "created_at": "2025-01-14T17:41:39.584492+00:00"} {"global_step": 3325, "acc_step": 0, "speed/wps": 12920.678571119832, "speed/FLOPS": 202937011832371.4, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1362884044647217, "optim/lr": 0.0019944, "optim/total_tokens": 871628800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328746, "loss/out": 5.215103626251221, "created_at": "2025-01-14T17:41:44.665068+00:00"} {"global_step": 3326, "acc_step": 0, "speed/wps": 12925.81464115817, "speed/FLOPS": 203017680870023.3, "speed/curr_iter_time": 1.2789, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1078615188598633, "optim/lr": 0.0019950000000000002, "optim/total_tokens": 871890944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 5.1880574226379395, "created_at": "2025-01-14T17:41:49.740580+00:00"} {"global_step": 3327, "acc_step": 0, "speed/wps": 12925.371504663532, "speed/FLOPS": 203010720802441.66, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9065707325935364, "optim/lr": 0.0019956, "optim/total_tokens": 872153088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450746, "loss/out": 5.146655559539795, "created_at": "2025-01-14T17:41:54.821575+00:00"} {"global_step": 3328, "acc_step": 0, "speed/wps": 12916.943489778818, "speed/FLOPS": 202878347247371.16, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1629234552383423, "optim/lr": 0.0019962, "optim/total_tokens": 872415232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315474, "loss/out": 5.158446311950684, "created_at": "2025-01-14T17:41:59.900547+00:00"} {"global_step": 3329, "acc_step": 0, "speed/wps": 12921.59222940093, "speed/FLOPS": 202951362091174.1, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2205238342285156, "optim/lr": 0.0019968, "optim/total_tokens": 872677376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460271, "loss/out": 5.157448768615723, "created_at": "2025-01-14T17:42:04.977056+00:00"} {"global_step": 3330, "acc_step": 0, "speed/wps": 12926.197221466544, "speed/FLOPS": 203023689819486.78, "speed/curr_iter_time": 1.2782, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0582184791564941, "optim/lr": 0.0019974, "optim/total_tokens": 872939520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456330, "loss/out": 5.191284656524658, "created_at": "2025-01-14T17:42:10.052036+00:00"} {"global_step": 3331, "acc_step": 0, "speed/wps": 12930.235543993307, "speed/FLOPS": 203087117224006.25, "speed/curr_iter_time": 1.2782, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9622957110404968, "optim/lr": 0.001998, "optim/total_tokens": 873201664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443306, "loss/out": 4.997122764587402, "created_at": "2025-01-14T17:42:15.124130+00:00"} {"global_step": 3332, "acc_step": 0, "speed/wps": 12922.052497901353, "speed/FLOPS": 202958591240448.44, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.17025625705719, "optim/lr": 0.0019986, "optim/total_tokens": 873463808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430506, "loss/out": 5.169713973999023, "created_at": "2025-01-14T17:42:20.201665+00:00"} {"global_step": 3333, "acc_step": 0, "speed/wps": 12931.751161447173, "speed/FLOPS": 203110922078795.84, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1029688119888306, "optim/lr": 0.0019992, "optim/total_tokens": 873725952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483664, "loss/out": 5.195382118225098, "created_at": "2025-01-14T17:42:25.273272+00:00"} {"global_step": 3334, "acc_step": 0, "speed/wps": 12922.171555542194, "speed/FLOPS": 202960461204300.2, "speed/curr_iter_time": 1.2782, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0912550687789917, "optim/lr": 0.0019998, "optim/total_tokens": 873988096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439586, "loss/out": 5.088690757751465, "created_at": "2025-01-14T17:42:30.349379+00:00"} {"global_step": 3335, "acc_step": 0, "speed/wps": 12928.442880121931, "speed/FLOPS": 203058960974528.0, "speed/curr_iter_time": 1.2784, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2746964693069458, "optim/lr": 0.0020004, "optim/total_tokens": 874250240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 5.137495040893555, "created_at": "2025-01-14T17:42:35.426973+00:00"} {"global_step": 3336, "acc_step": 0, "speed/wps": 12933.660052067, "speed/FLOPS": 203140903829070.8, "speed/curr_iter_time": 1.2774, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1225776672363281, "optim/lr": 0.002001, "optim/total_tokens": 874512384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361335, "loss/out": 5.232653617858887, "created_at": "2025-01-14T17:42:40.501924+00:00"} {"global_step": 3337, "acc_step": 0, "speed/wps": 12927.457297372679, "speed/FLOPS": 203043481043118.38, "speed/curr_iter_time": 1.2785, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.281583547592163, "optim/lr": 0.0020016, "optim/total_tokens": 874774528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 5.144375324249268, "created_at": "2025-01-14T17:42:45.574981+00:00"} {"global_step": 3338, "acc_step": 0, "speed/wps": 12929.658283682533, "speed/FLOPS": 203078050557586.75, "speed/curr_iter_time": 1.2778, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0310331583023071, "optim/lr": 0.0020022, "optim/total_tokens": 875036672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454182, "loss/out": 5.088800430297852, "created_at": "2025-01-14T17:42:50.650176+00:00"} {"global_step": 3339, "acc_step": 0, "speed/wps": 12923.276172294944, "speed/FLOPS": 202977810728304.6, "speed/curr_iter_time": 1.2794, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9829223155975342, "optim/lr": 0.0020028, "optim/total_tokens": 875298816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325819, "loss/out": 5.328011989593506, "created_at": "2025-01-14T17:42:55.725099+00:00"} {"global_step": 3340, "acc_step": 0, "speed/wps": 12929.084680926142, "speed/FLOPS": 203069041338084.84, "speed/curr_iter_time": 1.2774, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9232273697853088, "optim/lr": 0.0020034, "optim/total_tokens": 875560960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468245, "loss/out": 5.009457588195801, "created_at": "2025-01-14T17:43:00.800020+00:00"} {"global_step": 3341, "acc_step": 0, "speed/wps": 12932.022544969066, "speed/FLOPS": 203115184529927.22, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7639637589454651, "optim/lr": 0.002004, "optim/total_tokens": 875823104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459765, "loss/out": 5.12410831451416, "created_at": "2025-01-14T17:43:05.875455+00:00"} {"global_step": 3342, "acc_step": 0, "speed/wps": 12925.16859250383, "speed/FLOPS": 203007533788143.0, "speed/curr_iter_time": 1.2788, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8521656394004822, "optim/lr": 0.0020046, "optim/total_tokens": 876085248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442784, "loss/out": 5.09914493560791, "created_at": "2025-01-14T17:43:10.950385+00:00"} {"global_step": 3343, "acc_step": 0, "speed/wps": 12927.01862437389, "speed/FLOPS": 203036591080872.5, "speed/curr_iter_time": 1.2781, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8582857251167297, "optim/lr": 0.0020052, "optim/total_tokens": 876347392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331148, "loss/out": 5.101330757141113, "created_at": "2025-01-14T17:43:16.026334+00:00"} {"global_step": 3344, "acc_step": 0, "speed/wps": 12919.242783198208, "speed/FLOPS": 202914460810082.94, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0081156492233276, "optim/lr": 0.0020058, "optim/total_tokens": 876609536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 5.108908653259277, "created_at": "2025-01-14T17:43:21.102669+00:00"} {"global_step": 3345, "acc_step": 0, "speed/wps": 12936.33609946028, "speed/FLOPS": 203182934830656.78, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0217068195343018, "optim/lr": 0.0020063999999999998, "optim/total_tokens": 876871680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466369, "loss/out": 5.030872344970703, "created_at": "2025-01-14T17:43:26.174804+00:00"} {"global_step": 3346, "acc_step": 0, "speed/wps": 12928.30185724088, "speed/FLOPS": 203056746016395.34, "speed/curr_iter_time": 1.2783, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0583534240722656, "optim/lr": 0.002007, "optim/total_tokens": 877133824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 5.0249786376953125, "created_at": "2025-01-14T17:43:31.248100+00:00"} {"global_step": 3347, "acc_step": 0, "speed/wps": 12927.217733170315, "speed/FLOPS": 203039718358124.53, "speed/curr_iter_time": 1.2774, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4129587411880493, "optim/lr": 0.0020076, "optim/total_tokens": 877395968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309025, "loss/out": 5.111201763153076, "created_at": "2025-01-14T17:43:36.335954+00:00"} {"global_step": 3348, "acc_step": 0, "speed/wps": 12931.458062071282, "speed/FLOPS": 203106318550336.75, "speed/curr_iter_time": 1.2778, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.092414379119873, "optim/lr": 0.0020082, "optim/total_tokens": 877658112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 5.075135231018066, "created_at": "2025-01-14T17:43:41.408596+00:00"} {"global_step": 3349, "acc_step": 0, "speed/wps": 12906.502664288088, "speed/FLOPS": 202714359735841.53, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.103259801864624, "optim/lr": 0.0020088, "optim/total_tokens": 877920256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 5.230422019958496, "created_at": "2025-01-14T17:43:46.498244+00:00"} {"global_step": 3350, "acc_step": 0, "speed/wps": 12927.778347512543, "speed/FLOPS": 203048523576734.06, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0160424709320068, "optim/lr": 0.0020093999999999997, "optim/total_tokens": 878182400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444067, "loss/out": 5.067336082458496, "created_at": "2025-01-14T17:43:51.575664+00:00"} {"global_step": 3351, "acc_step": 0, "speed/wps": 12919.80323976114, "speed/FLOPS": 202923263550551.8, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0014, "optim/grad_norm": 1.1747978925704956, "optim/lr": 0.00201, "optim/total_tokens": 878444544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441158, "loss/out": 5.111091136932373, "created_at": "2025-01-14T17:43:56.658608+00:00"} {"global_step": 3352, "acc_step": 0, "speed/wps": 12931.3780190878, "speed/FLOPS": 203105061365290.56, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8745715022087097, "optim/lr": 0.0020106, "optim/total_tokens": 878706688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438251, "loss/out": 5.082679748535156, "created_at": "2025-01-14T17:44:01.737264+00:00"} {"global_step": 3353, "acc_step": 0, "speed/wps": 12931.034349070911, "speed/FLOPS": 203099663555423.22, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1617674827575684, "optim/lr": 0.0020112, "optim/total_tokens": 878968832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442200, "loss/out": 5.205699920654297, "created_at": "2025-01-14T17:44:06.811555+00:00"} {"global_step": 3354, "acc_step": 0, "speed/wps": 12924.043094001914, "speed/FLOPS": 202989856287573.44, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0239022970199585, "optim/lr": 0.0020118, "optim/total_tokens": 879230976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453945, "loss/out": 5.054997444152832, "created_at": "2025-01-14T17:44:11.890632+00:00"} {"global_step": 3355, "acc_step": 0, "speed/wps": 12924.393848539497, "speed/FLOPS": 202995365369573.4, "speed/curr_iter_time": 1.2785, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8545280694961548, "optim/lr": 0.0020123999999999997, "optim/total_tokens": 879493120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450462, "loss/out": 5.055810928344727, "created_at": "2025-01-14T17:44:16.967447+00:00"} {"global_step": 3356, "acc_step": 0, "speed/wps": 12920.58140955311, "speed/FLOPS": 202935485776452.03, "speed/curr_iter_time": 1.2801, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9778376817703247, "optim/lr": 0.002013, "optim/total_tokens": 879755264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364786, "loss/out": 5.051249027252197, "created_at": "2025-01-14T17:44:22.046456+00:00"} {"global_step": 3357, "acc_step": 0, "speed/wps": 12923.806058368746, "speed/FLOPS": 202986133317233.4, "speed/curr_iter_time": 1.2784, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0823181867599487, "optim/lr": 0.0020136, "optim/total_tokens": 880017408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 5.048698902130127, "created_at": "2025-01-14T17:44:27.124936+00:00"} {"global_step": 3358, "acc_step": 0, "speed/wps": 12925.728950775278, "speed/FLOPS": 203016334984809.06, "speed/curr_iter_time": 1.2789, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0394256114959717, "optim/lr": 0.0020142, "optim/total_tokens": 880279552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445170, "loss/out": 5.0417160987854, "created_at": "2025-01-14T17:44:32.206880+00:00"} {"global_step": 3359, "acc_step": 0, "speed/wps": 12929.122810252704, "speed/FLOPS": 203069640211553.7, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9906079173088074, "optim/lr": 0.0020147999999999998, "optim/total_tokens": 880541696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434515, "loss/out": 5.205700874328613, "created_at": "2025-01-14T17:44:37.280557+00:00"} {"global_step": 3360, "acc_step": 0, "speed/wps": 12924.695311458327, "speed/FLOPS": 203000100258967.1, "speed/curr_iter_time": 1.2785, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7924551963806152, "optim/lr": 0.0020154, "optim/total_tokens": 880803840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435044, "loss/out": 5.118573188781738, "created_at": "2025-01-14T17:44:42.358730+00:00"} {"global_step": 3361, "acc_step": 0, "speed/wps": 12924.212010012536, "speed/FLOPS": 202992509345635.16, "speed/curr_iter_time": 1.2784, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0212594270706177, "optim/lr": 0.002016, "optim/total_tokens": 881065984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441158, "loss/out": 4.980246543884277, "created_at": "2025-01-14T17:44:47.438350+00:00"} {"global_step": 3362, "acc_step": 0, "speed/wps": 12928.502107704053, "speed/FLOPS": 203059891225092.44, "speed/curr_iter_time": 1.2783, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.339761257171631, "optim/lr": 0.0020166, "optim/total_tokens": 881328128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 5.119205474853516, "created_at": "2025-01-14T17:44:52.511507+00:00"} {"global_step": 3363, "acc_step": 0, "speed/wps": 12926.83032410127, "speed/FLOPS": 203033633566339.34, "speed/curr_iter_time": 1.2786, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2069605588912964, "optim/lr": 0.0020172000000000002, "optim/total_tokens": 881590272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 407007, "loss/out": 5.109159469604492, "created_at": "2025-01-14T17:44:57.587040+00:00"} {"global_step": 3364, "acc_step": 0, "speed/wps": 12923.904163287212, "speed/FLOPS": 202987674189788.22, "speed/curr_iter_time": 1.2789, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.0780115127563477, "optim/lr": 0.0020178, "optim/total_tokens": 881852416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 5.0565080642700195, "created_at": "2025-01-14T17:45:02.664855+00:00"} {"global_step": 3365, "acc_step": 0, "speed/wps": 12925.525566798498, "speed/FLOPS": 203013140559974.88, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0011, "optim/grad_norm": 1.141379475593567, "optim/lr": 0.0020184, "optim/total_tokens": 882114560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 5.089571952819824, "created_at": "2025-01-14T17:45:07.768149+00:00"} {"global_step": 3366, "acc_step": 0, "speed/wps": 12925.23307718313, "speed/FLOPS": 203008546608642.66, "speed/curr_iter_time": 1.279, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1124260425567627, "optim/lr": 0.002019, "optim/total_tokens": 882376704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 5.038433074951172, "created_at": "2025-01-14T17:45:12.847383+00:00"} {"global_step": 3367, "acc_step": 0, "speed/wps": 12920.899387512953, "speed/FLOPS": 202940480057258.88, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1284486055374146, "optim/lr": 0.0020196000000000003, "optim/total_tokens": 882638848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480426, "loss/out": 5.187658309936523, "created_at": "2025-01-14T17:45:17.927763+00:00"} {"global_step": 3368, "acc_step": 0, "speed/wps": 12922.403626126501, "speed/FLOPS": 202964106191725.12, "speed/curr_iter_time": 1.2796, "speed/data_load_time": 0.002, "optim/grad_norm": 1.2657424211502075, "optim/lr": 0.0020202, "optim/total_tokens": 882900992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324322, "loss/out": 5.104199409484863, "created_at": "2025-01-14T17:45:23.004112+00:00"} {"global_step": 3369, "acc_step": 0, "speed/wps": 12916.76759623226, "speed/FLOPS": 202875584597519.38, "speed/curr_iter_time": 1.2785, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1325314044952393, "optim/lr": 0.0020208, "optim/total_tokens": 883163136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 5.17860746383667, "created_at": "2025-01-14T17:45:28.081701+00:00"} {"global_step": 3370, "acc_step": 0, "speed/wps": 12917.412373750183, "speed/FLOPS": 202885711714455.53, "speed/curr_iter_time": 1.2788, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.0169250965118408, "optim/lr": 0.0020214, "optim/total_tokens": 883425280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 5.041587829589844, "created_at": "2025-01-14T17:45:33.163955+00:00"} {"global_step": 3371, "acc_step": 0, "speed/wps": 12929.30259195288, "speed/FLOPS": 203072463930200.9, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.251105546951294, "optim/lr": 0.0020220000000000004, "optim/total_tokens": 883687424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 4.996942043304443, "created_at": "2025-01-14T17:45:38.236422+00:00"} {"global_step": 3372, "acc_step": 0, "speed/wps": 12933.327463834015, "speed/FLOPS": 203135680073847.7, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4689745903015137, "optim/lr": 0.0020226000000000003, "optim/total_tokens": 883949568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362184, "loss/out": 5.044646263122559, "created_at": "2025-01-14T17:45:43.307194+00:00"} {"global_step": 3373, "acc_step": 0, "speed/wps": 12925.846518621944, "speed/FLOPS": 203018181549396.12, "speed/curr_iter_time": 1.2783, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2271957397460938, "optim/lr": 0.0020232, "optim/total_tokens": 884211712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 4.904811859130859, "created_at": "2025-01-14T17:45:48.381033+00:00"} {"global_step": 3374, "acc_step": 0, "speed/wps": 12930.795830837464, "speed/FLOPS": 203095917298808.53, "speed/curr_iter_time": 1.2784, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1800059080123901, "optim/lr": 0.0020238, "optim/total_tokens": 884473856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443307, "loss/out": 5.017938613891602, "created_at": "2025-01-14T17:45:53.454408+00:00"} {"global_step": 3375, "acc_step": 0, "speed/wps": 12927.706859686856, "speed/FLOPS": 203047400762197.06, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1277921199798584, "optim/lr": 0.0020244, "optim/total_tokens": 884736000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428932, "loss/out": 5.008646488189697, "created_at": "2025-01-14T17:45:58.527729+00:00"} {"global_step": 3376, "acc_step": 0, "speed/wps": 12928.621440994395, "speed/FLOPS": 203061765518398.38, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4444012641906738, "optim/lr": 0.0020250000000000003, "optim/total_tokens": 884998144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 4.9989013671875, "created_at": "2025-01-14T17:46:03.600371+00:00"} {"global_step": 3377, "acc_step": 0, "speed/wps": 12930.900989825655, "speed/FLOPS": 203097568965221.7, "speed/curr_iter_time": 1.2781, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1427805423736572, "optim/lr": 0.0020256000000000002, "optim/total_tokens": 885260288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 4.962312698364258, "created_at": "2025-01-14T17:46:08.671981+00:00"} {"global_step": 3378, "acc_step": 0, "speed/wps": 12929.79235853461, "speed/FLOPS": 203080156387376.78, "speed/curr_iter_time": 1.2771, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9487513303756714, "optim/lr": 0.0020262, "optim/total_tokens": 885522432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 5.001970291137695, "created_at": "2025-01-14T17:46:13.748810+00:00"} {"global_step": 3379, "acc_step": 0, "speed/wps": 12927.332340526864, "speed/FLOPS": 203041518424146.12, "speed/curr_iter_time": 1.2782, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1126378774642944, "optim/lr": 0.0020268, "optim/total_tokens": 885784576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316000, "loss/out": 5.009018898010254, "created_at": "2025-01-14T17:46:18.828276+00:00"} {"global_step": 3380, "acc_step": 0, "speed/wps": 12920.138045864336, "speed/FLOPS": 202928522140474.28, "speed/curr_iter_time": 1.2785, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0739208459854126, "optim/lr": 0.0020274, "optim/total_tokens": 886046720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 4.944303512573242, "created_at": "2025-01-14T17:46:23.905466+00:00"} {"global_step": 3381, "acc_step": 0, "speed/wps": 12934.053402660631, "speed/FLOPS": 203147081940664.16, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8753260374069214, "optim/lr": 0.0020280000000000003, "optim/total_tokens": 886308864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447263, "loss/out": 5.023478031158447, "created_at": "2025-01-14T17:46:28.975958+00:00"} {"global_step": 3382, "acc_step": 0, "speed/wps": 12927.498813237646, "speed/FLOPS": 203044133106826.44, "speed/curr_iter_time": 1.2782, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.055533528327942, "optim/lr": 0.0020286, "optim/total_tokens": 886571008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 4.946606159210205, "created_at": "2025-01-14T17:46:34.052533+00:00"} {"global_step": 3383, "acc_step": 0, "speed/wps": 12928.23140633381, "speed/FLOPS": 203055639488090.12, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0269763469696045, "optim/lr": 0.0020292, "optim/total_tokens": 886833152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296822, "loss/out": 4.969162940979004, "created_at": "2025-01-14T17:46:39.126786+00:00"} {"global_step": 3384, "acc_step": 0, "speed/wps": 12942.524408109864, "speed/FLOPS": 203280130721625.7, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.814213216304779, "optim/lr": 0.0020298, "optim/total_tokens": 887095296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364683, "loss/out": 4.926223278045654, "created_at": "2025-01-14T17:46:44.198862+00:00"} {"global_step": 3385, "acc_step": 0, "speed/wps": 12932.504617284087, "speed/FLOPS": 203122756138071.4, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0238242149353027, "optim/lr": 0.0020304, "optim/total_tokens": 887357440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 5.1048808097839355, "created_at": "2025-01-14T17:46:49.270190+00:00"} {"global_step": 3386, "acc_step": 0, "speed/wps": 12923.909802964881, "speed/FLOPS": 202987762768675.75, "speed/curr_iter_time": 1.2785, "speed/data_load_time": 0.0011, "optim/grad_norm": 1.1044611930847168, "optim/lr": 0.0020310000000000003, "optim/total_tokens": 887619584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 5.046362400054932, "created_at": "2025-01-14T17:46:54.344638+00:00"} {"global_step": 3387, "acc_step": 0, "speed/wps": 12931.015201514834, "speed/FLOPS": 203099362816743.53, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3339115381240845, "optim/lr": 0.0020316, "optim/total_tokens": 887881728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327927, "loss/out": 4.961957931518555, "created_at": "2025-01-14T17:46:59.416265+00:00"} {"global_step": 3388, "acc_step": 0, "speed/wps": 12929.513915485002, "speed/FLOPS": 203075783056661.84, "speed/curr_iter_time": 1.2782, "speed/data_load_time": 0.001, "optim/grad_norm": 0.855763852596283, "optim/lr": 0.0020322, "optim/total_tokens": 888143872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 4.878522872924805, "created_at": "2025-01-14T17:47:04.489808+00:00"} {"global_step": 3389, "acc_step": 0, "speed/wps": 12933.562647622357, "speed/FLOPS": 203139373958419.97, "speed/curr_iter_time": 1.2781, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0249682664871216, "optim/lr": 0.0020328, "optim/total_tokens": 888406016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 5.126818656921387, "created_at": "2025-01-14T17:47:09.590403+00:00"} {"global_step": 3390, "acc_step": 0, "speed/wps": 12926.16983944745, "speed/FLOPS": 203023259747250.06, "speed/curr_iter_time": 1.2786, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.141236662864685, "optim/lr": 0.0020334, "optim/total_tokens": 888668160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337583, "loss/out": 4.994195938110352, "created_at": "2025-01-14T17:47:14.664119+00:00"} {"global_step": 3391, "acc_step": 0, "speed/wps": 12929.207498023212, "speed/FLOPS": 203070970349362.72, "speed/curr_iter_time": 1.2783, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3199093341827393, "optim/lr": 0.0020340000000000002, "optim/total_tokens": 888930304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 5.021871089935303, "created_at": "2025-01-14T17:47:19.744502+00:00"} {"global_step": 3392, "acc_step": 0, "speed/wps": 12919.038047376023, "speed/FLOPS": 202911245152659.03, "speed/curr_iter_time": 1.2803, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1841161251068115, "optim/lr": 0.0020346, "optim/total_tokens": 889192448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 5.0037078857421875, "created_at": "2025-01-14T17:47:24.823489+00:00"} {"global_step": 3393, "acc_step": 0, "speed/wps": 12927.96567786633, "speed/FLOPS": 203051465857359.16, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1258878707885742, "optim/lr": 0.0020352, "optim/total_tokens": 889454592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302343, "loss/out": 5.047472953796387, "created_at": "2025-01-14T17:47:29.897197+00:00"} {"global_step": 3394, "acc_step": 0, "speed/wps": 12920.664635563266, "speed/FLOPS": 202936792955306.06, "speed/curr_iter_time": 1.2785, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1807570457458496, "optim/lr": 0.0020358, "optim/total_tokens": 889716736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 4.994605541229248, "created_at": "2025-01-14T17:47:34.973381+00:00"} {"global_step": 3395, "acc_step": 0, "speed/wps": 12926.074663376796, "speed/FLOPS": 203021764876274.8, "speed/curr_iter_time": 1.2781, "speed/data_load_time": 0.001, "optim/grad_norm": 1.0751653909683228, "optim/lr": 0.0020364, "optim/total_tokens": 889978880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443307, "loss/out": 4.9686174392700195, "created_at": "2025-01-14T17:47:40.047204+00:00"} {"global_step": 3396, "acc_step": 0, "speed/wps": 12929.820036244673, "speed/FLOPS": 203080591103846.47, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9147518873214722, "optim/lr": 0.002037, "optim/total_tokens": 890241024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456330, "loss/out": 4.9750566482543945, "created_at": "2025-01-14T17:47:45.120127+00:00"} {"global_step": 3397, "acc_step": 0, "speed/wps": 12933.306155873695, "speed/FLOPS": 203135345403050.66, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1090041399002075, "optim/lr": 0.0020376, "optim/total_tokens": 890503168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327688, "loss/out": 4.893564224243164, "created_at": "2025-01-14T17:47:50.193057+00:00"} {"global_step": 3398, "acc_step": 0, "speed/wps": 12929.420837729838, "speed/FLOPS": 203074321142614.75, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2234238386154175, "optim/lr": 0.0020382, "optim/total_tokens": 890765312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463218, "loss/out": 4.937596797943115, "created_at": "2025-01-14T17:47:55.273176+00:00"} {"global_step": 3399, "acc_step": 0, "speed/wps": 12922.480978512085, "speed/FLOPS": 202965321117234.03, "speed/curr_iter_time": 1.2798, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9227579832077026, "optim/lr": 0.0020388, "optim/total_tokens": 891027456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460287, "loss/out": 4.971157073974609, "created_at": "2025-01-14T17:48:00.352703+00:00"} {"global_step": 3400, "acc_step": 0, "speed/wps": 12927.173946424866, "speed/FLOPS": 203039030627117.97, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.157476782798767, "optim/lr": 0.0020394, "optim/total_tokens": 891289600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445407, "loss/out": 4.977930068969727, "created_at": "2025-01-14T17:48:05.426970+00:00"} {"global_step": 3401, "acc_step": 0, "speed/wps": 12929.15506036517, "speed/FLOPS": 203070146743886.28, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.022200345993042, "optim/lr": 0.00204, "optim/total_tokens": 891551744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 373517, "loss/out": 4.907966613769531, "created_at": "2025-01-14T17:48:10.508912+00:00"} {"global_step": 3402, "acc_step": 0, "speed/wps": 12929.76884564088, "speed/FLOPS": 203079787085070.62, "speed/curr_iter_time": 1.2787, "speed/data_load_time": 0.001, "optim/grad_norm": 0.9424383640289307, "optim/lr": 0.0020406, "optim/total_tokens": 891813888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 4.949992656707764, "created_at": "2025-01-14T17:48:15.582424+00:00"} {"global_step": 3403, "acc_step": 0, "speed/wps": 12927.72106456363, "speed/FLOPS": 203047623869306.34, "speed/curr_iter_time": 1.2778, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8129037022590637, "optim/lr": 0.0020412, "optim/total_tokens": 892076032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 4.921763896942139, "created_at": "2025-01-14T17:48:20.656619+00:00"} {"global_step": 3404, "acc_step": 0, "speed/wps": 12920.849515444062, "speed/FLOPS": 202939696747885.56, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.007165551185608, "optim/lr": 0.0020418, "optim/total_tokens": 892338176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359885, "loss/out": 4.99824333190918, "created_at": "2025-01-14T17:48:25.732297+00:00"} {"global_step": 3405, "acc_step": 0, "speed/wps": 12926.506661541694, "speed/FLOPS": 203028550001082.22, "speed/curr_iter_time": 1.2786, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7971799969673157, "optim/lr": 0.0020423999999999998, "optim/total_tokens": 892600320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 4.891290664672852, "created_at": "2025-01-14T17:48:30.806749+00:00"} {"global_step": 3406, "acc_step": 0, "speed/wps": 12921.609972861272, "speed/FLOPS": 202951640776600.8, "speed/curr_iter_time": 1.2782, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0623897314071655, "optim/lr": 0.002043, "optim/total_tokens": 892862464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 4.912031173706055, "created_at": "2025-01-14T17:48:35.882148+00:00"} {"global_step": 3407, "acc_step": 0, "speed/wps": 12926.651575239288, "speed/FLOPS": 203030826069835.22, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.099815845489502, "optim/lr": 0.0020436, "optim/total_tokens": 893124608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322311, "loss/out": 4.865272521972656, "created_at": "2025-01-14T17:48:40.960570+00:00"} {"global_step": 3408, "acc_step": 0, "speed/wps": 12923.549751700597, "speed/FLOPS": 202982107668807.88, "speed/curr_iter_time": 1.2782, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.13044273853302, "optim/lr": 0.0020442, "optim/total_tokens": 893386752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 4.950529098510742, "created_at": "2025-01-14T17:48:46.039881+00:00"} {"global_step": 3409, "acc_step": 0, "speed/wps": 12930.802091950412, "speed/FLOPS": 203096015638191.06, "speed/curr_iter_time": 1.2785, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7521767616271973, "optim/lr": 0.0020448, "optim/total_tokens": 893648896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434294, "loss/out": 4.961922645568848, "created_at": "2025-01-14T17:48:51.112727+00:00"} {"global_step": 3410, "acc_step": 0, "speed/wps": 12928.390111380564, "speed/FLOPS": 203058132168932.97, "speed/curr_iter_time": 1.2779, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9468083381652832, "optim/lr": 0.0020453999999999997, "optim/total_tokens": 893911040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 5.003101825714111, "created_at": "2025-01-14T17:48:56.185454+00:00"} {"global_step": 3411, "acc_step": 0, "speed/wps": 12933.437071151693, "speed/FLOPS": 203137401607388.03, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9082731008529663, "optim/lr": 0.002046, "optim/total_tokens": 894173184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323796, "loss/out": 4.985447883605957, "created_at": "2025-01-14T17:49:01.257893+00:00"} {"global_step": 3412, "acc_step": 0, "speed/wps": 12928.33185629249, "speed/FLOPS": 203057217192723.9, "speed/curr_iter_time": 1.2787, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7716344594955444, "optim/lr": 0.0020466, "optim/total_tokens": 894435328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435320, "loss/out": 4.909121990203857, "created_at": "2025-01-14T17:49:06.330476+00:00"} {"global_step": 3413, "acc_step": 0, "speed/wps": 12928.126523525973, "speed/FLOPS": 203053992159468.94, "speed/curr_iter_time": 1.2779, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9727684855461121, "optim/lr": 0.0020472, "optim/total_tokens": 894697472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 4.985270977020264, "created_at": "2025-01-14T17:49:11.404358+00:00"} {"global_step": 3414, "acc_step": 0, "speed/wps": 12930.909220888772, "speed/FLOPS": 203097698245378.6, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8581885099411011, "optim/lr": 0.0020478, "optim/total_tokens": 894959616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 4.908967018127441, "created_at": "2025-01-14T17:49:16.477890+00:00"} {"global_step": 3415, "acc_step": 0, "speed/wps": 12932.812802569137, "speed/FLOPS": 203127596611464.06, "speed/curr_iter_time": 1.2778, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8988066911697388, "optim/lr": 0.0020484, "optim/total_tokens": 895221760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310904, "loss/out": 4.9866623878479, "created_at": "2025-01-14T17:49:21.550887+00:00"} {"global_step": 3416, "acc_step": 0, "speed/wps": 12929.932211772002, "speed/FLOPS": 203082352974648.5, "speed/curr_iter_time": 1.2779, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9573215246200562, "optim/lr": 0.002049, "optim/total_tokens": 895483904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 4.946046352386475, "created_at": "2025-01-14T17:49:26.624575+00:00"} {"global_step": 3417, "acc_step": 0, "speed/wps": 12926.256020066921, "speed/FLOPS": 203024613332303.44, "speed/curr_iter_time": 1.2787, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9510943293571472, "optim/lr": 0.0020496, "optim/total_tokens": 895746048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435329, "loss/out": 4.9198713302612305, "created_at": "2025-01-14T17:49:31.698194+00:00"} {"global_step": 3418, "acc_step": 0, "speed/wps": 12920.977666156265, "speed/FLOPS": 202941709530918.22, "speed/curr_iter_time": 1.2785, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0079777240753174, "optim/lr": 0.0020502, "optim/total_tokens": 896008192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437717, "loss/out": 4.975019454956055, "created_at": "2025-01-14T17:49:36.774975+00:00"} {"global_step": 3419, "acc_step": 0, "speed/wps": 12922.187663243936, "speed/FLOPS": 202960714197890.34, "speed/curr_iter_time": 1.2783, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9920469522476196, "optim/lr": 0.0020508, "optim/total_tokens": 896270336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446503, "loss/out": 4.9685797691345215, "created_at": "2025-01-14T17:49:41.850016+00:00"} {"global_step": 3420, "acc_step": 0, "speed/wps": 12935.318261229915, "speed/FLOPS": 203166948282592.56, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.262882947921753, "optim/lr": 0.0020514, "optim/total_tokens": 896532480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 5.057164192199707, "created_at": "2025-01-14T17:49:46.920166+00:00"} {"global_step": 3421, "acc_step": 0, "speed/wps": 12923.832281564111, "speed/FLOPS": 202986545188551.1, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.1173174381256104, "optim/lr": 0.002052, "optim/total_tokens": 896794624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 4.796419620513916, "created_at": "2025-01-14T17:49:52.000059+00:00"} {"global_step": 3422, "acc_step": 0, "speed/wps": 12926.220310591874, "speed/FLOPS": 203024052465927.88, "speed/curr_iter_time": 1.2785, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0037243366241455, "optim/lr": 0.0020526000000000003, "optim/total_tokens": 897056768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311191, "loss/out": 4.866168022155762, "created_at": "2025-01-14T17:49:57.075812+00:00"} {"global_step": 3423, "acc_step": 0, "speed/wps": 12929.26200760255, "speed/FLOPS": 203071826497210.22, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9292012453079224, "optim/lr": 0.0020532000000000003, "optim/total_tokens": 897318912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 4.881525039672852, "created_at": "2025-01-14T17:50:02.148196+00:00"} {"global_step": 3424, "acc_step": 0, "speed/wps": 12928.676453957925, "speed/FLOPS": 203062629572586.38, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9692018628120422, "optim/lr": 0.0020538, "optim/total_tokens": 897581056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 4.897002696990967, "created_at": "2025-01-14T17:50:07.227504+00:00"} {"global_step": 3425, "acc_step": 0, "speed/wps": 12934.36201838351, "speed/FLOPS": 203151929174674.4, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.03976309299469, "optim/lr": 0.0020544, "optim/total_tokens": 897843200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 4.748776435852051, "created_at": "2025-01-14T17:50:12.297857+00:00"} {"global_step": 3426, "acc_step": 0, "speed/wps": 12929.439289254895, "speed/FLOPS": 203074610949170.8, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9547758102416992, "optim/lr": 0.0020550000000000004, "optim/total_tokens": 898105344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317059, "loss/out": 4.853793621063232, "created_at": "2025-01-14T17:50:17.374316+00:00"} {"global_step": 3427, "acc_step": 0, "speed/wps": 12927.355770938882, "speed/FLOPS": 203041886430963.66, "speed/curr_iter_time": 1.2788, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.8657758831977844, "optim/lr": 0.0020556000000000003, "optim/total_tokens": 898367488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 4.962162971496582, "created_at": "2025-01-14T17:50:22.449572+00:00"} {"global_step": 3428, "acc_step": 0, "speed/wps": 12934.244141048981, "speed/FLOPS": 203150077749155.22, "speed/curr_iter_time": 1.2781, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0482685565948486, "optim/lr": 0.0020562, "optim/total_tokens": 898629632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446220, "loss/out": 5.037456035614014, "created_at": "2025-01-14T17:50:27.521490+00:00"} {"global_step": 3429, "acc_step": 0, "speed/wps": 12929.32464691272, "speed/FLOPS": 203072810333651.88, "speed/curr_iter_time": 1.2779, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1211010217666626, "optim/lr": 0.0020568, "optim/total_tokens": 898891776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449930, "loss/out": 4.963092803955078, "created_at": "2025-01-14T17:50:32.599805+00:00"} {"global_step": 3430, "acc_step": 0, "speed/wps": 12928.097572864472, "speed/FLOPS": 203053537449547.7, "speed/curr_iter_time": 1.2781, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9039326310157776, "optim/lr": 0.0020574, "optim/total_tokens": 899153920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 406769, "loss/out": 4.925717830657959, "created_at": "2025-01-14T17:50:37.674738+00:00"} {"global_step": 3431, "acc_step": 0, "speed/wps": 12927.351477177719, "speed/FLOPS": 203041818991544.34, "speed/curr_iter_time": 1.2788, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8875440359115601, "optim/lr": 0.0020580000000000004, "optim/total_tokens": 899416064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478284, "loss/out": 4.968012809753418, "created_at": "2025-01-14T17:50:42.752644+00:00"} {"global_step": 3432, "acc_step": 0, "speed/wps": 12932.760280994975, "speed/FLOPS": 203126771687969.66, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8140604496002197, "optim/lr": 0.0020586000000000003, "optim/total_tokens": 899678208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 5.107707500457764, "created_at": "2025-01-14T17:50:47.833725+00:00"} {"global_step": 3433, "acc_step": 0, "speed/wps": 12928.83672582684, "speed/FLOPS": 203065146862524.3, "speed/curr_iter_time": 1.2782, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8189588785171509, "optim/lr": 0.0020592, "optim/total_tokens": 899940352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362914, "loss/out": 4.901340484619141, "created_at": "2025-01-14T17:50:52.910473+00:00"} {"global_step": 3434, "acc_step": 0, "speed/wps": 12932.402041853253, "speed/FLOPS": 203121145049976.75, "speed/curr_iter_time": 1.2787, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9306824207305908, "optim/lr": 0.0020598, "optim/total_tokens": 900202496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 4.865944862365723, "created_at": "2025-01-14T17:50:57.982166+00:00"} {"global_step": 3435, "acc_step": 0, "speed/wps": 12930.144627414356, "speed/FLOPS": 203085689254201.38, "speed/curr_iter_time": 1.2781, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.054655909538269, "optim/lr": 0.0020604, "optim/total_tokens": 900464640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434807, "loss/out": 4.9686737060546875, "created_at": "2025-01-14T17:51:03.054988+00:00"} {"global_step": 3436, "acc_step": 0, "speed/wps": 12929.71312966585, "speed/FLOPS": 203078911989121.03, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.048837423324585, "optim/lr": 0.0020610000000000003, "optim/total_tokens": 900726784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460271, "loss/out": 4.955793857574463, "created_at": "2025-01-14T17:51:08.127700+00:00"} {"global_step": 3437, "acc_step": 0, "speed/wps": 12932.80385126739, "speed/FLOPS": 203127456018969.75, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.0648446083068848, "optim/lr": 0.0020616000000000002, "optim/total_tokens": 900988928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322598, "loss/out": 4.882024765014648, "created_at": "2025-01-14T17:51:13.201132+00:00"} {"global_step": 3438, "acc_step": 0, "speed/wps": 12933.473681027726, "speed/FLOPS": 203137976615798.38, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.072633147239685, "optim/lr": 0.0020622, "optim/total_tokens": 901251072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 4.814491271972656, "created_at": "2025-01-14T17:51:18.271956+00:00"} {"global_step": 3439, "acc_step": 0, "speed/wps": 12928.62051936497, "speed/FLOPS": 203061751042941.8, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0974935293197632, "optim/lr": 0.0020628, "optim/total_tokens": 901513216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461368, "loss/out": 4.88589334487915, "created_at": "2025-01-14T17:51:23.344592+00:00"} {"global_step": 3440, "acc_step": 0, "speed/wps": 12936.442884902594, "speed/FLOPS": 203184612042766.88, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9181907773017883, "optim/lr": 0.0020634, "optim/total_tokens": 901775360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 4.871243476867676, "created_at": "2025-01-14T17:51:28.415093+00:00"} {"global_step": 3441, "acc_step": 0, "speed/wps": 12933.443731799385, "speed/FLOPS": 203137506222012.75, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8115901350975037, "optim/lr": 0.002064, "optim/total_tokens": 902037504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296822, "loss/out": 4.976827621459961, "created_at": "2025-01-14T17:51:33.517357+00:00"} {"global_step": 3442, "acc_step": 0, "speed/wps": 12932.70119265422, "speed/FLOPS": 203125843624382.22, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.983881950378418, "optim/lr": 0.0020646, "optim/total_tokens": 902299648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 4.849175453186035, "created_at": "2025-01-14T17:51:38.594171+00:00"} {"global_step": 3443, "acc_step": 0, "speed/wps": 12923.146340766296, "speed/FLOPS": 202975771545739.97, "speed/curr_iter_time": 1.2791, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.93138188123703, "optim/lr": 0.0020652, "optim/total_tokens": 902561792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326058, "loss/out": 4.861839771270752, "created_at": "2025-01-14T17:51:43.668981+00:00"} {"global_step": 3444, "acc_step": 0, "speed/wps": 12934.228053201356, "speed/FLOPS": 203149825067401.22, "speed/curr_iter_time": 1.2782, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7821797132492065, "optim/lr": 0.0020658, "optim/total_tokens": 902823936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482898, "loss/out": 4.923407554626465, "created_at": "2025-01-14T17:51:48.743591+00:00"} {"global_step": 3445, "acc_step": 0, "speed/wps": 12932.244540705731, "speed/FLOPS": 203118671278025.66, "speed/curr_iter_time": 1.2778, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8251814842224121, "optim/lr": 0.0020664, "optim/total_tokens": 903086080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457665, "loss/out": 4.872678756713867, "created_at": "2025-01-14T17:51:53.816491+00:00"} {"global_step": 3446, "acc_step": 0, "speed/wps": 12930.574901784885, "speed/FLOPS": 203092447304449.47, "speed/curr_iter_time": 1.2779, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7436432242393494, "optim/lr": 0.002067, "optim/total_tokens": 903348224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451026, "loss/out": 4.917115688323975, "created_at": "2025-01-14T17:51:58.888220+00:00"} {"global_step": 3447, "acc_step": 0, "speed/wps": 12934.661999287508, "speed/FLOPS": 203156640786988.66, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.7767708897590637, "optim/lr": 0.0020676, "optim/total_tokens": 903610368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318405, "loss/out": 4.860980987548828, "created_at": "2025-01-14T17:52:03.960576+00:00"} {"global_step": 3448, "acc_step": 0, "speed/wps": 12942.0628395874, "speed/FLOPS": 203272881153718.75, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8526672720909119, "optim/lr": 0.0020682, "optim/total_tokens": 903872512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460034, "loss/out": 4.911328315734863, "created_at": "2025-01-14T17:52:09.029453+00:00"} {"global_step": 3449, "acc_step": 0, "speed/wps": 12940.43408170129, "speed/FLOPS": 203247299272969.53, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7399152517318726, "optim/lr": 0.0020688, "optim/total_tokens": 904134656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479095, "loss/out": 4.880321979522705, "created_at": "2025-01-14T17:52:14.100901+00:00"} {"global_step": 3450, "acc_step": 0, "speed/wps": 12935.150053432875, "speed/FLOPS": 203164306347998.34, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.911090612411499, "optim/lr": 0.0020694, "optim/total_tokens": 904396800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440390, "loss/out": 4.860603332519531, "created_at": "2025-01-14T17:52:19.172660+00:00"} {"global_step": 3451, "acc_step": 0, "speed/wps": 12935.656302177684, "speed/FLOPS": 203172257680194.22, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.112051248550415, "optim/lr": 0.00207, "optim/total_tokens": 904658944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 4.911078453063965, "created_at": "2025-01-14T17:52:24.250382+00:00"} {"global_step": 3452, "acc_step": 0, "speed/wps": 12934.340638068514, "speed/FLOPS": 203151593367447.78, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7741860151290894, "optim/lr": 0.0020706, "optim/total_tokens": 904921088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372023, "loss/out": 4.952365875244141, "created_at": "2025-01-14T17:52:29.320863+00:00"} {"global_step": 3453, "acc_step": 0, "speed/wps": 12935.712889773878, "speed/FLOPS": 203173146466151.97, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7998108863830566, "optim/lr": 0.0020712, "optim/total_tokens": 905183232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 4.848221778869629, "created_at": "2025-01-14T17:52:34.396152+00:00"} {"global_step": 3454, "acc_step": 0, "speed/wps": 12925.263187822007, "speed/FLOPS": 203009019537602.6, "speed/curr_iter_time": 1.2779, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9743775129318237, "optim/lr": 0.0020718, "optim/total_tokens": 905445376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 4.793968677520752, "created_at": "2025-01-14T17:52:39.470418+00:00"} {"global_step": 3455, "acc_step": 0, "speed/wps": 12926.143248227245, "speed/FLOPS": 203022842095596.56, "speed/curr_iter_time": 1.2781, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.9319717884063721, "optim/lr": 0.0020724, "optim/total_tokens": 905707520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308138, "loss/out": 4.9015302658081055, "created_at": "2025-01-14T17:52:44.551917+00:00"} {"global_step": 3456, "acc_step": 0, "speed/wps": 12933.819703442532, "speed/FLOPS": 203143411373307.53, "speed/curr_iter_time": 1.2778, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.0583522319793701, "optim/lr": 0.0020729999999999998, "optim/total_tokens": 905969664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 4.776409149169922, "created_at": "2025-01-14T17:52:49.627628+00:00"} {"global_step": 3457, "acc_step": 0, "speed/wps": 12935.73110244377, "speed/FLOPS": 203173432521159.44, "speed/curr_iter_time": 1.2788, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7964909672737122, "optim/lr": 0.0020736, "optim/total_tokens": 906231808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446747, "loss/out": 4.779252529144287, "created_at": "2025-01-14T17:52:54.700767+00:00"} {"global_step": 3458, "acc_step": 0, "speed/wps": 12931.657288637796, "speed/FLOPS": 203109447677329.8, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8794364333152771, "optim/lr": 0.0020742, "optim/total_tokens": 906493952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443307, "loss/out": 4.914573669433594, "created_at": "2025-01-14T17:52:59.776814+00:00"} {"global_step": 3459, "acc_step": 0, "speed/wps": 12941.109535726933, "speed/FLOPS": 203257908206613.34, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.07899808883667, "optim/lr": 0.0020748, "optim/total_tokens": 906756096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472789, "loss/out": 4.772579669952393, "created_at": "2025-01-14T17:53:04.845534+00:00"} {"global_step": 3460, "acc_step": 0, "speed/wps": 12936.176083935698, "speed/FLOPS": 203180421566958.8, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8438498377799988, "optim/lr": 0.0020754, "optim/total_tokens": 907018240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463222, "loss/out": 4.85032844543457, "created_at": "2025-01-14T17:53:09.921392+00:00"} {"global_step": 3461, "acc_step": 0, "speed/wps": 12935.521525486627, "speed/FLOPS": 203170140827058.6, "speed/curr_iter_time": 1.2783, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9374247193336487, "optim/lr": 0.0020759999999999997, "optim/total_tokens": 907280384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 4.814572334289551, "created_at": "2025-01-14T17:53:14.991375+00:00"} {"global_step": 3462, "acc_step": 0, "speed/wps": 12939.011082294362, "speed/FLOPS": 203224949111877.62, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7703835368156433, "optim/lr": 0.0020766, "optim/total_tokens": 907542528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323417, "loss/out": 4.787816047668457, "created_at": "2025-01-14T17:53:20.060917+00:00"} {"global_step": 3463, "acc_step": 0, "speed/wps": 12934.079698362924, "speed/FLOPS": 203147494950802.75, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1517764329910278, "optim/lr": 0.0020772, "optim/total_tokens": 907804672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370682, "loss/out": 4.848937034606934, "created_at": "2025-01-14T17:53:25.133197+00:00"} {"global_step": 3464, "acc_step": 0, "speed/wps": 12935.093836928583, "speed/FLOPS": 203163423390548.84, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7870774269104004, "optim/lr": 0.0020778, "optim/total_tokens": 908066816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 4.908437728881836, "created_at": "2025-01-14T17:53:30.207102+00:00"} {"global_step": 3465, "acc_step": 0, "speed/wps": 12934.281435434572, "speed/FLOPS": 203150663508728.75, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0174198150634766, "optim/lr": 0.0020783999999999998, "optim/total_tokens": 908328960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447269, "loss/out": 5.019526481628418, "created_at": "2025-01-14T17:53:35.278960+00:00"} {"global_step": 3466, "acc_step": 0, "speed/wps": 12938.310994931066, "speed/FLOPS": 203213953277816.22, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8434248566627502, "optim/lr": 0.0020789999999999997, "optim/total_tokens": 908591104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434230, "loss/out": 4.939535617828369, "created_at": "2025-01-14T17:53:40.353039+00:00"} {"global_step": 3467, "acc_step": 0, "speed/wps": 12948.292241857733, "speed/FLOPS": 203370722476470.1, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8769662976264954, "optim/lr": 0.0020796, "optim/total_tokens": 908853248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 410502, "loss/out": 4.928530216217041, "created_at": "2025-01-14T17:53:45.423365+00:00"} {"global_step": 3468, "acc_step": 0, "speed/wps": 12938.52901248612, "speed/FLOPS": 203217377543105.84, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9496393203735352, "optim/lr": 0.0020802, "optim/total_tokens": 909115392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 4.8275370597839355, "created_at": "2025-01-14T17:53:50.494267+00:00"} {"global_step": 3469, "acc_step": 0, "speed/wps": 12940.202876480982, "speed/FLOPS": 203243667877275.3, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7676687240600586, "optim/lr": 0.0020808, "optim/total_tokens": 909377536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472260, "loss/out": 4.882809638977051, "created_at": "2025-01-14T17:53:55.563052+00:00"} {"global_step": 3470, "acc_step": 0, "speed/wps": 12942.926961051773, "speed/FLOPS": 203286453368744.06, "speed/curr_iter_time": 1.2774, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9985725283622742, "optim/lr": 0.0020813999999999997, "optim/total_tokens": 909639680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443545, "loss/out": 4.731289863586426, "created_at": "2025-01-14T17:54:00.630674+00:00"} {"global_step": 3471, "acc_step": 0, "speed/wps": 12944.74521600909, "speed/FLOPS": 203315011561392.88, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7599278092384338, "optim/lr": 0.002082, "optim/total_tokens": 909901824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 347841, "loss/out": 4.928511619567871, "created_at": "2025-01-14T17:54:05.697771+00:00"} {"global_step": 3472, "acc_step": 0, "speed/wps": 12941.968719422104, "speed/FLOPS": 203271402867188.12, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9339888691902161, "optim/lr": 0.0020826, "optim/total_tokens": 910163968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 4.866155624389648, "created_at": "2025-01-14T17:54:10.765159+00:00"} {"global_step": 3473, "acc_step": 0, "speed/wps": 12935.807803576772, "speed/FLOPS": 203174637217851.66, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.084905982017517, "optim/lr": 0.0020832, "optim/total_tokens": 910426112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426015, "loss/out": 4.820919990539551, "created_at": "2025-01-14T17:54:15.836158+00:00"} {"global_step": 3474, "acc_step": 0, "speed/wps": 12941.15563503359, "speed/FLOPS": 203258632259571.5, "speed/curr_iter_time": 1.2778, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9255173206329346, "optim/lr": 0.0020838000000000002, "optim/total_tokens": 910688256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450990, "loss/out": 4.978703022003174, "created_at": "2025-01-14T17:54:20.905253+00:00"} {"global_step": 3475, "acc_step": 0, "speed/wps": 12940.997130477333, "speed/FLOPS": 203256142727707.38, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7395035028457642, "optim/lr": 0.0020844, "optim/total_tokens": 910950400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295714, "loss/out": 4.808485984802246, "created_at": "2025-01-14T17:54:25.974124+00:00"} {"global_step": 3476, "acc_step": 0, "speed/wps": 12950.475532430433, "speed/FLOPS": 203405014055070.0, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9334914684295654, "optim/lr": 0.002085, "optim/total_tokens": 911212544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423603, "loss/out": 4.817654609680176, "created_at": "2025-01-14T17:54:31.040659+00:00"} {"global_step": 3477, "acc_step": 0, "speed/wps": 12948.547322160075, "speed/FLOPS": 203374728863135.44, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9143263697624207, "optim/lr": 0.0020856, "optim/total_tokens": 911474688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 4.928630352020264, "created_at": "2025-01-14T17:54:36.106584+00:00"} {"global_step": 3478, "acc_step": 0, "speed/wps": 12949.411223149413, "speed/FLOPS": 203388297615297.34, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9512982368469238, "optim/lr": 0.0020862000000000003, "optim/total_tokens": 911736832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 4.811914443969727, "created_at": "2025-01-14T17:54:41.171132+00:00"} {"global_step": 3479, "acc_step": 0, "speed/wps": 12949.593495025118, "speed/FLOPS": 203391160445573.1, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9544190168380737, "optim/lr": 0.0020868, "optim/total_tokens": 911998976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459765, "loss/out": 4.919954776763916, "created_at": "2025-01-14T17:54:46.239492+00:00"} {"global_step": 3480, "acc_step": 0, "speed/wps": 12957.766066317912, "speed/FLOPS": 203519521908016.25, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9317376017570496, "optim/lr": 0.0020874, "optim/total_tokens": 912261120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445170, "loss/out": 4.819140434265137, "created_at": "2025-01-14T17:54:51.306617+00:00"} {"global_step": 3481, "acc_step": 0, "speed/wps": 12956.53864791303, "speed/FLOPS": 203500243615317.94, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1311427354812622, "optim/lr": 0.002088, "optim/total_tokens": 912523264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329804, "loss/out": 4.71848201751709, "created_at": "2025-01-14T17:54:56.368312+00:00"} {"global_step": 3482, "acc_step": 0, "speed/wps": 12962.44853457837, "speed/FLOPS": 203593066506432.4, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.806953489780426, "optim/lr": 0.0020886000000000004, "optim/total_tokens": 912785408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470119, "loss/out": 4.8617963790893555, "created_at": "2025-01-14T17:55:01.429282+00:00"} {"global_step": 3483, "acc_step": 0, "speed/wps": 12963.352723336006, "speed/FLOPS": 203607268033356.62, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9692904949188232, "optim/lr": 0.0020892000000000003, "optim/total_tokens": 913047552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 4.764316082000732, "created_at": "2025-01-14T17:55:06.497562+00:00"} {"global_step": 3484, "acc_step": 0, "speed/wps": 12969.067416743237, "speed/FLOPS": 203697025146128.84, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1667766571044922, "optim/lr": 0.0020898, "optim/total_tokens": 913309696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447791, "loss/out": 4.809419631958008, "created_at": "2025-01-14T17:55:11.568540+00:00"} {"global_step": 3485, "acc_step": 0, "speed/wps": 12959.40589214403, "speed/FLOPS": 203545277625972.25, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.001, "optim/grad_norm": 1.0834481716156006, "optim/lr": 0.0020904, "optim/total_tokens": 913571840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430791, "loss/out": 4.959591865539551, "created_at": "2025-01-14T17:55:16.629616+00:00"} {"global_step": 3486, "acc_step": 0, "speed/wps": 12966.126785321541, "speed/FLOPS": 203650838488800.1, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.110106348991394, "optim/lr": 0.002091, "optim/total_tokens": 913833984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345728, "loss/out": 4.686039924621582, "created_at": "2025-01-14T17:55:21.691888+00:00"} {"global_step": 3487, "acc_step": 0, "speed/wps": 12962.69991059293, "speed/FLOPS": 203597014712167.9, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.123854398727417, "optim/lr": 0.0020916000000000003, "optim/total_tokens": 914096128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 4.8764472007751465, "created_at": "2025-01-14T17:55:26.751120+00:00"} {"global_step": 3488, "acc_step": 0, "speed/wps": 12961.032492259943, "speed/FLOPS": 203570825615975.44, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2537002563476562, "optim/lr": 0.0020922000000000002, "optim/total_tokens": 914358272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 4.815857887268066, "created_at": "2025-01-14T17:55:31.811638+00:00"} {"global_step": 3489, "acc_step": 0, "speed/wps": 12957.776207060055, "speed/FLOPS": 203519681182306.44, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1744798421859741, "optim/lr": 0.0020928, "optim/total_tokens": 914620416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 4.9193878173828125, "created_at": "2025-01-14T17:55:36.874022+00:00"} {"global_step": 3490, "acc_step": 0, "speed/wps": 12957.185600905279, "speed/FLOPS": 203510404901067.78, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1355022192001343, "optim/lr": 0.0020934, "optim/total_tokens": 914882560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 4.657243251800537, "created_at": "2025-01-14T17:55:41.936449+00:00"} {"global_step": 3491, "acc_step": 0, "speed/wps": 12961.828842201201, "speed/FLOPS": 203583333386102.4, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.15960693359375, "optim/lr": 0.002094, "optim/total_tokens": 915144704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442208, "loss/out": 4.725399017333984, "created_at": "2025-01-14T17:55:47.000752+00:00"} {"global_step": 3492, "acc_step": 0, "speed/wps": 12961.527337262327, "speed/FLOPS": 203578597836726.2, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.101442575454712, "optim/lr": 0.0020946000000000003, "optim/total_tokens": 915406848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 4.76383638381958, "created_at": "2025-01-14T17:55:52.060972+00:00"} {"global_step": 3493, "acc_step": 0, "speed/wps": 12959.704983227157, "speed/FLOPS": 203549975262426.97, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2515476942062378, "optim/lr": 0.0020952, "optim/total_tokens": 915668992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463770, "loss/out": 4.831243991851807, "created_at": "2025-01-14T17:55:57.122834+00:00"} {"global_step": 3494, "acc_step": 0, "speed/wps": 12961.437911519897, "speed/FLOPS": 203577193282557.47, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.954014003276825, "optim/lr": 0.0020958, "optim/total_tokens": 915931136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318405, "loss/out": 4.7563886642456055, "created_at": "2025-01-14T17:56:02.182709+00:00"} {"global_step": 3495, "acc_step": 0, "speed/wps": 12961.325496634505, "speed/FLOPS": 203575427652308.16, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0873483419418335, "optim/lr": 0.0020964, "optim/total_tokens": 916193280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 408017, "loss/out": 4.770183086395264, "created_at": "2025-01-14T17:56:07.243052+00:00"} {"global_step": 3496, "acc_step": 0, "speed/wps": 12960.081688416945, "speed/FLOPS": 203555891935079.78, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8956140279769897, "optim/lr": 0.002097, "optim/total_tokens": 916455424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 4.781518459320068, "created_at": "2025-01-14T17:56:12.338205+00:00"} {"global_step": 3497, "acc_step": 0, "speed/wps": 12965.036800574086, "speed/FLOPS": 203633718780545.3, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1810401678085327, "optim/lr": 0.0020976000000000002, "optim/total_tokens": 916717568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 4.778180122375488, "created_at": "2025-01-14T17:56:17.398118+00:00"} {"global_step": 3498, "acc_step": 0, "speed/wps": 12969.676640477954, "speed/FLOPS": 203706593842043.72, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0161586999893188, "optim/lr": 0.0020982, "optim/total_tokens": 916979712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430268, "loss/out": 4.772464275360107, "created_at": "2025-01-14T17:56:22.456741+00:00"} {"global_step": 3499, "acc_step": 0, "speed/wps": 12955.962245052775, "speed/FLOPS": 203491190416337.1, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.120131254196167, "optim/lr": 0.0020988, "optim/total_tokens": 917241856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300225, "loss/out": 4.827392101287842, "created_at": "2025-01-14T17:56:27.519721+00:00"} {"global_step": 3500, "acc_step": 0, "speed/wps": 12956.821231147986, "speed/FLOPS": 203504681973332.97, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.0253492593765259, "optim/lr": 0.0020994, "optim/total_tokens": 917504000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470403, "loss/out": 4.773015022277832, "created_at": "2025-01-14T17:56:32.581464+00:00"} {"global_step": 3501, "acc_step": 0, "speed/wps": 12960.608662745586, "speed/FLOPS": 203564168791049.7, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9273744821548462, "optim/lr": 0.0021, "optim/total_tokens": 917766144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 4.916046142578125, "created_at": "2025-01-14T17:56:37.643028+00:00"} {"global_step": 3502, "acc_step": 0, "speed/wps": 12957.02781043834, "speed/FLOPS": 203507926584957.66, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0917072296142578, "optim/lr": 0.0021006, "optim/total_tokens": 918028288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 4.781401634216309, "created_at": "2025-01-14T17:56:42.705648+00:00"} {"global_step": 3503, "acc_step": 0, "speed/wps": 12951.835109386599, "speed/FLOPS": 203426368079421.1, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.827088475227356, "optim/lr": 0.0021012, "optim/total_tokens": 918290432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 4.812496662139893, "created_at": "2025-01-14T17:56:47.770194+00:00"} {"global_step": 3504, "acc_step": 0, "speed/wps": 12956.650283171604, "speed/FLOPS": 203501997000456.75, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8813210129737854, "optim/lr": 0.0021018, "optim/total_tokens": 918552576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457920, "loss/out": 4.81065559387207, "created_at": "2025-01-14T17:56:52.833901+00:00"} {"global_step": 3505, "acc_step": 0, "speed/wps": 12958.594385801724, "speed/FLOPS": 203532531803739.75, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7998049259185791, "optim/lr": 0.0021024, "optim/total_tokens": 918814720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463222, "loss/out": 4.735105991363525, "created_at": "2025-01-14T17:56:57.896075+00:00"} {"global_step": 3506, "acc_step": 0, "speed/wps": 12959.641314084016, "speed/FLOPS": 203548975251043.38, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.8071004748344421, "optim/lr": 0.002103, "optim/total_tokens": 919076864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338751, "loss/out": 4.701467990875244, "created_at": "2025-01-14T17:57:02.958335+00:00"} {"global_step": 3507, "acc_step": 0, "speed/wps": 12955.848825482084, "speed/FLOPS": 203489409006124.56, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0698126554489136, "optim/lr": 0.0021036, "optim/total_tokens": 919339008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328845, "loss/out": 4.732202053070068, "created_at": "2025-01-14T17:57:08.021143+00:00"} {"global_step": 3508, "acc_step": 0, "speed/wps": 12961.74647219485, "speed/FLOPS": 203582039651964.4, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7691813707351685, "optim/lr": 0.0021042, "optim/total_tokens": 919601152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 4.820446014404297, "created_at": "2025-01-14T17:57:13.081291+00:00"} {"global_step": 3509, "acc_step": 0, "speed/wps": 12953.236984607156, "speed/FLOPS": 203448386456138.38, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.001, "optim/grad_norm": 1.007099986076355, "optim/lr": 0.0021048, "optim/total_tokens": 919863296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 4.838728427886963, "created_at": "2025-01-14T17:57:18.147572+00:00"} {"global_step": 3510, "acc_step": 0, "speed/wps": 12956.068457835985, "speed/FLOPS": 203492858634049.06, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0865846872329712, "optim/lr": 0.0021054, "optim/total_tokens": 920125440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456045, "loss/out": 4.800759792327881, "created_at": "2025-01-14T17:57:23.211005+00:00"} {"global_step": 3511, "acc_step": 0, "speed/wps": 12957.679076104816, "speed/FLOPS": 203518155607182.5, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7829170227050781, "optim/lr": 0.002106, "optim/total_tokens": 920387584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445170, "loss/out": 4.768328666687012, "created_at": "2025-01-14T17:57:28.281032+00:00"} {"global_step": 3512, "acc_step": 0, "speed/wps": 12963.155451486224, "speed/FLOPS": 203604169607875.88, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9793981909751892, "optim/lr": 0.0021066, "optim/total_tokens": 920649728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363627, "loss/out": 4.884794235229492, "created_at": "2025-01-14T17:57:33.348030+00:00"} {"global_step": 3513, "acc_step": 0, "speed/wps": 12960.004651419655, "speed/FLOPS": 203554681963177.4, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1477476358413696, "optim/lr": 0.0021072, "optim/total_tokens": 920911872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 4.753610610961914, "created_at": "2025-01-14T17:57:38.409275+00:00"} {"global_step": 3514, "acc_step": 0, "speed/wps": 12963.813810327312, "speed/FLOPS": 203614510038155.03, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.905632495880127, "optim/lr": 0.0021078, "optim/total_tokens": 921174016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459765, "loss/out": 4.797450542449951, "created_at": "2025-01-14T17:57:43.474785+00:00"} {"global_step": 3515, "acc_step": 0, "speed/wps": 12962.38035904561, "speed/FLOPS": 203591995716007.4, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7840292453765869, "optim/lr": 0.0021084, "optim/total_tokens": 921436160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 4.881089210510254, "created_at": "2025-01-14T17:57:48.534670+00:00"} {"global_step": 3516, "acc_step": 0, "speed/wps": 12967.009714517319, "speed/FLOPS": 203664706105091.62, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.001, "optim/grad_norm": 0.8974134922027588, "optim/lr": 0.0021089999999999998, "optim/total_tokens": 921698304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 4.736234664916992, "created_at": "2025-01-14T17:57:53.598394+00:00"} {"global_step": 3517, "acc_step": 0, "speed/wps": 12956.39493362002, "speed/FLOPS": 203497986384862.1, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8139622211456299, "optim/lr": 0.0021096, "optim/total_tokens": 921960448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450471, "loss/out": 4.833184719085693, "created_at": "2025-01-14T17:57:58.660159+00:00"} {"global_step": 3518, "acc_step": 0, "speed/wps": 12959.136823401192, "speed/FLOPS": 203541051531626.78, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0308750867843628, "optim/lr": 0.0021102, "optim/total_tokens": 922222592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328985, "loss/out": 4.787096977233887, "created_at": "2025-01-14T17:58:03.720761+00:00"} {"global_step": 3519, "acc_step": 0, "speed/wps": 12965.365914302787, "speed/FLOPS": 203638887963904.22, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1388989686965942, "optim/lr": 0.0021108, "optim/total_tokens": 922484736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 344264, "loss/out": 4.6933183670043945, "created_at": "2025-01-14T17:58:08.779071+00:00"} {"global_step": 3520, "acc_step": 0, "speed/wps": 12963.309486694074, "speed/FLOPS": 203606588942481.72, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0487151145935059, "optim/lr": 0.0021114, "optim/total_tokens": 922746880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466420, "loss/out": 4.714719772338867, "created_at": "2025-01-14T17:58:13.854622+00:00"} {"global_step": 3521, "acc_step": 0, "speed/wps": 12965.387937192727, "speed/FLOPS": 203639233863653.34, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.952914834022522, "optim/lr": 0.0021119999999999997, "optim/total_tokens": 923009024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 4.699906349182129, "created_at": "2025-01-14T17:58:18.913955+00:00"} {"global_step": 3522, "acc_step": 0, "speed/wps": 12961.4649274803, "speed/FLOPS": 203577617605339.4, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9230966567993164, "optim/lr": 0.0021126, "optim/total_tokens": 923271168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445166, "loss/out": 4.72752571105957, "created_at": "2025-01-14T17:58:23.978403+00:00"} {"global_step": 3523, "acc_step": 0, "speed/wps": 12965.317910443031, "speed/FLOPS": 203638133997322.53, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9358010292053223, "optim/lr": 0.0021132, "optim/total_tokens": 923533312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500101, "loss/out": 4.702901363372803, "created_at": "2025-01-14T17:58:29.038165+00:00"} {"global_step": 3524, "acc_step": 0, "speed/wps": 12955.934913302204, "speed/FLOPS": 203490761133636.62, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8962720632553101, "optim/lr": 0.0021138, "optim/total_tokens": 923795456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431842, "loss/out": 4.703262805938721, "created_at": "2025-01-14T17:58:34.099990+00:00"} {"global_step": 3525, "acc_step": 0, "speed/wps": 12957.088996523955, "speed/FLOPS": 203508887596510.75, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7299972772598267, "optim/lr": 0.0021144, "optim/total_tokens": 924057600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315181, "loss/out": 4.709747791290283, "created_at": "2025-01-14T17:58:39.169988+00:00"} {"global_step": 3526, "acc_step": 0, "speed/wps": 12959.635172990042, "speed/FLOPS": 203548878796723.75, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0599064826965332, "optim/lr": 0.002115, "optim/total_tokens": 924319744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482092, "loss/out": 4.758744716644287, "created_at": "2025-01-14T17:58:44.233985+00:00"} {"global_step": 3527, "acc_step": 0, "speed/wps": 12963.838213847937, "speed/FLOPS": 203614893328980.38, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7599015831947327, "optim/lr": 0.0021156, "optim/total_tokens": 924581888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 4.712515830993652, "created_at": "2025-01-14T17:58:49.292871+00:00"} {"global_step": 3528, "acc_step": 0, "speed/wps": 12960.371002444523, "speed/FLOPS": 203560436009442.0, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8624536991119385, "optim/lr": 0.0021162, "optim/total_tokens": 924844032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462126, "loss/out": 4.788138389587402, "created_at": "2025-01-14T17:58:54.353721+00:00"} {"global_step": 3529, "acc_step": 0, "speed/wps": 12959.775696424926, "speed/FLOPS": 203551085910368.03, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9179143309593201, "optim/lr": 0.0021168, "optim/total_tokens": 925106176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 4.712552070617676, "created_at": "2025-01-14T17:58:59.414214+00:00"} {"global_step": 3530, "acc_step": 0, "speed/wps": 12955.261500449757, "speed/FLOPS": 203480184259422.7, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9023142457008362, "optim/lr": 0.0021174, "optim/total_tokens": 925368320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476224, "loss/out": 4.7489776611328125, "created_at": "2025-01-14T17:59:04.476551+00:00"} {"global_step": 3531, "acc_step": 0, "speed/wps": 12958.1004066, "speed/FLOPS": 203524773181578.12, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0014, "optim/grad_norm": 1.1279181241989136, "optim/lr": 0.002118, "optim/total_tokens": 925630464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445407, "loss/out": 4.6683478355407715, "created_at": "2025-01-14T17:59:09.537737+00:00"} {"global_step": 3532, "acc_step": 0, "speed/wps": 12958.994003681017, "speed/FLOPS": 203538808351666.56, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.8726956248283386, "optim/lr": 0.0021186, "optim/total_tokens": 925892608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331435, "loss/out": 4.693364143371582, "created_at": "2025-01-14T17:59:14.602518+00:00"} {"global_step": 3533, "acc_step": 0, "speed/wps": 12954.848668582224, "speed/FLOPS": 203473700167652.7, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.141046166419983, "optim/lr": 0.0021192000000000003, "optim/total_tokens": 926154752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430753, "loss/out": 4.828360557556152, "created_at": "2025-01-14T17:59:19.670138+00:00"} {"global_step": 3534, "acc_step": 0, "speed/wps": 12955.141469753184, "speed/FLOPS": 203478299012393.56, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.4724787473678589, "optim/lr": 0.0021198000000000002, "optim/total_tokens": 926416896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 4.831971645355225, "created_at": "2025-01-14T17:59:24.737259+00:00"} {"global_step": 3535, "acc_step": 0, "speed/wps": 12959.881327297566, "speed/FLOPS": 203552744988374.88, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.7688586115837097, "optim/lr": 0.0021204, "optim/total_tokens": 926679040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428932, "loss/out": 4.7468414306640625, "created_at": "2025-01-14T17:59:29.797743+00:00"} {"global_step": 3536, "acc_step": 0, "speed/wps": 12963.2433080855, "speed/FLOPS": 203605549516495.12, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.217354655265808, "optim/lr": 0.002121, "optim/total_tokens": 926941184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 4.741488933563232, "created_at": "2025-01-14T17:59:34.858860+00:00"} {"global_step": 3537, "acc_step": 0, "speed/wps": 12959.221394091794, "speed/FLOPS": 203542379830534.97, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4475805759429932, "optim/lr": 0.0021216000000000004, "optim/total_tokens": 927203328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435866, "loss/out": 4.7347869873046875, "created_at": "2025-01-14T17:59:39.921563+00:00"} {"global_step": 3538, "acc_step": 0, "speed/wps": 12960.42100104426, "speed/FLOPS": 203561221306156.16, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1208200454711914, "optim/lr": 0.0021222000000000003, "optim/total_tokens": 927465472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312776, "loss/out": 4.8252410888671875, "created_at": "2025-01-14T17:59:44.986780+00:00"} {"global_step": 3539, "acc_step": 0, "speed/wps": 12967.585454324178, "speed/FLOPS": 203673748889907.78, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8724901080131531, "optim/lr": 0.0021228, "optim/total_tokens": 927727616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 4.730232238769531, "created_at": "2025-01-14T17:59:50.045358+00:00"} {"global_step": 3540, "acc_step": 0, "speed/wps": 12957.973441686046, "speed/FLOPS": 203522779023136.4, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1791670322418213, "optim/lr": 0.0021234, "optim/total_tokens": 927989760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 4.810080051422119, "created_at": "2025-01-14T17:59:55.109427+00:00"} {"global_step": 3541, "acc_step": 0, "speed/wps": 12963.684471768916, "speed/FLOPS": 203612478598365.6, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.022235631942749, "optim/lr": 0.002124, "optim/total_tokens": 928251904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445691, "loss/out": 4.728635787963867, "created_at": "2025-01-14T18:00:00.169400+00:00"} {"global_step": 3542, "acc_step": 0, "speed/wps": 12958.041418156256, "speed/FLOPS": 203523846687010.5, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0776805877685547, "optim/lr": 0.0021246000000000004, "optim/total_tokens": 928514048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439054, "loss/out": 4.772238731384277, "created_at": "2025-01-14T18:00:05.232773+00:00"} {"global_step": 3543, "acc_step": 0, "speed/wps": 12965.85227710952, "speed/FLOPS": 203646526960116.47, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0844281911849976, "optim/lr": 0.0021252000000000003, "optim/total_tokens": 928776192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 4.723661422729492, "created_at": "2025-01-14T18:00:10.295302+00:00"} {"global_step": 3544, "acc_step": 0, "speed/wps": 12961.471220715952, "speed/FLOPS": 203577716449253.16, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1234586238861084, "optim/lr": 0.0021258, "optim/total_tokens": 929038336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437723, "loss/out": 4.698293685913086, "created_at": "2025-01-14T18:00:15.355402+00:00"} {"global_step": 3545, "acc_step": 0, "speed/wps": 12962.987235331988, "speed/FLOPS": 203601527542020.38, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7341985106468201, "optim/lr": 0.0021264, "optim/total_tokens": 929300480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320571, "loss/out": 4.788287162780762, "created_at": "2025-01-14T18:00:20.416194+00:00"} {"global_step": 3546, "acc_step": 0, "speed/wps": 12961.984490945844, "speed/FLOPS": 203585778063520.8, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9175881743431091, "optim/lr": 0.002127, "optim/total_tokens": 929562624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436096, "loss/out": 4.688901901245117, "created_at": "2025-01-14T18:00:25.481573+00:00"} {"global_step": 3547, "acc_step": 0, "speed/wps": 12963.848879968305, "speed/FLOPS": 203615060855057.56, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1104201078414917, "optim/lr": 0.0021276000000000003, "optim/total_tokens": 929824768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 4.559781074523926, "created_at": "2025-01-14T18:00:30.546446+00:00"} {"global_step": 3548, "acc_step": 0, "speed/wps": 12960.708943293548, "speed/FLOPS": 203565743834855.34, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1480450630187988, "optim/lr": 0.0021282000000000002, "optim/total_tokens": 930086912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438531, "loss/out": 4.681765079498291, "created_at": "2025-01-14T18:00:35.609968+00:00"} {"global_step": 3549, "acc_step": 0, "speed/wps": 12960.645901645321, "speed/FLOPS": 203564753679141.62, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9326664805412292, "optim/lr": 0.0021288, "optim/total_tokens": 930349056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 4.771065711975098, "created_at": "2025-01-14T18:00:40.670541+00:00"} {"global_step": 3550, "acc_step": 0, "speed/wps": 12963.756771217746, "speed/FLOPS": 203613614160559.34, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1685583591461182, "optim/lr": 0.0021294, "optim/total_tokens": 930611200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431299, "loss/out": 4.800025939941406, "created_at": "2025-01-14T18:00:45.734675+00:00"} {"global_step": 3551, "acc_step": 0, "speed/wps": 12955.071324676854, "speed/FLOPS": 203477197287580.38, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1399445533752441, "optim/lr": 0.00213, "optim/total_tokens": 930873344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358828, "loss/out": 4.903903961181641, "created_at": "2025-01-14T18:00:50.802149+00:00"} {"global_step": 3552, "acc_step": 0, "speed/wps": 12962.322236579615, "speed/FLOPS": 203591082822810.3, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1422287225723267, "optim/lr": 0.0021306000000000003, "optim/total_tokens": 931135488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 4.621436595916748, "created_at": "2025-01-14T18:00:55.868127+00:00"} {"global_step": 3553, "acc_step": 0, "speed/wps": 12961.920460365882, "speed/FLOPS": 203584772375275.25, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4086501598358154, "optim/lr": 0.0021312, "optim/total_tokens": 931397632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466369, "loss/out": 4.870212554931641, "created_at": "2025-01-14T18:01:00.940191+00:00"} {"global_step": 3554, "acc_step": 0, "speed/wps": 12966.929326393076, "speed/FLOPS": 203663443499135.53, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1585693359375, "optim/lr": 0.0021318, "optim/total_tokens": 931659776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 4.781312942504883, "created_at": "2025-01-14T18:01:05.998120+00:00"} {"global_step": 3555, "acc_step": 0, "speed/wps": 12963.647248841247, "speed/FLOPS": 203611893961136.9, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1849852800369263, "optim/lr": 0.0021324, "optim/total_tokens": 931921920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445404, "loss/out": 4.782633304595947, "created_at": "2025-01-14T18:01:11.058934+00:00"} {"global_step": 3556, "acc_step": 0, "speed/wps": 12961.216530996906, "speed/FLOPS": 203573716197235.2, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0638985633850098, "optim/lr": 0.002133, "optim/total_tokens": 932184064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313963, "loss/out": 4.7824320793151855, "created_at": "2025-01-14T18:01:16.120777+00:00"} {"global_step": 3557, "acc_step": 0, "speed/wps": 12958.691091383007, "speed/FLOPS": 203534050697781.2, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3104857206344604, "optim/lr": 0.0021336000000000003, "optim/total_tokens": 932446208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301104, "loss/out": 4.725075721740723, "created_at": "2025-01-14T18:01:21.184778+00:00"} {"global_step": 3558, "acc_step": 0, "speed/wps": 12964.000390449426, "speed/FLOPS": 203617440535360.78, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.141969919204712, "optim/lr": 0.0021342, "optim/total_tokens": 932708352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472789, "loss/out": 4.671319007873535, "created_at": "2025-01-14T18:01:26.245087+00:00"} {"global_step": 3559, "acc_step": 0, "speed/wps": 12964.527583193174, "speed/FLOPS": 203625720821840.06, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1973316669464111, "optim/lr": 0.0021348, "optim/total_tokens": 932970496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 4.757707118988037, "created_at": "2025-01-14T18:01:31.305207+00:00"} {"global_step": 3560, "acc_step": 0, "speed/wps": 12957.97529332434, "speed/FLOPS": 203522808105660.22, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.139609694480896, "optim/lr": 0.0021354, "optim/total_tokens": 933232640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455517, "loss/out": 4.7274675369262695, "created_at": "2025-01-14T18:01:36.366652+00:00"} {"global_step": 3561, "acc_step": 0, "speed/wps": 12963.849030430945, "speed/FLOPS": 203615063218280.06, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1873756647109985, "optim/lr": 0.002136, "optim/total_tokens": 933494784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 4.643508434295654, "created_at": "2025-01-14T18:01:41.427112+00:00"} {"global_step": 3562, "acc_step": 0, "speed/wps": 12954.798231204562, "speed/FLOPS": 203472907979328.28, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1641640663146973, "optim/lr": 0.0021366000000000002, "optim/total_tokens": 933756928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431320, "loss/out": 4.687623977661133, "created_at": "2025-01-14T18:01:46.491987+00:00"} {"global_step": 3563, "acc_step": 0, "speed/wps": 12954.579612560958, "speed/FLOPS": 203469474273116.97, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.3662548065185547, "optim/lr": 0.0021372, "optim/total_tokens": 934019072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296475, "loss/out": 4.590682506561279, "created_at": "2025-01-14T18:01:51.559452+00:00"} {"global_step": 3564, "acc_step": 0, "speed/wps": 12963.957402650658, "speed/FLOPS": 203616765352909.7, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.123704433441162, "optim/lr": 0.0021378, "optim/total_tokens": 934281216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 4.804512023925781, "created_at": "2025-01-14T18:01:56.618950+00:00"} {"global_step": 3565, "acc_step": 0, "speed/wps": 12964.539041057988, "speed/FLOPS": 203625900783351.72, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0879061222076416, "optim/lr": 0.0021384, "optim/total_tokens": 934543360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 4.766940116882324, "created_at": "2025-01-14T18:02:01.677628+00:00"} {"global_step": 3566, "acc_step": 0, "speed/wps": 12945.302533932856, "speed/FLOPS": 203323765005220.7, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1961448192596436, "optim/lr": 0.002139, "optim/total_tokens": 934805504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425730, "loss/out": 4.66285514831543, "created_at": "2025-01-14T18:02:06.743776+00:00"} {"global_step": 3567, "acc_step": 0, "speed/wps": 12954.956890375213, "speed/FLOPS": 203475399939623.9, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5056254863739014, "optim/lr": 0.0021395999999999997, "optim/total_tokens": 935067648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 4.755212306976318, "created_at": "2025-01-14T18:02:11.809631+00:00"} {"global_step": 3568, "acc_step": 0, "speed/wps": 12950.83667800883, "speed/FLOPS": 203410686342644.75, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.187817931175232, "optim/lr": 0.0021402, "optim/total_tokens": 935329792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320721, "loss/out": 4.732905387878418, "created_at": "2025-01-14T18:02:16.879198+00:00"} {"global_step": 3569, "acc_step": 0, "speed/wps": 12963.302293749644, "speed/FLOPS": 203606475967405.3, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0917398929595947, "optim/lr": 0.0021408, "optim/total_tokens": 935591936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331435, "loss/out": 4.7468438148498535, "created_at": "2025-01-14T18:02:21.943862+00:00"} {"global_step": 3570, "acc_step": 0, "speed/wps": 12960.739243320199, "speed/FLOPS": 203566219738410.5, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4148906469345093, "optim/lr": 0.0021414, "optim/total_tokens": 935854080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 412655, "loss/out": 4.64113712310791, "created_at": "2025-01-14T18:02:27.005622+00:00"} {"global_step": 3571, "acc_step": 0, "speed/wps": 12961.589134067272, "speed/FLOPS": 203579568440465.53, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0212668180465698, "optim/lr": 0.002142, "optim/total_tokens": 936116224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 4.772511959075928, "created_at": "2025-01-14T18:02:32.065496+00:00"} {"global_step": 3572, "acc_step": 0, "speed/wps": 12966.680880511292, "speed/FLOPS": 203659541315161.72, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2077677249908447, "optim/lr": 0.0021425999999999997, "optim/total_tokens": 936378368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 4.7326154708862305, "created_at": "2025-01-14T18:02:37.123638+00:00"} {"global_step": 3573, "acc_step": 0, "speed/wps": 12958.346428633622, "speed/FLOPS": 203528637295685.88, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1037524938583374, "optim/lr": 0.0021432, "optim/total_tokens": 936640512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453144, "loss/out": 4.736118316650391, "created_at": "2025-01-14T18:02:42.186050+00:00"} {"global_step": 3574, "acc_step": 0, "speed/wps": 12970.484445575528, "speed/FLOPS": 203719281531141.66, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0962142944335938, "optim/lr": 0.0021438, "optim/total_tokens": 936902656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 4.7174296379089355, "created_at": "2025-01-14T18:02:47.244992+00:00"} {"global_step": 3575, "acc_step": 0, "speed/wps": 12961.868291836161, "speed/FLOPS": 203583952996828.94, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0869662761688232, "optim/lr": 0.0021444, "optim/total_tokens": 937164800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441157, "loss/out": 4.765049934387207, "created_at": "2025-01-14T18:02:52.304759+00:00"} {"global_step": 3576, "acc_step": 0, "speed/wps": 12964.10310949098, "speed/FLOPS": 203619053879059.12, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1977221965789795, "optim/lr": 0.0021449999999999998, "optim/total_tokens": 937426944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300751, "loss/out": 4.790326118469238, "created_at": "2025-01-14T18:02:57.364726+00:00"} {"global_step": 3577, "acc_step": 0, "speed/wps": 12960.956093285331, "speed/FLOPS": 203569625665095.97, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0915203094482422, "optim/lr": 0.0021455999999999997, "optim/total_tokens": 937689088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481474, "loss/out": 4.823334693908691, "created_at": "2025-01-14T18:03:02.424928+00:00"} {"global_step": 3578, "acc_step": 0, "speed/wps": 12965.765073175458, "speed/FLOPS": 203645157302501.28, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0919376611709595, "optim/lr": 0.0021462, "optim/total_tokens": 937951232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 4.666390419006348, "created_at": "2025-01-14T18:03:07.482991+00:00"} {"global_step": 3579, "acc_step": 0, "speed/wps": 12956.545599564171, "speed/FLOPS": 203500352800551.7, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.2407292127609253, "optim/lr": 0.0021468, "optim/total_tokens": 938213376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447032, "loss/out": 4.817821502685547, "created_at": "2025-01-14T18:03:12.547636+00:00"} {"global_step": 3580, "acc_step": 0, "speed/wps": 12957.362701942502, "speed/FLOPS": 203513186516219.7, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.379724383354187, "optim/lr": 0.0021474, "optim/total_tokens": 938475520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 4.76618766784668, "created_at": "2025-01-14T18:03:17.609043+00:00"} {"global_step": 3581, "acc_step": 0, "speed/wps": 12965.263378298509, "speed/FLOPS": 203637277495057.78, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0405197143554688, "optim/lr": 0.002148, "optim/total_tokens": 938737664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443306, "loss/out": 4.74461555480957, "created_at": "2025-01-14T18:03:22.667158+00:00"} {"global_step": 3582, "acc_step": 0, "speed/wps": 12960.372078988175, "speed/FLOPS": 203560452918039.4, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9655197262763977, "optim/lr": 0.0021486, "optim/total_tokens": 938999808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328460, "loss/out": 4.897970676422119, "created_at": "2025-01-14T18:03:27.727397+00:00"} {"global_step": 3583, "acc_step": 0, "speed/wps": 12961.387494442888, "speed/FLOPS": 203576401413082.72, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9615708589553833, "optim/lr": 0.0021492, "optim/total_tokens": 939261952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325819, "loss/out": 4.704809665679932, "created_at": "2025-01-14T18:03:32.787235+00:00"} {"global_step": 3584, "acc_step": 0, "speed/wps": 12962.617582047817, "speed/FLOPS": 203595721629235.62, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.237685203552246, "optim/lr": 0.0021498, "optim/total_tokens": 939524096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371118, "loss/out": 4.753803253173828, "created_at": "2025-01-14T18:03:37.852307+00:00"} {"global_step": 3585, "acc_step": 0, "speed/wps": 12957.95147022463, "speed/FLOPS": 203522433931142.94, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0852025747299194, "optim/lr": 0.0021504000000000002, "optim/total_tokens": 939786240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 4.835722923278809, "created_at": "2025-01-14T18:03:42.914437+00:00"} {"global_step": 3586, "acc_step": 0, "speed/wps": 12961.34201245465, "speed/FLOPS": 203575687055958.66, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1250309944152832, "optim/lr": 0.002151, "optim/total_tokens": 940048384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463770, "loss/out": 4.713921546936035, "created_at": "2025-01-14T18:03:47.977983+00:00"} {"global_step": 3587, "acc_step": 0, "speed/wps": 12963.721112177625, "speed/FLOPS": 203613054086333.6, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1438305377960205, "optim/lr": 0.0021516, "optim/total_tokens": 940310528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459779, "loss/out": 4.763246536254883, "created_at": "2025-01-14T18:03:53.045984+00:00"} {"global_step": 3588, "acc_step": 0, "speed/wps": 12961.641169038387, "speed/FLOPS": 203580385721190.44, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3350270986557007, "optim/lr": 0.0021522, "optim/total_tokens": 940572672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436381, "loss/out": 4.794243335723877, "created_at": "2025-01-14T18:03:58.107017+00:00"} {"global_step": 3589, "acc_step": 0, "speed/wps": 12967.288341166703, "speed/FLOPS": 203669082319493.34, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9518304467201233, "optim/lr": 0.0021528000000000003, "optim/total_tokens": 940834816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 4.702200889587402, "created_at": "2025-01-14T18:04:03.167472+00:00"} {"global_step": 3590, "acc_step": 0, "speed/wps": 12964.281913959898, "speed/FLOPS": 203621862248946.56, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1049654483795166, "optim/lr": 0.0021534, "optim/total_tokens": 941096960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427570, "loss/out": 4.761444091796875, "created_at": "2025-01-14T18:04:08.228150+00:00"} {"global_step": 3591, "acc_step": 0, "speed/wps": 12957.842585171004, "speed/FLOPS": 203520723741751.97, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8095051646232605, "optim/lr": 0.002154, "optim/total_tokens": 941359104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 4.706171035766602, "created_at": "2025-01-14T18:04:13.289686+00:00"} {"global_step": 3592, "acc_step": 0, "speed/wps": 12961.506332594705, "speed/FLOPS": 203578267929557.75, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1188722848892212, "optim/lr": 0.0021546, "optim/total_tokens": 941621248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458972, "loss/out": 4.755622863769531, "created_at": "2025-01-14T18:04:18.352655+00:00"} {"global_step": 3593, "acc_step": 0, "speed/wps": 12961.794423348898, "speed/FLOPS": 203582792790730.53, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1919734477996826, "optim/lr": 0.0021552000000000003, "optim/total_tokens": 941883392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 4.77689790725708, "created_at": "2025-01-14T18:04:23.414159+00:00"} {"global_step": 3594, "acc_step": 0, "speed/wps": 12963.466665985685, "speed/FLOPS": 203609057659243.44, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9306422472000122, "optim/lr": 0.0021558000000000003, "optim/total_tokens": 942145536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420921, "loss/out": 4.767070770263672, "created_at": "2025-01-14T18:04:28.473215+00:00"} {"global_step": 3595, "acc_step": 0, "speed/wps": 12967.926859134646, "speed/FLOPS": 203679111121595.53, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.116722822189331, "optim/lr": 0.0021564, "optim/total_tokens": 942407680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326870, "loss/out": 4.734244346618652, "created_at": "2025-01-14T18:04:33.533888+00:00"} {"global_step": 3596, "acc_step": 0, "speed/wps": 12966.795983179987, "speed/FLOPS": 203661349160741.22, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1450787782669067, "optim/lr": 0.002157, "optim/total_tokens": 942669824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295129, "loss/out": 4.6164679527282715, "created_at": "2025-01-14T18:04:38.592149+00:00"} {"global_step": 3597, "acc_step": 0, "speed/wps": 12962.654930394758, "speed/FLOPS": 203596308236346.34, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1659914255142212, "optim/lr": 0.0021576, "optim/total_tokens": 942931968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368187, "loss/out": 4.72014856338501, "created_at": "2025-01-14T18:04:43.651514+00:00"} {"global_step": 3598, "acc_step": 0, "speed/wps": 12963.039154965361, "speed/FLOPS": 203602343011207.56, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1309409141540527, "optim/lr": 0.0021582000000000003, "optim/total_tokens": 943194112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 4.8303937911987305, "created_at": "2025-01-14T18:04:48.716200+00:00"} {"global_step": 3599, "acc_step": 0, "speed/wps": 12970.035513358482, "speed/FLOPS": 203712230433775.44, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1248806715011597, "optim/lr": 0.0021588, "optim/total_tokens": 943456256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 4.715487957000732, "created_at": "2025-01-14T18:04:53.772530+00:00"} {"global_step": 3600, "acc_step": 0, "speed/wps": 12964.480736811698, "speed/FLOPS": 203624985035045.3, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1130448579788208, "optim/lr": 0.0021594, "optim/total_tokens": 943718400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443543, "loss/out": 4.782237529754639, "created_at": "2025-01-14T18:04:58.836251+00:00"} {"global_step": 3601, "acc_step": 0, "speed/wps": 12963.92452012665, "speed/FLOPS": 203616248887684.8, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0037376880645752, "optim/lr": 0.00216, "optim/total_tokens": 943980544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 4.649499416351318, "created_at": "2025-01-14T18:05:03.897024+00:00"} {"global_step": 3602, "acc_step": 0, "speed/wps": 12964.704548197407, "speed/FLOPS": 203628500300406.97, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9715257287025452, "optim/lr": 0.0021606, "optim/total_tokens": 944242688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 4.715878486633301, "created_at": "2025-01-14T18:05:08.957527+00:00"} {"global_step": 3603, "acc_step": 0, "speed/wps": 12962.381748719035, "speed/FLOPS": 203592017542738.16, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3914157152175903, "optim/lr": 0.0021612000000000003, "optim/total_tokens": 944504832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444359, "loss/out": 4.683588981628418, "created_at": "2025-01-14T18:05:14.018219+00:00"} {"global_step": 3604, "acc_step": 0, "speed/wps": 12959.59272966912, "speed/FLOPS": 203548212166045.47, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.001, "optim/grad_norm": 1.141080379486084, "optim/lr": 0.0021618, "optim/total_tokens": 944766976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361573, "loss/out": 4.719799518585205, "created_at": "2025-01-14T18:05:19.079662+00:00"} {"global_step": 3605, "acc_step": 0, "speed/wps": 12956.149830133494, "speed/FLOPS": 203494136697798.72, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0326330661773682, "optim/lr": 0.0021624, "optim/total_tokens": 945029120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450746, "loss/out": 4.590835094451904, "created_at": "2025-01-14T18:05:24.150611+00:00"} {"global_step": 3606, "acc_step": 0, "speed/wps": 12969.97757562072, "speed/FLOPS": 203711320441989.1, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2008943557739258, "optim/lr": 0.002163, "optim/total_tokens": 945291264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 4.721774101257324, "created_at": "2025-01-14T18:05:29.207495+00:00"} {"global_step": 3607, "acc_step": 0, "speed/wps": 12964.597111748682, "speed/FLOPS": 203626812863346.6, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1331114768981934, "optim/lr": 0.0021636, "optim/total_tokens": 945553408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451278, "loss/out": 4.684101104736328, "created_at": "2025-01-14T18:05:34.268073+00:00"} {"global_step": 3608, "acc_step": 0, "speed/wps": 12961.17631184025, "speed/FLOPS": 203573084500113.0, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2391295433044434, "optim/lr": 0.0021642000000000002, "optim/total_tokens": 945815552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 4.701924800872803, "created_at": "2025-01-14T18:05:39.327968+00:00"} {"global_step": 3609, "acc_step": 0, "speed/wps": 12961.4229377169, "speed/FLOPS": 203576958098405.22, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3718154430389404, "optim/lr": 0.0021648, "optim/total_tokens": 946077696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430283, "loss/out": 4.773691177368164, "created_at": "2025-01-14T18:05:44.388135+00:00"} {"global_step": 3610, "acc_step": 0, "speed/wps": 12947.453039286836, "speed/FLOPS": 203357541646908.97, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.144051194190979, "optim/lr": 0.0021654, "optim/total_tokens": 946339840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320967, "loss/out": 4.693440914154053, "created_at": "2025-01-14T18:05:49.454412+00:00"} {"global_step": 3611, "acc_step": 0, "speed/wps": 12960.19631352599, "speed/FLOPS": 203557692279928.72, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.012050986289978, "optim/lr": 0.002166, "optim/total_tokens": 946601984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369048, "loss/out": 4.683246612548828, "created_at": "2025-01-14T18:05:54.514987+00:00"} {"global_step": 3612, "acc_step": 0, "speed/wps": 12962.272094373084, "speed/FLOPS": 203590295270554.12, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.067551851272583, "optim/lr": 0.0021666, "optim/total_tokens": 946864128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 4.667786598205566, "created_at": "2025-01-14T18:05:59.576499+00:00"} {"global_step": 3613, "acc_step": 0, "speed/wps": 12963.779251965854, "speed/FLOPS": 203613967251600.16, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9595398902893066, "optim/lr": 0.0021672, "optim/total_tokens": 947126272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480143, "loss/out": 4.566190242767334, "created_at": "2025-01-14T18:06:04.635315+00:00"} {"global_step": 3614, "acc_step": 0, "speed/wps": 12955.75241754595, "speed/FLOPS": 203487894787009.2, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.279036283493042, "optim/lr": 0.0021678, "optim/total_tokens": 947388416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443016, "loss/out": 4.732030868530273, "created_at": "2025-01-14T18:06:09.697290+00:00"} {"global_step": 3615, "acc_step": 0, "speed/wps": 12938.712686292012, "speed/FLOPS": 203220262392621.66, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.230875015258789, "optim/lr": 0.0021684, "optim/total_tokens": 947650560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431096, "loss/out": 4.740571022033691, "created_at": "2025-01-14T18:06:14.767653+00:00"} {"global_step": 3616, "acc_step": 0, "speed/wps": 12953.591812634279, "speed/FLOPS": 203453959517889.12, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1416947841644287, "optim/lr": 0.002169, "optim/total_tokens": 947912704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289499, "loss/out": 4.706396102905273, "created_at": "2025-01-14T18:06:19.834938+00:00"} {"global_step": 3617, "acc_step": 0, "speed/wps": 12945.620984418938, "speed/FLOPS": 203328766707701.44, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.145818829536438, "optim/lr": 0.0021696, "optim/total_tokens": 948174848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 4.76246452331543, "created_at": "2025-01-14T18:06:24.903264+00:00"} {"global_step": 3618, "acc_step": 0, "speed/wps": 12951.982515364854, "speed/FLOPS": 203428683292866.62, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1255747079849243, "optim/lr": 0.0021702, "optim/total_tokens": 948436992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 4.777753829956055, "created_at": "2025-01-14T18:06:29.971257+00:00"} {"global_step": 3619, "acc_step": 0, "speed/wps": 12948.425778986883, "speed/FLOPS": 203372819860582.62, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1407661437988281, "optim/lr": 0.0021708, "optim/total_tokens": 948699136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 4.7914862632751465, "created_at": "2025-01-14T18:06:35.037305+00:00"} {"global_step": 3620, "acc_step": 0, "speed/wps": 12945.860845371746, "speed/FLOPS": 203332534053568.78, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2152825593948364, "optim/lr": 0.0021714, "optim/total_tokens": 948961280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322598, "loss/out": 4.647245407104492, "created_at": "2025-01-14T18:06:40.104564+00:00"} {"global_step": 3621, "acc_step": 0, "speed/wps": 12958.835961545332, "speed/FLOPS": 203536326082752.97, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.529330015182495, "optim/lr": 0.002172, "optim/total_tokens": 949223424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 385346, "loss/out": 4.766358375549316, "created_at": "2025-01-14T18:06:45.175288+00:00"} {"global_step": 3622, "acc_step": 0, "speed/wps": 12957.642987606388, "speed/FLOPS": 203517588787723.9, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3837101459503174, "optim/lr": 0.0021726, "optim/total_tokens": 949485568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 4.690268516540527, "created_at": "2025-01-14T18:06:50.237409+00:00"} {"global_step": 3623, "acc_step": 0, "speed/wps": 12952.160475472523, "speed/FLOPS": 203431478400900.38, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1357109546661377, "optim/lr": 0.0021732, "optim/total_tokens": 949747712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 4.796294212341309, "created_at": "2025-01-14T18:06:55.301012+00:00"} {"global_step": 3624, "acc_step": 0, "speed/wps": 12950.186481277078, "speed/FLOPS": 203400474109508.66, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1551294326782227, "optim/lr": 0.0021738, "optim/total_tokens": 950009856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439871, "loss/out": 4.674899578094482, "created_at": "2025-01-14T18:07:00.367091+00:00"} {"global_step": 3625, "acc_step": 0, "speed/wps": 12957.225129645141, "speed/FLOPS": 203511025754245.5, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3959356546401978, "optim/lr": 0.0021744, "optim/total_tokens": 950272000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 4.661494255065918, "created_at": "2025-01-14T18:07:05.428961+00:00"} {"global_step": 3626, "acc_step": 0, "speed/wps": 12956.63249821849, "speed/FLOPS": 203501717663329.03, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3221111297607422, "optim/lr": 0.002175, "optim/total_tokens": 950534144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435028, "loss/out": 4.737176895141602, "created_at": "2025-01-14T18:07:10.491764+00:00"} {"global_step": 3627, "acc_step": 0, "speed/wps": 12955.864833579502, "speed/FLOPS": 203489660435291.97, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.169565200805664, "optim/lr": 0.0021755999999999998, "optim/total_tokens": 950796288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355742, "loss/out": 4.717320442199707, "created_at": "2025-01-14T18:07:15.554084+00:00"} {"global_step": 3628, "acc_step": 0, "speed/wps": 12955.156636549435, "speed/FLOPS": 203478537227770.03, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2675217390060425, "optim/lr": 0.0021762, "optim/total_tokens": 951058432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456567, "loss/out": 4.666539669036865, "created_at": "2025-01-14T18:07:20.616385+00:00"} {"global_step": 3629, "acc_step": 0, "speed/wps": 12955.786594409, "speed/FLOPS": 203488431581607.34, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1892614364624023, "optim/lr": 0.0021768, "optim/total_tokens": 951320576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 4.754411220550537, "created_at": "2025-01-14T18:07:25.705654+00:00"} {"global_step": 3630, "acc_step": 0, "speed/wps": 12957.711724140749, "speed/FLOPS": 203518668389449.9, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1486587524414062, "optim/lr": 0.0021774, "optim/total_tokens": 951582720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432127, "loss/out": 4.795299530029297, "created_at": "2025-01-14T18:07:30.776010+00:00"} {"global_step": 3631, "acc_step": 0, "speed/wps": 12955.879796928504, "speed/FLOPS": 203489895455249.84, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1009188890457153, "optim/lr": 0.002178, "optim/total_tokens": 951844864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456584, "loss/out": 4.706394672393799, "created_at": "2025-01-14T18:07:35.839220+00:00"} {"global_step": 3632, "acc_step": 0, "speed/wps": 12955.047280903533, "speed/FLOPS": 203476819647080.75, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.180638313293457, "optim/lr": 0.0021785999999999997, "optim/total_tokens": 952107008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 4.65934944152832, "created_at": "2025-01-14T18:07:40.904966+00:00"} {"global_step": 3633, "acc_step": 0, "speed/wps": 12946.766592156986, "speed/FLOPS": 203346760051457.47, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2792165279388428, "optim/lr": 0.0021792, "optim/total_tokens": 952369152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448606, "loss/out": 4.557226181030273, "created_at": "2025-01-14T18:07:45.971762+00:00"} {"global_step": 3634, "acc_step": 0, "speed/wps": 12953.241624626737, "speed/FLOPS": 203448459334021.94, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.107552409172058, "optim/lr": 0.0021798, "optim/total_tokens": 952631296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 4.740410804748535, "created_at": "2025-01-14T18:07:51.034890+00:00"} {"global_step": 3635, "acc_step": 0, "speed/wps": 12952.296168900457, "speed/FLOPS": 203433609652648.4, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4244492053985596, "optim/lr": 0.0021804, "optim/total_tokens": 952893440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435329, "loss/out": 4.654635906219482, "created_at": "2025-01-14T18:07:56.098729+00:00"} {"global_step": 3636, "acc_step": 0, "speed/wps": 12950.879161720308, "speed/FLOPS": 203411353607712.25, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.2242450714111328, "optim/lr": 0.0021809999999999998, "optim/total_tokens": 953155584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330862, "loss/out": 4.574557781219482, "created_at": "2025-01-14T18:08:01.163367+00:00"} {"global_step": 3637, "acc_step": 0, "speed/wps": 12950.626167989614, "speed/FLOPS": 203407379993521.88, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8371798992156982, "optim/lr": 0.0021816, "optim/total_tokens": 953417728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367993, "loss/out": 4.627293586730957, "created_at": "2025-01-14T18:08:06.228207+00:00"} {"global_step": 3638, "acc_step": 0, "speed/wps": 12956.22396826639, "speed/FLOPS": 203495301139052.28, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7028071880340576, "optim/lr": 0.0021822, "optim/total_tokens": 953679872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 4.696932792663574, "created_at": "2025-01-14T18:08:11.292870+00:00"} {"global_step": 3639, "acc_step": 0, "speed/wps": 12959.23343445512, "speed/FLOPS": 203542568940986.22, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.001, "optim/grad_norm": 1.1468608379364014, "optim/lr": 0.0021828, "optim/total_tokens": 953942016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459507, "loss/out": 4.748628616333008, "created_at": "2025-01-14T18:08:16.357569+00:00"} {"global_step": 3640, "acc_step": 0, "speed/wps": 12946.984072950652, "speed/FLOPS": 203350175886171.94, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.001, "optim/grad_norm": 1.183834433555603, "optim/lr": 0.0021834, "optim/total_tokens": 954204160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 4.801169395446777, "created_at": "2025-01-14T18:08:21.423685+00:00"} {"global_step": 3641, "acc_step": 0, "speed/wps": 12952.938366042286, "speed/FLOPS": 203443696241232.2, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4221194982528687, "optim/lr": 0.002184, "optim/total_tokens": 954466304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334369, "loss/out": 4.527905464172363, "created_at": "2025-01-14T18:08:26.488231+00:00"} {"global_step": 3642, "acc_step": 0, "speed/wps": 12953.395324173678, "speed/FLOPS": 203450873396611.97, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3208929300308228, "optim/lr": 0.0021846, "optim/total_tokens": 954728448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 391766, "loss/out": 4.6560492515563965, "created_at": "2025-01-14T18:08:31.559889+00:00"} {"global_step": 3643, "acc_step": 0, "speed/wps": 12956.255275476238, "speed/FLOPS": 203495792861803.6, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.112242579460144, "optim/lr": 0.0021852, "optim/total_tokens": 954990592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 4.6917724609375, "created_at": "2025-01-14T18:08:36.629754+00:00"} {"global_step": 3644, "acc_step": 0, "speed/wps": 12957.889144981495, "speed/FLOPS": 203521455027555.62, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1937155723571777, "optim/lr": 0.0021858000000000003, "optim/total_tokens": 955252736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 4.718353271484375, "created_at": "2025-01-14T18:08:41.691508+00:00"} {"global_step": 3645, "acc_step": 0, "speed/wps": 12950.044652376737, "speed/FLOPS": 203398246491735.44, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1941595077514648, "optim/lr": 0.0021864000000000002, "optim/total_tokens": 955514880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456567, "loss/out": 4.670590877532959, "created_at": "2025-01-14T18:08:46.758496+00:00"} {"global_step": 3646, "acc_step": 0, "speed/wps": 12954.392838470732, "speed/FLOPS": 203466540729375.47, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.001, "optim/grad_norm": 1.220839023590088, "optim/lr": 0.002187, "optim/total_tokens": 955777024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302343, "loss/out": 4.774523735046387, "created_at": "2025-01-14T18:08:51.820943+00:00"} {"global_step": 3647, "acc_step": 0, "speed/wps": 12948.264842391245, "speed/FLOPS": 203370292130198.1, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3083312511444092, "optim/lr": 0.0021876, "optim/total_tokens": 956039168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460840, "loss/out": 4.685824394226074, "created_at": "2025-01-14T18:08:56.887358+00:00"} {"global_step": 3648, "acc_step": 0, "speed/wps": 12957.30672404468, "speed/FLOPS": 203512307306412.7, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9960774183273315, "optim/lr": 0.0021882000000000004, "optim/total_tokens": 956301312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495573, "loss/out": 4.592887878417969, "created_at": "2025-01-14T18:09:01.949749+00:00"} {"global_step": 3649, "acc_step": 0, "speed/wps": 12952.295987713338, "speed/FLOPS": 203433606806855.72, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.988960862159729, "optim/lr": 0.0021888000000000003, "optim/total_tokens": 956563456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 4.852331161499023, "created_at": "2025-01-14T18:09:07.014035+00:00"} {"global_step": 3650, "acc_step": 0, "speed/wps": 12956.040672407518, "speed/FLOPS": 203492422225713.8, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0899875164031982, "optim/lr": 0.0021894, "optim/total_tokens": 956825600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437723, "loss/out": 4.703383445739746, "created_at": "2025-01-14T18:09:12.084074+00:00"} {"global_step": 3651, "acc_step": 0, "speed/wps": 12947.861840608752, "speed/FLOPS": 203363962433421.53, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.181409478187561, "optim/lr": 0.00219, "optim/total_tokens": 957087744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454995, "loss/out": 4.588867664337158, "created_at": "2025-01-14T18:09:17.150371+00:00"} {"global_step": 3652, "acc_step": 0, "speed/wps": 12955.206463743973, "speed/FLOPS": 203479319832330.0, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4161746501922607, "optim/lr": 0.0021906, "optim/total_tokens": 957349888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310540, "loss/out": 4.678043365478516, "created_at": "2025-01-14T18:09:22.215808+00:00"} {"global_step": 3653, "acc_step": 0, "speed/wps": 12947.572775089486, "speed/FLOPS": 203359422262223.8, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1930261850357056, "optim/lr": 0.0021912000000000004, "optim/total_tokens": 957612032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468255, "loss/out": 4.614470958709717, "created_at": "2025-01-14T18:09:27.285392+00:00"} {"global_step": 3654, "acc_step": 0, "speed/wps": 12951.854828800373, "speed/FLOPS": 203426677799911.72, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.52714204788208, "optim/lr": 0.0021918000000000003, "optim/total_tokens": 957874176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 4.773327827453613, "created_at": "2025-01-14T18:09:32.354751+00:00"} {"global_step": 3655, "acc_step": 0, "speed/wps": 12955.228917232387, "speed/FLOPS": 203479672495219.88, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3741542100906372, "optim/lr": 0.0021924, "optim/total_tokens": 958136320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445170, "loss/out": 4.746721267700195, "created_at": "2025-01-14T18:09:37.421739+00:00"} {"global_step": 3656, "acc_step": 0, "speed/wps": 12961.009147303817, "speed/FLOPS": 203570458951360.16, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.46388840675354, "optim/lr": 0.002193, "optim/total_tokens": 958398464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429470, "loss/out": 4.681801795959473, "created_at": "2025-01-14T18:09:42.482521+00:00"} {"global_step": 3657, "acc_step": 0, "speed/wps": 12961.097645638914, "speed/FLOPS": 203571848939322.3, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3522459268569946, "optim/lr": 0.0021936, "optim/total_tokens": 958660608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319464, "loss/out": 4.632554531097412, "created_at": "2025-01-14T18:09:47.544949+00:00"} {"global_step": 3658, "acc_step": 0, "speed/wps": 12954.554239853838, "speed/FLOPS": 203469075759885.9, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.1859740018844604, "optim/lr": 0.0021942000000000003, "optim/total_tokens": 958922752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 4.643560409545898, "created_at": "2025-01-14T18:09:52.607544+00:00"} {"global_step": 3659, "acc_step": 0, "speed/wps": 12961.178415061762, "speed/FLOPS": 203573117534096.97, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2630351781845093, "optim/lr": 0.0021948000000000002, "optim/total_tokens": 959184896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 4.750783920288086, "created_at": "2025-01-14T18:09:57.668431+00:00"} {"global_step": 3660, "acc_step": 0, "speed/wps": 12951.57889834693, "speed/FLOPS": 203422343932972.22, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2050615549087524, "optim/lr": 0.0021954, "optim/total_tokens": 959447040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430791, "loss/out": 4.76663875579834, "created_at": "2025-01-14T18:10:02.731961+00:00"} {"global_step": 3661, "acc_step": 0, "speed/wps": 12963.61991570345, "speed/FLOPS": 203611464656648.12, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1472176313400269, "optim/lr": 0.002196, "optim/total_tokens": 959709184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 4.727142333984375, "created_at": "2025-01-14T18:10:07.793176+00:00"} {"global_step": 3662, "acc_step": 0, "speed/wps": 12950.703574989411, "speed/FLOPS": 203408595776822.06, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1567258834838867, "optim/lr": 0.0021966, "optim/total_tokens": 959971328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430776, "loss/out": 4.543984413146973, "created_at": "2025-01-14T18:10:12.857827+00:00"} {"global_step": 3663, "acc_step": 0, "speed/wps": 12951.503783584598, "speed/FLOPS": 203421164152411.3, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.27424955368042, "optim/lr": 0.0021972000000000003, "optim/total_tokens": 960233472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318558, "loss/out": 4.674549102783203, "created_at": "2025-01-14T18:10:17.924323+00:00"} {"global_step": 3664, "acc_step": 0, "speed/wps": 12956.606802566292, "speed/FLOPS": 203501314077801.72, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2019357681274414, "optim/lr": 0.0021978, "optim/total_tokens": 960495616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427047, "loss/out": 4.703693389892578, "created_at": "2025-01-14T18:10:22.987257+00:00"} {"global_step": 3665, "acc_step": 0, "speed/wps": 12957.13036414212, "speed/FLOPS": 203509537331799.1, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.318773865699768, "optim/lr": 0.0021984, "optim/total_tokens": 960757760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 4.497052192687988, "created_at": "2025-01-14T18:10:28.049843+00:00"} {"global_step": 3666, "acc_step": 0, "speed/wps": 12953.860672624982, "speed/FLOPS": 203458182333495.47, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0783780813217163, "optim/lr": 0.002199, "optim/total_tokens": 961019904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421768, "loss/out": 4.7700676918029785, "created_at": "2025-01-14T18:10:33.121657+00:00"} {"global_step": 3667, "acc_step": 0, "speed/wps": 12959.732450225652, "speed/FLOPS": 203550406669381.97, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2037568092346191, "optim/lr": 0.0021996, "optim/total_tokens": 961282048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469590, "loss/out": 4.651789665222168, "created_at": "2025-01-14T18:10:38.187467+00:00"} {"global_step": 3668, "acc_step": 0, "speed/wps": 12954.439042002008, "speed/FLOPS": 203467266419324.47, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3082349300384521, "optim/lr": 0.0022002000000000002, "optim/total_tokens": 961544192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291719, "loss/out": 4.728002548217773, "created_at": "2025-01-14T18:10:43.251306+00:00"} {"global_step": 3669, "acc_step": 0, "speed/wps": 12957.921741581536, "speed/FLOPS": 203521967001951.7, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.116117000579834, "optim/lr": 0.0022008, "optim/total_tokens": 961806336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457361, "loss/out": 4.68795919418335, "created_at": "2025-01-14T18:10:48.314990+00:00"} {"global_step": 3670, "acc_step": 0, "speed/wps": 12957.566750270458, "speed/FLOPS": 203516391375602.03, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.2441902160644531, "optim/lr": 0.0022014, "optim/total_tokens": 962068480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 4.68595027923584, "created_at": "2025-01-14T18:10:53.376594+00:00"} {"global_step": 3671, "acc_step": 0, "speed/wps": 12956.985331707217, "speed/FLOPS": 203507259398113.5, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.067007303237915, "optim/lr": 0.002202, "optim/total_tokens": 962330624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415364, "loss/out": 4.679096221923828, "created_at": "2025-01-14T18:10:58.438135+00:00"} {"global_step": 3672, "acc_step": 0, "speed/wps": 12957.605353183439, "speed/FLOPS": 203516997687396.34, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.337255597114563, "optim/lr": 0.0022026, "optim/total_tokens": 962592768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 4.5759382247924805, "created_at": "2025-01-14T18:11:03.503049+00:00"} {"global_step": 3673, "acc_step": 0, "speed/wps": 12958.096886988053, "speed/FLOPS": 203524717901236.03, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2090771198272705, "optim/lr": 0.0022032, "optim/total_tokens": 962854912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444359, "loss/out": 4.644629001617432, "created_at": "2025-01-14T18:11:08.564492+00:00"} {"global_step": 3674, "acc_step": 0, "speed/wps": 12954.713556058918, "speed/FLOPS": 203471578039810.1, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0011, "optim/grad_norm": 1.2652086019515991, "optim/lr": 0.0022038, "optim/total_tokens": 963117056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362390, "loss/out": 4.684818267822266, "created_at": "2025-01-14T18:11:13.631410+00:00"} {"global_step": 3675, "acc_step": 0, "speed/wps": 12958.086332936025, "speed/FLOPS": 203524552135346.16, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1405609846115112, "optim/lr": 0.0022044, "optim/total_tokens": 963379200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 4.722487449645996, "created_at": "2025-01-14T18:11:18.693089+00:00"} {"global_step": 3676, "acc_step": 0, "speed/wps": 12953.821421947463, "speed/FLOPS": 203457565847668.94, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1758900880813599, "optim/lr": 0.002205, "optim/total_tokens": 963641344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462150, "loss/out": 4.635061264038086, "created_at": "2025-01-14T18:11:23.756022+00:00"} {"global_step": 3677, "acc_step": 0, "speed/wps": 12963.078401822855, "speed/FLOPS": 203602959437035.34, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2646105289459229, "optim/lr": 0.0022056, "optim/total_tokens": 963903488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 4.738041400909424, "created_at": "2025-01-14T18:11:28.815140+00:00"} {"global_step": 3678, "acc_step": 0, "speed/wps": 12959.878038240897, "speed/FLOPS": 203552693329220.25, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1979026794433594, "optim/lr": 0.0022062, "optim/total_tokens": 964165632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423627, "loss/out": 4.657873153686523, "created_at": "2025-01-14T18:11:33.877974+00:00"} {"global_step": 3679, "acc_step": 0, "speed/wps": 12960.39374834613, "speed/FLOPS": 203560793265082.9, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1528338193893433, "optim/lr": 0.0022068, "optim/total_tokens": 964427776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483617, "loss/out": 4.675717353820801, "created_at": "2025-01-14T18:11:38.946251+00:00"} {"global_step": 3680, "acc_step": 0, "speed/wps": 12959.030557917293, "speed/FLOPS": 203539382486178.06, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1129063367843628, "optim/lr": 0.0022074, "optim/total_tokens": 964689920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304748, "loss/out": 4.639442443847656, "created_at": "2025-01-14T18:11:44.009736+00:00"} {"global_step": 3681, "acc_step": 0, "speed/wps": 12956.549624082158, "speed/FLOPS": 203500416011136.9, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.111890435218811, "optim/lr": 0.002208, "optim/total_tokens": 964952064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460071, "loss/out": 4.582405090332031, "created_at": "2025-01-14T18:11:49.071506+00:00"} {"global_step": 3682, "acc_step": 0, "speed/wps": 12952.752895752557, "speed/FLOPS": 203440783175469.1, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.229921579360962, "optim/lr": 0.0022086, "optim/total_tokens": 965214208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 4.694931507110596, "created_at": "2025-01-14T18:11:54.134826+00:00"} {"global_step": 3683, "acc_step": 0, "speed/wps": 12953.80225930942, "speed/FLOPS": 203457264872106.25, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1366541385650635, "optim/lr": 0.0022092, "optim/total_tokens": 965476352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436381, "loss/out": 4.7010908126831055, "created_at": "2025-01-14T18:11:59.198407+00:00"} {"global_step": 3684, "acc_step": 0, "speed/wps": 12949.193992030227, "speed/FLOPS": 203384885702063.7, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2963534593582153, "optim/lr": 0.0022098, "optim/total_tokens": 965738496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 4.772195816040039, "created_at": "2025-01-14T18:12:04.262964+00:00"} {"global_step": 3685, "acc_step": 0, "speed/wps": 12950.88521109552, "speed/FLOPS": 203411448621462.66, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1248406171798706, "optim/lr": 0.0022104, "optim/total_tokens": 966000640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331674, "loss/out": 4.6927690505981445, "created_at": "2025-01-14T18:12:09.330777+00:00"} {"global_step": 3686, "acc_step": 0, "speed/wps": 12953.210145997964, "speed/FLOPS": 203447964918900.88, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9777114987373352, "optim/lr": 0.002211, "optim/total_tokens": 966262784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374420, "loss/out": 4.713546276092529, "created_at": "2025-01-14T18:12:14.393816+00:00"} {"global_step": 3687, "acc_step": 0, "speed/wps": 12962.577167945143, "speed/FLOPS": 203595086870218.34, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3119664192199707, "optim/lr": 0.0022115999999999998, "optim/total_tokens": 966524928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472504, "loss/out": 4.669869422912598, "created_at": "2025-01-14T18:12:19.461274+00:00"} {"global_step": 3688, "acc_step": 0, "speed/wps": 12953.983808265466, "speed/FLOPS": 203460116347935.38, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.113284707069397, "optim/lr": 0.0022122, "optim/total_tokens": 966787072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 4.6629638671875, "created_at": "2025-01-14T18:12:24.525140+00:00"} {"global_step": 3689, "acc_step": 0, "speed/wps": 12958.470876135498, "speed/FLOPS": 203530591914711.6, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5428348779678345, "optim/lr": 0.0022128, "optim/total_tokens": 967049216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 4.631288528442383, "created_at": "2025-01-14T18:12:29.598313+00:00"} {"global_step": 3690, "acc_step": 0, "speed/wps": 12928.57164231433, "speed/FLOPS": 203060983361697.4, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.5138423442840576, "optim/lr": 0.0022134, "optim/total_tokens": 967311360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428932, "loss/out": 4.730353832244873, "created_at": "2025-01-14T18:12:34.671240+00:00"} {"global_step": 3691, "acc_step": 0, "speed/wps": 12956.219779352195, "speed/FLOPS": 203495235346398.72, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1915020942687988, "optim/lr": 0.002214, "optim/total_tokens": 967573504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295714, "loss/out": 4.75549840927124, "created_at": "2025-01-14T18:12:39.733062+00:00"} {"global_step": 3692, "acc_step": 0, "speed/wps": 12956.512205942314, "speed/FLOPS": 203499828307832.75, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.546557068824768, "optim/lr": 0.0022146, "optim/total_tokens": 967835648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456584, "loss/out": 4.6742753982543945, "created_at": "2025-01-14T18:12:44.795208+00:00"} {"global_step": 3693, "acc_step": 0, "speed/wps": 12959.054998323096, "speed/FLOPS": 203539766356335.84, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1210134029388428, "optim/lr": 0.0022152, "optim/total_tokens": 968097792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 4.634154319763184, "created_at": "2025-01-14T18:12:49.859379+00:00"} {"global_step": 3694, "acc_step": 0, "speed/wps": 12950.418791768232, "speed/FLOPS": 203404122865000.28, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2900367975234985, "optim/lr": 0.0022158, "optim/total_tokens": 968359936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454466, "loss/out": 4.724421977996826, "created_at": "2025-01-14T18:12:54.925491+00:00"} {"global_step": 3695, "acc_step": 0, "speed/wps": 12960.190787738617, "speed/FLOPS": 203557605489844.8, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.001, "optim/grad_norm": 1.3212062120437622, "optim/lr": 0.0022164, "optim/total_tokens": 968622080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 4.687380790710449, "created_at": "2025-01-14T18:12:59.986653+00:00"} {"global_step": 3696, "acc_step": 0, "speed/wps": 12955.757619916943, "speed/FLOPS": 203487976497394.47, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.300248146057129, "optim/lr": 0.0022170000000000002, "optim/total_tokens": 968884224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438536, "loss/out": 4.75861930847168, "created_at": "2025-01-14T18:13:05.052122+00:00"} {"global_step": 3697, "acc_step": 0, "speed/wps": 12953.869242944287, "speed/FLOPS": 203458316942137.0, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2839254140853882, "optim/lr": 0.0022176, "optim/total_tokens": 969146368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353872, "loss/out": 4.799399375915527, "created_at": "2025-01-14T18:13:10.122837+00:00"} {"global_step": 3698, "acc_step": 0, "speed/wps": 12955.294780293347, "speed/FLOPS": 203480706965097.6, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4598422050476074, "optim/lr": 0.0022182, "optim/total_tokens": 969408512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 4.586240768432617, "created_at": "2025-01-14T18:13:15.185247+00:00"} {"global_step": 3699, "acc_step": 0, "speed/wps": 12958.862040714086, "speed/FLOPS": 203536735691934.8, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2555406093597412, "optim/lr": 0.0022188, "optim/total_tokens": 969670656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442778, "loss/out": 4.60904598236084, "created_at": "2025-01-14T18:13:20.250122+00:00"} {"global_step": 3700, "acc_step": 0, "speed/wps": 12957.964868321309, "speed/FLOPS": 203522644366662.2, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8464938402175903, "optim/lr": 0.0022194000000000003, "optim/total_tokens": 969932800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 4.7071733474731445, "created_at": "2025-01-14T18:13:25.316139+00:00"} {"global_step": 3701, "acc_step": 0, "speed/wps": 12955.058722017353, "speed/FLOPS": 203476999345495.03, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.3202342987060547, "optim/lr": 0.00222, "optim/total_tokens": 970194944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433179, "loss/out": 4.643060207366943, "created_at": "2025-01-14T18:13:30.378396+00:00"} {"global_step": 3702, "acc_step": 0, "speed/wps": 12954.354691114795, "speed/FLOPS": 203465941572730.53, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.2286237478256226, "optim/lr": 0.0022206, "optim/total_tokens": 970457088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338116, "loss/out": 4.623969078063965, "created_at": "2025-01-14T18:13:35.441803+00:00"} {"global_step": 3703, "acc_step": 0, "speed/wps": 12950.053467885027, "speed/FLOPS": 203398384951406.94, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1209839582443237, "optim/lr": 0.0022212, "optim/total_tokens": 970719232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325000, "loss/out": 4.656749725341797, "created_at": "2025-01-14T18:13:40.508044+00:00"} {"global_step": 3704, "acc_step": 0, "speed/wps": 12956.88372266258, "speed/FLOPS": 203505663488442.0, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4555219411849976, "optim/lr": 0.0022218000000000003, "optim/total_tokens": 970981376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374134, "loss/out": 4.7586164474487305, "created_at": "2025-01-14T18:13:45.572756+00:00"} {"global_step": 3705, "acc_step": 0, "speed/wps": 12957.320345081846, "speed/FLOPS": 203512521243518.66, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.181663155555725, "optim/lr": 0.0022224000000000002, "optim/total_tokens": 971243520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453124, "loss/out": 4.6617326736450195, "created_at": "2025-01-14T18:13:50.665921+00:00"} {"global_step": 3706, "acc_step": 0, "speed/wps": 12953.768860750506, "speed/FLOPS": 203456740301844.1, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1181732416152954, "optim/lr": 0.002223, "optim/total_tokens": 971505664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 4.669792652130127, "created_at": "2025-01-14T18:13:55.728966+00:00"} {"global_step": 3707, "acc_step": 0, "speed/wps": 12949.80283948992, "speed/FLOPS": 203394448488062.25, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1834278106689453, "optim/lr": 0.0022236, "optim/total_tokens": 971767808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328460, "loss/out": 4.640666961669922, "created_at": "2025-01-14T18:14:00.793506+00:00"} {"global_step": 3708, "acc_step": 0, "speed/wps": 12949.889130576354, "speed/FLOPS": 203395803808151.1, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4245558977127075, "optim/lr": 0.0022242, "optim/total_tokens": 972029952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441147, "loss/out": 4.590191841125488, "created_at": "2025-01-14T18:14:05.858784+00:00"} {"global_step": 3709, "acc_step": 0, "speed/wps": 12952.989711661436, "speed/FLOPS": 203444502694737.2, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1529467105865479, "optim/lr": 0.0022248000000000003, "optim/total_tokens": 972292096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 4.747645378112793, "created_at": "2025-01-14T18:14:10.923057+00:00"} {"global_step": 3710, "acc_step": 0, "speed/wps": 12954.359513180016, "speed/FLOPS": 203466017309891.03, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2970331907272339, "optim/lr": 0.0022254, "optim/total_tokens": 972554240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439820, "loss/out": 4.622020721435547, "created_at": "2025-01-14T18:14:15.987154+00:00"} {"global_step": 3711, "acc_step": 0, "speed/wps": 12960.959401987067, "speed/FLOPS": 203569677632803.38, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.368464708328247, "optim/lr": 0.002226, "optim/total_tokens": 972816384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 4.636009216308594, "created_at": "2025-01-14T18:14:21.048963+00:00"} {"global_step": 3712, "acc_step": 0, "speed/wps": 12958.711597714475, "speed/FLOPS": 203534372777895.56, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1099809408187866, "optim/lr": 0.0022266, "optim/total_tokens": 973078528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444881, "loss/out": 4.4609375, "created_at": "2025-01-14T18:14:26.119133+00:00"} {"global_step": 3713, "acc_step": 0, "speed/wps": 12959.94521122938, "speed/FLOPS": 203553748373309.75, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2417596578598022, "optim/lr": 0.0022272, "optim/total_tokens": 973340672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364445, "loss/out": 4.526149272918701, "created_at": "2025-01-14T18:14:31.181200+00:00"} {"global_step": 3714, "acc_step": 0, "speed/wps": 12958.742687851804, "speed/FLOPS": 203534861091224.62, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3759074211120605, "optim/lr": 0.0022278000000000003, "optim/total_tokens": 973602816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 4.703272819519043, "created_at": "2025-01-14T18:14:36.243725+00:00"} {"global_step": 3715, "acc_step": 0, "speed/wps": 12941.745232953877, "speed/FLOPS": 203267892705102.03, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.1561975479125977, "optim/lr": 0.0022284, "optim/total_tokens": 973864960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 4.7252068519592285, "created_at": "2025-01-14T18:14:41.311233+00:00"} {"global_step": 3716, "acc_step": 0, "speed/wps": 12955.118887773551, "speed/FLOPS": 203477944331372.53, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1336055994033813, "optim/lr": 0.002229, "optim/total_tokens": 974127104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 4.618935585021973, "created_at": "2025-01-14T18:14:46.381330+00:00"} {"global_step": 3717, "acc_step": 0, "speed/wps": 12959.116898042732, "speed/FLOPS": 203540738576491.84, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.505129098892212, "optim/lr": 0.0022296, "optim/total_tokens": 974389248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455808, "loss/out": 4.620952606201172, "created_at": "2025-01-14T18:14:51.443868+00:00"} {"global_step": 3718, "acc_step": 0, "speed/wps": 12953.990141931827, "speed/FLOPS": 203460215826868.94, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1295968294143677, "optim/lr": 0.0022302, "optim/total_tokens": 974651392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435338, "loss/out": 4.589671611785889, "created_at": "2025-01-14T18:14:56.510891+00:00"} {"global_step": 3719, "acc_step": 0, "speed/wps": 12956.541720570542, "speed/FLOPS": 203500291875626.44, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1705753803253174, "optim/lr": 0.0022308000000000002, "optim/total_tokens": 974913536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333837, "loss/out": 4.689095497131348, "created_at": "2025-01-14T18:15:01.574420+00:00"} {"global_step": 3720, "acc_step": 0, "speed/wps": 12951.235221646035, "speed/FLOPS": 203416946018123.4, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.186133861541748, "optim/lr": 0.0022314, "optim/total_tokens": 975175680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463770, "loss/out": 4.583495140075684, "created_at": "2025-01-14T18:15:06.663175+00:00"} {"global_step": 3721, "acc_step": 0, "speed/wps": 12958.632889837505, "speed/FLOPS": 203533136562531.78, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2008315324783325, "optim/lr": 0.002232, "optim/total_tokens": 975437824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 4.579405784606934, "created_at": "2025-01-14T18:15:11.724041+00:00"} {"global_step": 3722, "acc_step": 0, "speed/wps": 12957.018921016128, "speed/FLOPS": 203507786964366.5, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3247050046920776, "optim/lr": 0.0022326, "optim/total_tokens": 975699968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432365, "loss/out": 4.67841100692749, "created_at": "2025-01-14T18:15:16.788054+00:00"} {"global_step": 3723, "acc_step": 0, "speed/wps": 12955.23452942409, "speed/FLOPS": 203479760642402.53, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1100362539291382, "optim/lr": 0.0022332, "optim/total_tokens": 975962112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352687, "loss/out": 4.561008453369141, "created_at": "2025-01-14T18:15:21.851530+00:00"} {"global_step": 3724, "acc_step": 0, "speed/wps": 12952.507188581389, "speed/FLOPS": 203436924006708.84, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1813666820526123, "optim/lr": 0.0022338, "optim/total_tokens": 976224256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475174, "loss/out": 4.5162553787231445, "created_at": "2025-01-14T18:15:26.915955+00:00"} {"global_step": 3725, "acc_step": 0, "speed/wps": 12955.818254920481, "speed/FLOPS": 203488928853446.25, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.132632851600647, "optim/lr": 0.0022344, "optim/total_tokens": 976486400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 4.4689130783081055, "created_at": "2025-01-14T18:15:31.978638+00:00"} {"global_step": 3726, "acc_step": 0, "speed/wps": 12949.438882447625, "speed/FLOPS": 203388732042583.62, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8806861639022827, "optim/lr": 0.002235, "optim/total_tokens": 976748544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458430, "loss/out": 4.604708671569824, "created_at": "2025-01-14T18:15:37.045420+00:00"} {"global_step": 3727, "acc_step": 0, "speed/wps": 12957.838087405567, "speed/FLOPS": 203520653098165.22, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1453574895858765, "optim/lr": 0.0022356, "optim/total_tokens": 977010688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453375, "loss/out": 4.5155110359191895, "created_at": "2025-01-14T18:15:42.106606+00:00"} {"global_step": 3728, "acc_step": 0, "speed/wps": 12953.304951927195, "speed/FLOPS": 203449453976296.47, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2742856740951538, "optim/lr": 0.0022362, "optim/total_tokens": 977272832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424894, "loss/out": 4.709588050842285, "created_at": "2025-01-14T18:15:47.192810+00:00"} {"global_step": 3729, "acc_step": 0, "speed/wps": 12953.307980131705, "speed/FLOPS": 203449501538409.47, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.308817744255066, "optim/lr": 0.0022368, "optim/total_tokens": 977534976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455270, "loss/out": 4.685149192810059, "created_at": "2025-01-14T18:15:52.257632+00:00"} {"global_step": 3730, "acc_step": 0, "speed/wps": 12948.144988957614, "speed/FLOPS": 203368409667327.03, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.001, "optim/grad_norm": 1.153725504875183, "optim/lr": 0.0022374, "optim/total_tokens": 977797120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 4.61872673034668, "created_at": "2025-01-14T18:15:57.323519+00:00"} {"global_step": 3731, "acc_step": 0, "speed/wps": 12954.668888839191, "speed/FLOPS": 203470876479745.1, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1680474281311035, "optim/lr": 0.002238, "optim/total_tokens": 978059264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431842, "loss/out": 4.5730085372924805, "created_at": "2025-01-14T18:16:02.388220+00:00"} {"global_step": 3732, "acc_step": 0, "speed/wps": 12957.161370345972, "speed/FLOPS": 203510024326837.56, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.239070177078247, "optim/lr": 0.0022386, "optim/total_tokens": 978321408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459507, "loss/out": 4.779026508331299, "created_at": "2025-01-14T18:16:07.450451+00:00"} {"global_step": 3733, "acc_step": 0, "speed/wps": 12948.72574543864, "speed/FLOPS": 203377531245904.6, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2315176725387573, "optim/lr": 0.0022392, "optim/total_tokens": 978583552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305035, "loss/out": 4.662691116333008, "created_at": "2025-01-14T18:16:12.515281+00:00"} {"global_step": 3734, "acc_step": 0, "speed/wps": 12958.650730396732, "speed/FLOPS": 203533416773029.94, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.244112253189087, "optim/lr": 0.0022398, "optim/total_tokens": 978845696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 4.645678520202637, "created_at": "2025-01-14T18:16:17.581221+00:00"} {"global_step": 3735, "acc_step": 0, "speed/wps": 12955.199656636345, "speed/FLOPS": 203479212917350.78, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2078287601470947, "optim/lr": 0.0022404, "optim/total_tokens": 979107840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 4.53928279876709, "created_at": "2025-01-14T18:16:22.643701+00:00"} {"global_step": 3736, "acc_step": 0, "speed/wps": 12950.327924803465, "speed/FLOPS": 203402695674454.3, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3792213201522827, "optim/lr": 0.002241, "optim/total_tokens": 979369984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456045, "loss/out": 4.605246543884277, "created_at": "2025-01-14T18:16:27.707908+00:00"} {"global_step": 3737, "acc_step": 0, "speed/wps": 12960.162235127491, "speed/FLOPS": 203557157031851.7, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.226741075515747, "optim/lr": 0.0022416, "optim/total_tokens": 979632128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461889, "loss/out": 4.620452404022217, "created_at": "2025-01-14T18:16:32.768377+00:00"} {"global_step": 3738, "acc_step": 0, "speed/wps": 12954.54769338242, "speed/FLOPS": 203468972938556.47, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.637317657470703, "optim/lr": 0.0022421999999999997, "optim/total_tokens": 979894272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443545, "loss/out": 4.49941873550415, "created_at": "2025-01-14T18:16:37.831498+00:00"} {"global_step": 3739, "acc_step": 0, "speed/wps": 12956.717262864477, "speed/FLOPS": 203503049008573.06, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4117532968521118, "optim/lr": 0.0022428, "optim/total_tokens": 980156416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367376, "loss/out": 4.572971820831299, "created_at": "2025-01-14T18:16:42.894296+00:00"} {"global_step": 3740, "acc_step": 0, "speed/wps": 12956.431289506636, "speed/FLOPS": 203498557404019.6, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.220630168914795, "optim/lr": 0.0022434, "optim/total_tokens": 980418560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 4.663820266723633, "created_at": "2025-01-14T18:16:47.974375+00:00"} {"global_step": 3741, "acc_step": 0, "speed/wps": 12959.133226856331, "speed/FLOPS": 203540995042947.56, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.134661912918091, "optim/lr": 0.002244, "optim/total_tokens": 980680704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444359, "loss/out": 4.651419639587402, "created_at": "2025-01-14T18:16:53.036320+00:00"} {"global_step": 3742, "acc_step": 0, "speed/wps": 12956.26017529251, "speed/FLOPS": 203495869820151.2, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.269403338432312, "optim/lr": 0.0022446, "optim/total_tokens": 980942848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472504, "loss/out": 4.744248390197754, "created_at": "2025-01-14T18:16:58.099882+00:00"} {"global_step": 3743, "acc_step": 0, "speed/wps": 12947.60721136486, "speed/FLOPS": 203359963131249.0, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.177209496498108, "optim/lr": 0.0022451999999999997, "optim/total_tokens": 981204992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453660, "loss/out": 4.707125186920166, "created_at": "2025-01-14T18:17:03.166457+00:00"} {"global_step": 3744, "acc_step": 0, "speed/wps": 12954.042089676084, "speed/FLOPS": 203461031737576.2, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.193265438079834, "optim/lr": 0.0022458, "optim/total_tokens": 981467136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335427, "loss/out": 4.6318678855896, "created_at": "2025-01-14T18:17:08.229342+00:00"} {"global_step": 3745, "acc_step": 0, "speed/wps": 12953.521085830484, "speed/FLOPS": 203452848656246.72, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.349368691444397, "optim/lr": 0.0022464, "optim/total_tokens": 981729280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 391020, "loss/out": 4.6001482009887695, "created_at": "2025-01-14T18:17:13.292807+00:00"} {"global_step": 3746, "acc_step": 0, "speed/wps": 12951.172729781845, "speed/FLOPS": 203415964497523.47, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1823744773864746, "optim/lr": 0.002247, "optim/total_tokens": 981991424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 4.661105155944824, "created_at": "2025-01-14T18:17:18.358976+00:00"} {"global_step": 3747, "acc_step": 0, "speed/wps": 12949.116014505984, "speed/FLOPS": 203383660957893.12, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.513455867767334, "optim/lr": 0.0022475999999999998, "optim/total_tokens": 982253568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444641, "loss/out": 4.706178665161133, "created_at": "2025-01-14T18:17:23.426329+00:00"} {"global_step": 3748, "acc_step": 0, "speed/wps": 12955.195504161098, "speed/FLOPS": 203479147697020.9, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1576175689697266, "optim/lr": 0.0022482, "optim/total_tokens": 982515712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 4.727390289306641, "created_at": "2025-01-14T18:17:28.488895+00:00"} {"global_step": 3749, "acc_step": 0, "speed/wps": 12950.37753611105, "speed/FLOPS": 203403474888213.0, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5607962608337402, "optim/lr": 0.0022488, "optim/total_tokens": 982777856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432941, "loss/out": 4.6860504150390625, "created_at": "2025-01-14T18:17:33.554174+00:00"} {"global_step": 3750, "acc_step": 0, "speed/wps": 12953.902847099229, "speed/FLOPS": 203458844741567.3, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2583718299865723, "optim/lr": 0.0022494, "optim/total_tokens": 983040000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330478, "loss/out": 4.643374443054199, "created_at": "2025-01-14T18:17:38.617116+00:00"} {"global_step": 3751, "acc_step": 0, "speed/wps": 12962.894367519024, "speed/FLOPS": 203600068925404.03, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.2678439617156982, "optim/lr": 0.0022500000000000003, "optim/total_tokens": 983302144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 4.683685302734375, "created_at": "2025-01-14T18:17:43.676834+00:00"} {"global_step": 3752, "acc_step": 0, "speed/wps": 12961.818869405888, "speed/FLOPS": 203583176749647.97, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7477363348007202, "optim/lr": 0.0022506, "optim/total_tokens": 983564288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 4.610996723175049, "created_at": "2025-01-14T18:17:48.737474+00:00"} {"global_step": 3753, "acc_step": 0, "speed/wps": 12956.862215340825, "speed/FLOPS": 203505325686399.7, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.140677571296692, "optim/lr": 0.0022512, "optim/total_tokens": 983826432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448361, "loss/out": 4.634852409362793, "created_at": "2025-01-14T18:17:53.801379+00:00"} {"global_step": 3754, "acc_step": 0, "speed/wps": 12954.07439029912, "speed/FLOPS": 203461539063246.66, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.157853364944458, "optim/lr": 0.0022518, "optim/total_tokens": 984088576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473073, "loss/out": 4.667214393615723, "created_at": "2025-01-14T18:17:58.864998+00:00"} {"global_step": 3755, "acc_step": 0, "speed/wps": 12954.02970125543, "speed/FLOPS": 203460837160406.3, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3911855220794678, "optim/lr": 0.0022524000000000003, "optim/total_tokens": 984350720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457902, "loss/out": 4.744215965270996, "created_at": "2025-01-14T18:18:03.928136+00:00"} {"global_step": 3756, "acc_step": 0, "speed/wps": 12954.954908423355, "speed/FLOPS": 203475368810346.47, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2806625366210938, "optim/lr": 0.0022530000000000002, "optim/total_tokens": 984612864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429193, "loss/out": 4.637065887451172, "created_at": "2025-01-14T18:18:08.993132+00:00"} {"global_step": 3757, "acc_step": 0, "speed/wps": 12946.507709020329, "speed/FLOPS": 203342693936053.56, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2744516134262085, "optim/lr": 0.0022536, "optim/total_tokens": 984875008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364976, "loss/out": 4.713211536407471, "created_at": "2025-01-14T18:18:14.060626+00:00"} {"global_step": 3758, "acc_step": 0, "speed/wps": 12950.012471941049, "speed/FLOPS": 203397741053772.4, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.257473111152649, "optim/lr": 0.0022542, "optim/total_tokens": 985137152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 4.75929594039917, "created_at": "2025-01-14T18:18:19.134714+00:00"} {"global_step": 3759, "acc_step": 0, "speed/wps": 12946.40716220654, "speed/FLOPS": 203341114710177.22, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0014, "optim/grad_norm": 1.1424320936203003, "optim/lr": 0.0022548000000000004, "optim/total_tokens": 985399296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438240, "loss/out": 4.601628303527832, "created_at": "2025-01-14T18:18:24.202491+00:00"} {"global_step": 3760, "acc_step": 0, "speed/wps": 12949.823782411318, "speed/FLOPS": 203394777425421.4, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.294477105140686, "optim/lr": 0.0022554000000000003, "optim/total_tokens": 985661440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442208, "loss/out": 4.602420806884766, "created_at": "2025-01-14T18:18:29.267195+00:00"} {"global_step": 3761, "acc_step": 0, "speed/wps": 12951.746937737475, "speed/FLOPS": 203424983222511.06, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3926887512207031, "optim/lr": 0.002256, "optim/total_tokens": 985923584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299756, "loss/out": 4.659891128540039, "created_at": "2025-01-14T18:18:34.330704+00:00"} {"global_step": 3762, "acc_step": 0, "speed/wps": 12953.50744416604, "speed/FLOPS": 203452634395161.0, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.285111427307129, "optim/lr": 0.0022566, "optim/total_tokens": 986185728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477030, "loss/out": 4.5585126876831055, "created_at": "2025-01-14T18:18:39.393715+00:00"} {"global_step": 3763, "acc_step": 0, "speed/wps": 12955.781384858763, "speed/FLOPS": 203488349758462.2, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6482685804367065, "optim/lr": 0.0022572, "optim/total_tokens": 986447872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 4.657195091247559, "created_at": "2025-01-14T18:18:44.456082+00:00"} {"global_step": 3764, "acc_step": 0, "speed/wps": 12950.655548552719, "speed/FLOPS": 203407841455638.56, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.351019024848938, "optim/lr": 0.0022578000000000003, "optim/total_tokens": 986710016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449135, "loss/out": 4.579231262207031, "created_at": "2025-01-14T18:18:49.520143+00:00"} {"global_step": 3765, "acc_step": 0, "speed/wps": 12952.75933312061, "speed/FLOPS": 203440884283180.22, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.203998327255249, "optim/lr": 0.0022584000000000003, "optim/total_tokens": 986972160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454182, "loss/out": 4.699708461761475, "created_at": "2025-01-14T18:18:54.583723+00:00"} {"global_step": 3766, "acc_step": 0, "speed/wps": 12962.570338762725, "speed/FLOPS": 203594979608524.2, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.226988434791565, "optim/lr": 0.002259, "optim/total_tokens": 987234304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314942, "loss/out": 4.521583080291748, "created_at": "2025-01-14T18:18:59.648167+00:00"} {"global_step": 3767, "acc_step": 0, "speed/wps": 12950.5782160127, "speed/FLOPS": 203406626841831.53, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1281200647354126, "optim/lr": 0.0022596, "optim/total_tokens": 987496448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475458, "loss/out": 4.649691104888916, "created_at": "2025-01-14T18:19:04.715581+00:00"} {"global_step": 3768, "acc_step": 0, "speed/wps": 12957.39400749808, "speed/FLOPS": 203513678212988.38, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5007548332214355, "optim/lr": 0.0022602, "optim/total_tokens": 987758592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 4.697025299072266, "created_at": "2025-01-14T18:19:09.778104+00:00"} {"global_step": 3769, "acc_step": 0, "speed/wps": 12958.636712810347, "speed/FLOPS": 203533196607573.6, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2594425678253174, "optim/lr": 0.0022608000000000003, "optim/total_tokens": 988020736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 4.604737281799316, "created_at": "2025-01-14T18:19:14.838913+00:00"} {"global_step": 3770, "acc_step": 0, "speed/wps": 12953.510827753964, "speed/FLOPS": 203452687539058.9, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3396098613739014, "optim/lr": 0.0022614, "optim/total_tokens": 988282880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426544, "loss/out": 4.703132152557373, "created_at": "2025-01-14T18:19:19.909034+00:00"} {"global_step": 3771, "acc_step": 0, "speed/wps": 12948.209371930718, "speed/FLOPS": 203369420890391.1, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8176841735839844, "optim/lr": 0.002262, "optim/total_tokens": 988545024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438251, "loss/out": 4.593035697937012, "created_at": "2025-01-14T18:19:24.976854+00:00"} {"global_step": 3772, "acc_step": 0, "speed/wps": 12959.18978615625, "speed/FLOPS": 203541883384473.4, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4350228309631348, "optim/lr": 0.0022626, "optim/total_tokens": 988807168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313754, "loss/out": 4.578393936157227, "created_at": "2025-01-14T18:19:30.038033+00:00"} {"global_step": 3773, "acc_step": 0, "speed/wps": 12955.270115555195, "speed/FLOPS": 203480319571492.0, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3250586986541748, "optim/lr": 0.0022632, "optim/total_tokens": 989069312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479178, "loss/out": 4.518569469451904, "created_at": "2025-01-14T18:19:35.104850+00:00"} {"global_step": 3774, "acc_step": 0, "speed/wps": 12962.162974333776, "speed/FLOPS": 203588581390390.16, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7082980871200562, "optim/lr": 0.0022638000000000003, "optim/total_tokens": 989331456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 4.7075090408325195, "created_at": "2025-01-14T18:19:40.169042+00:00"} {"global_step": 3775, "acc_step": 0, "speed/wps": 12954.989956385913, "speed/FLOPS": 203475919286759.38, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.4046212434768677, "optim/lr": 0.0022644, "optim/total_tokens": 989593600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448606, "loss/out": 4.585680961608887, "created_at": "2025-01-14T18:19:45.234951+00:00"} {"global_step": 3776, "acc_step": 0, "speed/wps": 12954.204324133578, "speed/FLOPS": 203463579852664.62, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3701492547988892, "optim/lr": 0.002265, "optim/total_tokens": 989855744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430506, "loss/out": 4.612112522125244, "created_at": "2025-01-14T18:19:50.298829+00:00"} {"global_step": 3777, "acc_step": 0, "speed/wps": 12955.361101976516, "speed/FLOPS": 203481748638267.28, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5033289194107056, "optim/lr": 0.0022656, "optim/total_tokens": 990117888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355107, "loss/out": 4.558953762054443, "created_at": "2025-01-14T18:19:55.361193+00:00"} {"global_step": 3778, "acc_step": 0, "speed/wps": 12954.802559912461, "speed/FLOPS": 203472975967634.1, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.001, "optim/grad_norm": 1.5488190650939941, "optim/lr": 0.0022662, "optim/total_tokens": 990380032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464298, "loss/out": 4.609216690063477, "created_at": "2025-01-14T18:20:00.427872+00:00"} {"global_step": 3779, "acc_step": 0, "speed/wps": 12951.448159031732, "speed/FLOPS": 203420290492372.4, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4228475093841553, "optim/lr": 0.0022668000000000002, "optim/total_tokens": 990642176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 4.509882926940918, "created_at": "2025-01-14T18:20:05.497948+00:00"} {"global_step": 3780, "acc_step": 0, "speed/wps": 12953.030623932355, "speed/FLOPS": 203445145278171.3, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1471937894821167, "optim/lr": 0.0022674, "optim/total_tokens": 990904320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470403, "loss/out": 4.5676069259643555, "created_at": "2025-01-14T18:20:10.562667+00:00"} {"global_step": 3781, "acc_step": 0, "speed/wps": 12957.99225397395, "speed/FLOPS": 203523074495968.72, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1897988319396973, "optim/lr": 0.002268, "optim/total_tokens": 991166464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 4.557445526123047, "created_at": "2025-01-14T18:20:15.625815+00:00"} {"global_step": 3782, "acc_step": 0, "speed/wps": 12955.80745411985, "speed/FLOPS": 203488759212030.47, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1218223571777344, "optim/lr": 0.0022686, "optim/total_tokens": 991428608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442208, "loss/out": 4.480410575866699, "created_at": "2025-01-14T18:20:20.693557+00:00"} {"global_step": 3783, "acc_step": 0, "speed/wps": 12955.687186704059, "speed/FLOPS": 203486870247000.72, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2476341724395752, "optim/lr": 0.0022692, "optim/total_tokens": 991690752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316527, "loss/out": 4.554447174072266, "created_at": "2025-01-14T18:20:25.757058+00:00"} {"global_step": 3784, "acc_step": 0, "speed/wps": 12955.4395411351, "speed/FLOPS": 203482980633039.56, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.680298089981079, "optim/lr": 0.0022698, "optim/total_tokens": 991952896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472260, "loss/out": 4.699032783508301, "created_at": "2025-01-14T18:20:30.820659+00:00"} {"global_step": 3785, "acc_step": 0, "speed/wps": 12947.679920034643, "speed/FLOPS": 203361105120820.1, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4110517501831055, "optim/lr": 0.0022704, "optim/total_tokens": 992215040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495052, "loss/out": 4.600330829620361, "created_at": "2025-01-14T18:20:35.885873+00:00"} {"global_step": 3786, "acc_step": 0, "speed/wps": 12958.506740158125, "speed/FLOPS": 203531155208469.3, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1597893238067627, "optim/lr": 0.002271, "optim/total_tokens": 992477184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434515, "loss/out": 4.655276298522949, "created_at": "2025-01-14T18:20:40.947976+00:00"} {"global_step": 3787, "acc_step": 0, "speed/wps": 12957.080984993163, "speed/FLOPS": 203508761764410.66, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2640578746795654, "optim/lr": 0.0022716, "optim/total_tokens": 992739328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 4.701748371124268, "created_at": "2025-01-14T18:20:46.010582+00:00"} {"global_step": 3788, "acc_step": 0, "speed/wps": 12943.601161136534, "speed/FLOPS": 203297042607523.22, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5372836589813232, "optim/lr": 0.0022722, "optim/total_tokens": 993001472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297001, "loss/out": 4.599131107330322, "created_at": "2025-01-14T18:20:51.082507+00:00"} {"global_step": 3789, "acc_step": 0, "speed/wps": 12952.291648760169, "speed/FLOPS": 203433538657633.84, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3986684083938599, "optim/lr": 0.0022728, "optim/total_tokens": 993263616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468777, "loss/out": 4.647385597229004, "created_at": "2025-01-14T18:20:56.148458+00:00"} {"global_step": 3790, "acc_step": 0, "speed/wps": 12952.355279084464, "speed/FLOPS": 203434538059314.12, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3565430641174316, "optim/lr": 0.0022734, "optim/total_tokens": 993525760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 4.581731796264648, "created_at": "2025-01-14T18:21:01.212305+00:00"} {"global_step": 3791, "acc_step": 0, "speed/wps": 12958.892694849264, "speed/FLOPS": 203537217157250.97, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.392885446548462, "optim/lr": 0.002274, "optim/total_tokens": 993787904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431842, "loss/out": 4.6752166748046875, "created_at": "2025-01-14T18:21:06.279008+00:00"} {"global_step": 3792, "acc_step": 0, "speed/wps": 12956.713333669042, "speed/FLOPS": 203502987295159.5, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.496439814567566, "optim/lr": 0.0022746, "optim/total_tokens": 994050048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447554, "loss/out": 4.774661540985107, "created_at": "2025-01-14T18:21:11.341183+00:00"} {"global_step": 3793, "acc_step": 0, "speed/wps": 12952.123692978048, "speed/FLOPS": 203430900681280.22, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9169715642929077, "optim/lr": 0.0022752, "optim/total_tokens": 994312192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313070, "loss/out": 4.5567626953125, "created_at": "2025-01-14T18:21:16.404671+00:00"} {"global_step": 3794, "acc_step": 0, "speed/wps": 12952.116905797446, "speed/FLOPS": 203430794079282.28, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2347240447998047, "optim/lr": 0.0022758, "optim/total_tokens": 994574336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463248, "loss/out": 4.6729841232299805, "created_at": "2025-01-14T18:21:21.468159+00:00"} {"global_step": 3795, "acc_step": 0, "speed/wps": 12953.468758364623, "speed/FLOPS": 203452026781489.9, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5325732231140137, "optim/lr": 0.0022764, "optim/total_tokens": 994836480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 4.580097198486328, "created_at": "2025-01-14T18:21:26.532699+00:00"} {"global_step": 3796, "acc_step": 0, "speed/wps": 12952.524781017546, "speed/FLOPS": 203437200320093.25, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.026423215866089, "optim/lr": 0.002277, "optim/total_tokens": 995098624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 4.626712322235107, "created_at": "2025-01-14T18:21:31.597000+00:00"} {"global_step": 3797, "acc_step": 0, "speed/wps": 12946.855248642407, "speed/FLOPS": 203348152523387.78, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6635682582855225, "optim/lr": 0.0022776, "optim/total_tokens": 995360768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445404, "loss/out": 4.70155143737793, "created_at": "2025-01-14T18:21:36.662494+00:00"} {"global_step": 3798, "acc_step": 0, "speed/wps": 12954.48134902405, "speed/FLOPS": 203467930909241.6, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.395220994949341, "optim/lr": 0.0022781999999999998, "optim/total_tokens": 995622912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461605, "loss/out": 4.70876932144165, "created_at": "2025-01-14T18:21:41.726629+00:00"} {"global_step": 3799, "acc_step": 0, "speed/wps": 12951.702650930083, "speed/FLOPS": 203424287637344.5, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4651405811309814, "optim/lr": 0.0022788, "optim/total_tokens": 995885056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325819, "loss/out": 4.7038960456848145, "created_at": "2025-01-14T18:21:46.790613+00:00"} {"global_step": 3800, "acc_step": 0, "speed/wps": 12952.174603019288, "speed/FLOPS": 203431700293435.62, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0014, "optim/grad_norm": 1.6564847230911255, "optim/lr": 0.0022794, "optim/total_tokens": 996147200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415046, "loss/out": 4.577022552490234, "created_at": "2025-01-14T18:21:51.854258+00:00"} {"global_step": 3801, "acc_step": 0, "speed/wps": 12956.915367064801, "speed/FLOPS": 203506160507262.88, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.857873797416687, "optim/lr": 0.00228, "optim/total_tokens": 996409344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 4.574592113494873, "created_at": "2025-01-14T18:21:56.915876+00:00"} {"global_step": 3802, "acc_step": 0, "speed/wps": 12953.088799008658, "speed/FLOPS": 203446058997685.6, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3192644119262695, "optim/lr": 0.0022806, "optim/total_tokens": 996671488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434515, "loss/out": 4.788237571716309, "created_at": "2025-01-14T18:22:01.981417+00:00"} {"global_step": 3803, "acc_step": 0, "speed/wps": 12952.189500629407, "speed/FLOPS": 203431934280874.0, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5764882564544678, "optim/lr": 0.0022812, "optim/total_tokens": 996933632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 4.6418538093566895, "created_at": "2025-01-14T18:22:07.046755+00:00"} {"global_step": 3804, "acc_step": 0, "speed/wps": 12951.687210929615, "speed/FLOPS": 203424045130920.38, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.5814987421035767, "optim/lr": 0.0022818, "optim/total_tokens": 997195776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326583, "loss/out": 4.783454418182373, "created_at": "2025-01-14T18:22:12.111260+00:00"} {"global_step": 3805, "acc_step": 0, "speed/wps": 12952.724297226294, "speed/FLOPS": 203440333996315.62, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7400727272033691, "optim/lr": 0.0022824, "optim/total_tokens": 997457920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456045, "loss/out": 4.739587783813477, "created_at": "2025-01-14T18:22:17.177537+00:00"} {"global_step": 3806, "acc_step": 0, "speed/wps": 12955.087841257702, "speed/FLOPS": 203477456703178.75, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8096222877502441, "optim/lr": 0.002283, "optim/total_tokens": 997720064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 4.750861644744873, "created_at": "2025-01-14T18:22:22.241559+00:00"} {"global_step": 3807, "acc_step": 0, "speed/wps": 12952.128487158727, "speed/FLOPS": 203430975980475.66, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4299979209899902, "optim/lr": 0.0022836, "optim/total_tokens": 997982208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441679, "loss/out": 4.66285514831543, "created_at": "2025-01-14T18:22:27.309744+00:00"} {"global_step": 3808, "acc_step": 0, "speed/wps": 12957.88519121428, "speed/FLOPS": 203521392928208.47, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4122687578201294, "optim/lr": 0.0022842, "optim/total_tokens": 998244352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 4.616608142852783, "created_at": "2025-01-14T18:22:32.374554+00:00"} {"global_step": 3809, "acc_step": 0, "speed/wps": 12956.42215519021, "speed/FLOPS": 203498413937028.28, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5490312576293945, "optim/lr": 0.0022848, "optim/total_tokens": 998506496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429454, "loss/out": 4.586349010467529, "created_at": "2025-01-14T18:22:37.436843+00:00"} {"global_step": 3810, "acc_step": 0, "speed/wps": 12950.199603544712, "speed/FLOPS": 203400680212753.72, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7128210067749023, "optim/lr": 0.0022854, "optim/total_tokens": 998768640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288973, "loss/out": 4.593209266662598, "created_at": "2025-01-14T18:22:42.502363+00:00"} {"global_step": 3811, "acc_step": 0, "speed/wps": 12949.635198468735, "speed/FLOPS": 203391815455461.66, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8674657344818115, "optim/lr": 0.0022860000000000003, "optim/total_tokens": 999030784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 4.572671413421631, "created_at": "2025-01-14T18:22:47.566816+00:00"} {"global_step": 3812, "acc_step": 0, "speed/wps": 12957.741626001549, "speed/FLOPS": 203519138039263.25, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3774678707122803, "optim/lr": 0.0022866, "optim/total_tokens": 999292928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 4.600211143493652, "created_at": "2025-01-14T18:22:52.630851+00:00"} {"global_step": 3813, "acc_step": 0, "speed/wps": 12958.20387450084, "speed/FLOPS": 203526398287140.6, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3658874034881592, "optim/lr": 0.0022872, "optim/total_tokens": 999555072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454481, "loss/out": 4.658444404602051, "created_at": "2025-01-14T18:22:57.696156+00:00"} {"global_step": 3814, "acc_step": 0, "speed/wps": 12957.234365299892, "speed/FLOPS": 203511170812894.47, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.340396761894226, "optim/lr": 0.0022878, "optim/total_tokens": 999817216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 4.577974796295166, "created_at": "2025-01-14T18:23:02.757523+00:00"} {"global_step": 3815, "acc_step": 0, "speed/wps": 12961.383328450422, "speed/FLOPS": 203576335980446.38, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5731472969055176, "optim/lr": 0.0022884000000000003, "optim/total_tokens": 1000079360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423864, "loss/out": 4.554727554321289, "created_at": "2025-01-14T18:23:07.822560+00:00"} {"global_step": 3816, "acc_step": 0, "speed/wps": 12958.960007972917, "speed/FLOPS": 203538274402355.84, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0219526290893555, "optim/lr": 0.0022890000000000002, "optim/total_tokens": 1000341504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319751, "loss/out": 4.55194616317749, "created_at": "2025-01-14T18:23:12.884812+00:00"} {"global_step": 3817, "acc_step": 0, "speed/wps": 12959.803218619689, "speed/FLOPS": 203551518184256.25, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2691307067871094, "optim/lr": 0.0022896, "optim/total_tokens": 1000603648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 4.687202453613281, "created_at": "2025-01-14T18:23:17.950348+00:00"} {"global_step": 3818, "acc_step": 0, "speed/wps": 12960.174766562293, "speed/FLOPS": 203557353855255.25, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6117507219314575, "optim/lr": 0.0022902, "optim/total_tokens": 1000865792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 4.577998161315918, "created_at": "2025-01-14T18:23:23.012508+00:00"} {"global_step": 3819, "acc_step": 0, "speed/wps": 12958.17502282323, "speed/FLOPS": 203525945131897.4, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.6290264129638672, "optim/lr": 0.0022908, "optim/total_tokens": 1001127936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433179, "loss/out": 4.638729572296143, "created_at": "2025-01-14T18:23:28.077159+00:00"} {"global_step": 3820, "acc_step": 0, "speed/wps": 12954.699523139789, "speed/FLOPS": 203471357633531.9, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.048903226852417, "optim/lr": 0.0022914000000000003, "optim/total_tokens": 1001390080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459507, "loss/out": 4.664795875549316, "created_at": "2025-01-14T18:23:33.145678+00:00"} {"global_step": 3821, "acc_step": 0, "speed/wps": 12956.591327014705, "speed/FLOPS": 203501071012998.4, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5390355587005615, "optim/lr": 0.002292, "optim/total_tokens": 1001652224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323417, "loss/out": 4.559889793395996, "created_at": "2025-01-14T18:23:38.209332+00:00"} {"global_step": 3822, "acc_step": 0, "speed/wps": 12955.989848854373, "speed/FLOPS": 203491623971972.78, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.324617624282837, "optim/lr": 0.0022926, "optim/total_tokens": 1001914368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320284, "loss/out": 4.601858615875244, "created_at": "2025-01-14T18:23:43.271270+00:00"} {"global_step": 3823, "acc_step": 0, "speed/wps": 12953.8495388576, "speed/FLOPS": 203458007462379.3, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8979283571243286, "optim/lr": 0.0022932, "optim/total_tokens": 1002176512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466183, "loss/out": 4.686366081237793, "created_at": "2025-01-14T18:23:48.339773+00:00"} {"global_step": 3824, "acc_step": 0, "speed/wps": 12960.079988938684, "speed/FLOPS": 203555865242438.34, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1878206729888916, "optim/lr": 0.0022938, "optim/total_tokens": 1002438656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 4.682599067687988, "created_at": "2025-01-14T18:23:53.400917+00:00"} {"global_step": 3825, "acc_step": 0, "speed/wps": 12959.807144915781, "speed/FLOPS": 203551579852131.7, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.6748912334442139, "optim/lr": 0.0022944000000000003, "optim/total_tokens": 1002700800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426829, "loss/out": 4.675583839416504, "created_at": "2025-01-14T18:23:58.467054+00:00"} {"global_step": 3826, "acc_step": 0, "speed/wps": 12953.680746230151, "speed/FLOPS": 203455356342219.88, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.390985131263733, "optim/lr": 0.002295, "optim/total_tokens": 1002962944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447026, "loss/out": 4.586085319519043, "created_at": "2025-01-14T18:24:03.531273+00:00"} {"global_step": 3827, "acc_step": 0, "speed/wps": 12958.126583028175, "speed/FLOPS": 203525184318352.8, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3115034103393555, "optim/lr": 0.0022956, "optim/total_tokens": 1003225088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302050, "loss/out": 4.5321784019470215, "created_at": "2025-01-14T18:24:08.596494+00:00"} {"global_step": 3828, "acc_step": 0, "speed/wps": 12954.911935276967, "speed/FLOPS": 203474693858031.16, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3796182870864868, "optim/lr": 0.0022962, "optim/total_tokens": 1003487232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 4.666533470153809, "created_at": "2025-01-14T18:24:13.662327+00:00"} {"global_step": 3829, "acc_step": 0, "speed/wps": 12957.35374048367, "speed/FLOPS": 203513045764194.4, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3452717065811157, "optim/lr": 0.0022968, "optim/total_tokens": 1003749376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 4.629149436950684, "created_at": "2025-01-14T18:24:18.723675+00:00"} {"global_step": 3830, "acc_step": 0, "speed/wps": 12957.4421101712, "speed/FLOPS": 203514433731569.72, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8670859336853027, "optim/lr": 0.0022974000000000002, "optim/total_tokens": 1004011520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460578, "loss/out": 4.606456756591797, "created_at": "2025-01-14T18:24:23.790690+00:00"} {"global_step": 3831, "acc_step": 0, "speed/wps": 12945.626385854588, "speed/FLOPS": 203328851544670.66, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.254647135734558, "optim/lr": 0.002298, "optim/total_tokens": 1004273664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330862, "loss/out": 4.645898818969727, "created_at": "2025-01-14T18:24:28.859027+00:00"} {"global_step": 3832, "acc_step": 0, "speed/wps": 12955.215720372373, "speed/FLOPS": 203479465220398.97, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.001, "optim/grad_norm": 2.1939077377319336, "optim/lr": 0.0022986, "optim/total_tokens": 1004535808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371444, "loss/out": 4.6909685134887695, "created_at": "2025-01-14T18:24:33.922010+00:00"} {"global_step": 3833, "acc_step": 0, "speed/wps": 12958.310462889354, "speed/FLOPS": 203528072404250.03, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4100111722946167, "optim/lr": 0.0022992, "optim/total_tokens": 1004797952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464298, "loss/out": 4.6660919189453125, "created_at": "2025-01-14T18:24:38.987695+00:00"} {"global_step": 3834, "acc_step": 0, "speed/wps": 12956.644941716439, "speed/FLOPS": 203501913105563.47, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2710530757904053, "optim/lr": 0.0022998, "optim/total_tokens": 1005060096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 4.617088794708252, "created_at": "2025-01-14T18:24:44.050402+00:00"} {"global_step": 3835, "acc_step": 0, "speed/wps": 12950.612935220115, "speed/FLOPS": 203407172154693.16, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.2626967430114746, "optim/lr": 0.0023004, "optim/total_tokens": 1005322240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447554, "loss/out": 4.683246612548828, "created_at": "2025-01-14T18:24:49.117895+00:00"} {"global_step": 3836, "acc_step": 0, "speed/wps": 12953.304515579732, "speed/FLOPS": 203449447122859.94, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5234655141830444, "optim/lr": 0.002301, "optim/total_tokens": 1005584384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 4.6376848220825195, "created_at": "2025-01-14T18:24:54.185519+00:00"} {"global_step": 3837, "acc_step": 0, "speed/wps": 12958.617509673053, "speed/FLOPS": 203532894995914.5, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.318229079246521, "optim/lr": 0.0023016, "optim/total_tokens": 1005846528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328221, "loss/out": 4.575292587280273, "created_at": "2025-01-14T18:24:59.251412+00:00"} {"global_step": 3838, "acc_step": 0, "speed/wps": 12961.543696025388, "speed/FLOPS": 203578854773579.34, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3938984870910645, "optim/lr": 0.0023022, "optim/total_tokens": 1006108672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425179, "loss/out": 4.540832996368408, "created_at": "2025-01-14T18:25:04.311154+00:00"} {"global_step": 3839, "acc_step": 0, "speed/wps": 12956.346552457675, "speed/FLOPS": 203497226492224.78, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.9046680927276611, "optim/lr": 0.0023028, "optim/total_tokens": 1006370816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 4.598214149475098, "created_at": "2025-01-14T18:25:09.373964+00:00"} {"global_step": 3840, "acc_step": 0, "speed/wps": 12953.485022873667, "speed/FLOPS": 203452282237954.22, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0011, "optim/grad_norm": 2.0244593620300293, "optim/lr": 0.0023034, "optim/total_tokens": 1006632960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439820, "loss/out": 4.586877822875977, "created_at": "2025-01-14T18:25:14.438608+00:00"} {"global_step": 3841, "acc_step": 0, "speed/wps": 12953.833478509367, "speed/FLOPS": 203457755212541.1, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.9479796886444092, "optim/lr": 0.002304, "optim/total_tokens": 1006895104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 4.594348430633545, "created_at": "2025-01-14T18:25:19.510127+00:00"} {"global_step": 3842, "acc_step": 0, "speed/wps": 12951.149437015625, "speed/FLOPS": 203415598652622.38, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9923266172409058, "optim/lr": 0.0023046, "optim/total_tokens": 1007157248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324322, "loss/out": 4.664270877838135, "created_at": "2025-01-14T18:25:24.575370+00:00"} {"global_step": 3843, "acc_step": 0, "speed/wps": 12945.308804339087, "speed/FLOPS": 203323863490567.0, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6719937324523926, "optim/lr": 0.0023052, "optim/total_tokens": 1007419392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 406484, "loss/out": 4.674205303192139, "created_at": "2025-01-14T18:25:29.642761+00:00"} {"global_step": 3844, "acc_step": 0, "speed/wps": 12956.65453916226, "speed/FLOPS": 203502063846638.38, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.432531476020813, "optim/lr": 0.0023057999999999998, "optim/total_tokens": 1007681536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 4.636080265045166, "created_at": "2025-01-14T18:25:34.708962+00:00"} {"global_step": 3845, "acc_step": 0, "speed/wps": 12958.199023324822, "speed/FLOPS": 203526322092755.03, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7613282203674316, "optim/lr": 0.0023064, "optim/total_tokens": 1007943680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 4.713173866271973, "created_at": "2025-01-14T18:25:39.774101+00:00"} {"global_step": 3846, "acc_step": 0, "speed/wps": 12958.873104346305, "speed/FLOPS": 203536909461481.8, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.320797920227051, "optim/lr": 0.002307, "optim/total_tokens": 1008205824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 4.618816375732422, "created_at": "2025-01-14T18:25:44.834942+00:00"} {"global_step": 3847, "acc_step": 0, "speed/wps": 12953.004565844958, "speed/FLOPS": 203444736000101.2, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.184856414794922, "optim/lr": 0.0023076, "optim/total_tokens": 1008467968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435567, "loss/out": 4.840738296508789, "created_at": "2025-01-14T18:25:49.900427+00:00"} {"global_step": 3848, "acc_step": 0, "speed/wps": 12955.817205371928, "speed/FLOPS": 203488912368844.0, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.6850025653839111, "optim/lr": 0.0023082, "optim/total_tokens": 1008730112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350412, "loss/out": 4.682117938995361, "created_at": "2025-01-14T18:25:54.963157+00:00"} {"global_step": 3849, "acc_step": 0, "speed/wps": 12950.481045169652, "speed/FLOPS": 203405100640214.72, "speed/curr_iter_time": 1.2771, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6768133640289307, "optim/lr": 0.0023087999999999997, "optim/total_tokens": 1008992256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402512, "loss/out": 4.758278846740723, "created_at": "2025-01-14T18:26:00.027385+00:00"} {"global_step": 3850, "acc_step": 0, "speed/wps": 12955.19175954182, "speed/FLOPS": 203479088882629.5, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5620338916778564, "optim/lr": 0.0023094, "optim/total_tokens": 1009254400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468255, "loss/out": 4.689561367034912, "created_at": "2025-01-14T18:26:05.111142+00:00"} {"global_step": 3851, "acc_step": 0, "speed/wps": 12959.753714077731, "speed/FLOPS": 203550740647398.44, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6800404787063599, "optim/lr": 0.00231, "optim/total_tokens": 1009516544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453375, "loss/out": 4.778379917144775, "created_at": "2025-01-14T18:26:10.172222+00:00"} {"global_step": 3852, "acc_step": 0, "speed/wps": 12956.456621855592, "speed/FLOPS": 203498955283370.25, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.240001678466797, "optim/lr": 0.0023106, "optim/total_tokens": 1009778688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445166, "loss/out": 4.648202896118164, "created_at": "2025-01-14T18:26:15.235209+00:00"} {"global_step": 3853, "acc_step": 0, "speed/wps": 12956.895679953213, "speed/FLOPS": 203505851294122.44, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.393308401107788, "optim/lr": 0.0023112, "optim/total_tokens": 1010040832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299119, "loss/out": 4.637218952178955, "created_at": "2025-01-14T18:26:20.299025+00:00"} {"global_step": 3854, "acc_step": 0, "speed/wps": 12955.084580869381, "speed/FLOPS": 203477405494299.94, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9931936264038086, "optim/lr": 0.0023117999999999997, "optim/total_tokens": 1010302976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 4.628526210784912, "created_at": "2025-01-14T18:26:25.361457+00:00"} {"global_step": 3855, "acc_step": 0, "speed/wps": 12954.102274606237, "speed/FLOPS": 203461977024607.0, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.728358030319214, "optim/lr": 0.0023124, "optim/total_tokens": 1010565120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 4.752053260803223, "created_at": "2025-01-14T18:26:30.424642+00:00"} {"global_step": 3856, "acc_step": 0, "speed/wps": 12947.433068844484, "speed/FLOPS": 203357227983669.6, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.3040683269500732, "optim/lr": 0.002313, "optim/total_tokens": 1010827264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456045, "loss/out": 4.732285499572754, "created_at": "2025-01-14T18:26:35.494072+00:00"} {"global_step": 3857, "acc_step": 0, "speed/wps": 12957.599013595382, "speed/FLOPS": 203516898115454.4, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1893911361694336, "optim/lr": 0.0023136, "optim/total_tokens": 1011089408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424393, "loss/out": 4.844090461730957, "created_at": "2025-01-14T18:26:40.563638+00:00"} {"global_step": 3858, "acc_step": 0, "speed/wps": 12954.28169998537, "speed/FLOPS": 203464795146742.56, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.2331433296203613, "optim/lr": 0.0023141999999999998, "optim/total_tokens": 1011351552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370300, "loss/out": 4.629655361175537, "created_at": "2025-01-14T18:26:45.636557+00:00"} {"global_step": 3859, "acc_step": 0, "speed/wps": 12952.814530237985, "speed/FLOPS": 203441751229757.4, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.8715038299560547, "optim/lr": 0.0023148, "optim/total_tokens": 1011613696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 4.717662811279297, "created_at": "2025-01-14T18:26:50.700898+00:00"} {"global_step": 3860, "acc_step": 0, "speed/wps": 12953.44705731449, "speed/FLOPS": 203451685936677.2, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.34720516204834, "optim/lr": 0.0023154, "optim/total_tokens": 1011875840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437717, "loss/out": 4.732826232910156, "created_at": "2025-01-14T18:26:55.766933+00:00"} {"global_step": 3861, "acc_step": 0, "speed/wps": 12956.674478420286, "speed/FLOPS": 203502377020085.06, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.925532102584839, "optim/lr": 0.002316, "optim/total_tokens": 1012137984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 4.723873138427734, "created_at": "2025-01-14T18:27:00.831479+00:00"} {"global_step": 3862, "acc_step": 0, "speed/wps": 12950.63857611901, "speed/FLOPS": 203407574880244.6, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.020087242126465, "optim/lr": 0.0023166000000000003, "optim/total_tokens": 1012400128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440921, "loss/out": 4.693332672119141, "created_at": "2025-01-14T18:27:05.896289+00:00"} {"global_step": 3863, "acc_step": 0, "speed/wps": 12958.618051378768, "speed/FLOPS": 203532903504147.12, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.0494754314422607, "optim/lr": 0.0023172, "optim/total_tokens": 1012662272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337583, "loss/out": 4.749546051025391, "created_at": "2025-01-14T18:27:10.958514+00:00"} {"global_step": 3864, "acc_step": 0, "speed/wps": 12950.975108623434, "speed/FLOPS": 203412860585670.7, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7309006452560425, "optim/lr": 0.0023178, "optim/total_tokens": 1012924416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423841, "loss/out": 4.754140853881836, "created_at": "2025-01-14T18:27:16.027185+00:00"} {"global_step": 3865, "acc_step": 0, "speed/wps": 12959.60540798003, "speed/FLOPS": 203548411296340.12, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.1476192474365234, "optim/lr": 0.0023184, "optim/total_tokens": 1013186560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 4.656505107879639, "created_at": "2025-01-14T18:27:21.093497+00:00"} {"global_step": 3866, "acc_step": 0, "speed/wps": 12951.765760556675, "speed/FLOPS": 203425278860751.84, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.6948864459991455, "optim/lr": 0.0023190000000000003, "optim/total_tokens": 1013448704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438240, "loss/out": 4.735711574554443, "created_at": "2025-01-14T18:27:26.162518+00:00"} {"global_step": 3867, "acc_step": 0, "speed/wps": 12959.422870858853, "speed/FLOPS": 203545544300019.72, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8657265901565552, "optim/lr": 0.0023196000000000002, "optim/total_tokens": 1013710848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439054, "loss/out": 4.7000932693481445, "created_at": "2025-01-14T18:27:31.228666+00:00"} {"global_step": 3868, "acc_step": 0, "speed/wps": 12958.92215350243, "speed/FLOPS": 203537679845879.38, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.395183563232422, "optim/lr": 0.0023202, "optim/total_tokens": 1013972992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318931, "loss/out": 4.766694068908691, "created_at": "2025-01-14T18:27:36.293588+00:00"} {"global_step": 3869, "acc_step": 0, "speed/wps": 12948.318319109841, "speed/FLOPS": 203371132055548.34, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6561641693115234, "optim/lr": 0.0023208, "optim/total_tokens": 1014235136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462963, "loss/out": 4.709787845611572, "created_at": "2025-01-14T18:27:41.362217+00:00"} {"global_step": 3870, "acc_step": 0, "speed/wps": 12949.533926292112, "speed/FLOPS": 203390224836765.22, "speed/curr_iter_time": 1.2771, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.720062017440796, "optim/lr": 0.0023214000000000004, "optim/total_tokens": 1014497280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466420, "loss/out": 4.755847930908203, "created_at": "2025-01-14T18:27:46.430235+00:00"} {"global_step": 3871, "acc_step": 0, "speed/wps": 12949.234582462635, "speed/FLOPS": 203385523230581.78, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.155827045440674, "optim/lr": 0.0023220000000000003, "optim/total_tokens": 1014759424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 4.722754001617432, "created_at": "2025-01-14T18:27:51.497378+00:00"} {"global_step": 3872, "acc_step": 0, "speed/wps": 12958.911960850557, "speed/FLOPS": 203537519756275.53, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.9566595554351807, "optim/lr": 0.0023226, "optim/total_tokens": 1015021568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421165, "loss/out": 4.7591094970703125, "created_at": "2025-01-14T18:27:56.558128+00:00"} {"global_step": 3873, "acc_step": 0, "speed/wps": 12957.968165951492, "speed/FLOPS": 203522696160475.62, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8297449350357056, "optim/lr": 0.0023232, "optim/total_tokens": 1015283712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331250, "loss/out": 4.726282119750977, "created_at": "2025-01-14T18:28:01.619200+00:00"} {"global_step": 3874, "acc_step": 0, "speed/wps": 12951.193259998208, "speed/FLOPS": 203416286952782.94, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.716111660003662, "optim/lr": 0.0023238, "optim/total_tokens": 1015545856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468777, "loss/out": 4.630012035369873, "created_at": "2025-01-14T18:28:06.687781+00:00"} {"global_step": 3875, "acc_step": 0, "speed/wps": 12955.349469515384, "speed/FLOPS": 203481565934480.53, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.906267523765564, "optim/lr": 0.0023244000000000003, "optim/total_tokens": 1015808000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 4.6976318359375, "created_at": "2025-01-14T18:28:11.752797+00:00"} {"global_step": 3876, "acc_step": 0, "speed/wps": 12944.668835094304, "speed/FLOPS": 203313811894167.78, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.9411733150482178, "optim/lr": 0.0023250000000000002, "optim/total_tokens": 1016070144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438531, "loss/out": 4.714179992675781, "created_at": "2025-01-14T18:28:16.824549+00:00"} {"global_step": 3877, "acc_step": 0, "speed/wps": 12949.915652763295, "speed/FLOPS": 203396220375542.3, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.3776631355285645, "optim/lr": 0.0023256, "optim/total_tokens": 1016332288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460034, "loss/out": 4.631787300109863, "created_at": "2025-01-14T18:28:21.890102+00:00"} {"global_step": 3878, "acc_step": 0, "speed/wps": 12950.151035081193, "speed/FLOPS": 203399917378294.1, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0188324451446533, "optim/lr": 0.0023262, "optim/total_tokens": 1016594432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322156, "loss/out": 4.702865123748779, "created_at": "2025-01-14T18:28:26.954269+00:00"} {"global_step": 3879, "acc_step": 0, "speed/wps": 12956.920281694815, "speed/FLOPS": 203506237698280.62, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.4323782920837402, "optim/lr": 0.0023268, "optim/total_tokens": 1016856576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430506, "loss/out": 4.692783355712891, "created_at": "2025-01-14T18:28:32.015958+00:00"} {"global_step": 3880, "acc_step": 0, "speed/wps": 12950.653331956548, "speed/FLOPS": 203407806640949.78, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9814825057983398, "optim/lr": 0.0023274000000000003, "optim/total_tokens": 1017118720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 4.769598484039307, "created_at": "2025-01-14T18:28:37.081554+00:00"} {"global_step": 3881, "acc_step": 0, "speed/wps": 12951.659761336186, "speed/FLOPS": 203423613997335.88, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.81864595413208, "optim/lr": 0.002328, "optim/total_tokens": 1017380864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440390, "loss/out": 4.668231964111328, "created_at": "2025-01-14T18:28:42.146153+00:00"} {"global_step": 3882, "acc_step": 0, "speed/wps": 12950.95779212114, "speed/FLOPS": 203412588606206.8, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0013, "optim/grad_norm": 4.213150978088379, "optim/lr": 0.0023286, "optim/total_tokens": 1017643008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 4.708372116088867, "created_at": "2025-01-14T18:28:47.217187+00:00"} {"global_step": 3883, "acc_step": 0, "speed/wps": 12961.453794961943, "speed/FLOPS": 203577442753840.94, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.997825264930725, "optim/lr": 0.0023292, "optim/total_tokens": 1017905152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337297, "loss/out": 4.773622512817383, "created_at": "2025-01-14T18:28:52.277327+00:00"} {"global_step": 3884, "acc_step": 0, "speed/wps": 12957.52366912806, "speed/FLOPS": 203515714727060.97, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5779480934143066, "optim/lr": 0.0023298, "optim/total_tokens": 1018167296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361473, "loss/out": 4.873408317565918, "created_at": "2025-01-14T18:28:57.340736+00:00"} {"global_step": 3885, "acc_step": 0, "speed/wps": 12951.56393062826, "speed/FLOPS": 203422108844382.7, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.317445755004883, "optim/lr": 0.0023304000000000003, "optim/total_tokens": 1018429440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451797, "loss/out": 4.6945695877075195, "created_at": "2025-01-14T18:29:02.410041+00:00"} {"global_step": 3886, "acc_step": 0, "speed/wps": 12958.693978922556, "speed/FLOPS": 203534096050557.72, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.881315231323242, "optim/lr": 0.002331, "optim/total_tokens": 1018691584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452081, "loss/out": 4.840229034423828, "created_at": "2025-01-14T18:29:07.471217+00:00"} {"global_step": 3887, "acc_step": 0, "speed/wps": 12955.625484327717, "speed/FLOPS": 203485901126392.3, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.471934795379639, "optim/lr": 0.0023316, "optim/total_tokens": 1018953728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452330, "loss/out": 4.782081127166748, "created_at": "2025-01-14T18:29:12.538340+00:00"} {"global_step": 3888, "acc_step": 0, "speed/wps": 12956.299889227348, "speed/FLOPS": 203496493582070.5, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5699093341827393, "optim/lr": 0.0023322, "optim/total_tokens": 1019215872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454695, "loss/out": 4.785656929016113, "created_at": "2025-01-14T18:29:17.605662+00:00"} {"global_step": 3889, "acc_step": 0, "speed/wps": 12957.140876077952, "speed/FLOPS": 203509702436196.25, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.783521056175232, "optim/lr": 0.0023328, "optim/total_tokens": 1019478016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426015, "loss/out": 4.837742328643799, "created_at": "2025-01-14T18:29:22.667872+00:00"} {"global_step": 3890, "acc_step": 0, "speed/wps": 12959.295192682293, "speed/FLOPS": 203543538938808.66, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3431406021118164, "optim/lr": 0.0023334000000000002, "optim/total_tokens": 1019740160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321630, "loss/out": 4.58181095123291, "created_at": "2025-01-14T18:29:27.758859+00:00"} {"global_step": 3891, "acc_step": 0, "speed/wps": 12955.377767177142, "speed/FLOPS": 203482010388143.53, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.137810707092285, "optim/lr": 0.002334, "optim/total_tokens": 1020002304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464298, "loss/out": 4.692835807800293, "created_at": "2025-01-14T18:29:32.821703+00:00"} {"global_step": 3892, "acc_step": 0, "speed/wps": 12958.285163856846, "speed/FLOPS": 203527675048179.97, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.692126989364624, "optim/lr": 0.0023346, "optim/total_tokens": 1020264448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461385, "loss/out": 4.890214920043945, "created_at": "2025-01-14T18:29:37.883304+00:00"} {"global_step": 3893, "acc_step": 0, "speed/wps": 12952.674503428361, "speed/FLOPS": 203439551916295.25, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.4731295108795166, "optim/lr": 0.0023352, "optim/total_tokens": 1020526592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 4.844330787658691, "created_at": "2025-01-14T18:29:42.951627+00:00"} {"global_step": 3894, "acc_step": 0, "speed/wps": 12966.295783002683, "speed/FLOPS": 203653492829609.56, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.827231764793396, "optim/lr": 0.0023358, "optim/total_tokens": 1020788736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 339460, "loss/out": 4.694446563720703, "created_at": "2025-01-14T18:29:48.010604+00:00"} {"global_step": 3895, "acc_step": 0, "speed/wps": 12957.540509296827, "speed/FLOPS": 203515979225052.28, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.200214147567749, "optim/lr": 0.0023364, "optim/total_tokens": 1021050880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371206, "loss/out": 4.8803935050964355, "created_at": "2025-01-14T18:29:53.075602+00:00"} {"global_step": 3896, "acc_step": 0, "speed/wps": 12959.648763121866, "speed/FLOPS": 203549092248418.84, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7562698125839233, "optim/lr": 0.002337, "optim/total_tokens": 1021313024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 4.706954479217529, "created_at": "2025-01-14T18:29:58.136346+00:00"} {"global_step": 3897, "acc_step": 0, "speed/wps": 12966.976873854153, "speed/FLOPS": 203664190297348.8, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9747964143753052, "optim/lr": 0.0023376, "optim/total_tokens": 1021575168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 4.724847793579102, "created_at": "2025-01-14T18:30:03.195127+00:00"} {"global_step": 3898, "acc_step": 0, "speed/wps": 12956.233574630523, "speed/FLOPS": 203495452020201.56, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.736588001251221, "optim/lr": 0.0023382, "optim/total_tokens": 1021837312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425464, "loss/out": 4.629613876342773, "created_at": "2025-01-14T18:30:08.258341+00:00"} {"global_step": 3899, "acc_step": 0, "speed/wps": 12967.915710733716, "speed/FLOPS": 203678936020639.4, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8299081325531006, "optim/lr": 0.0023388, "optim/total_tokens": 1022099456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 4.796936988830566, "created_at": "2025-01-14T18:30:13.316658+00:00"} {"global_step": 3900, "acc_step": 0, "speed/wps": 12952.7818425012, "speed/FLOPS": 203441237823933.5, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6561534404754639, "optim/lr": 0.0023394, "optim/total_tokens": 1022361600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440105, "loss/out": 4.762404918670654, "created_at": "2025-01-14T18:30:18.380052+00:00"} {"global_step": 3901, "acc_step": 0, "speed/wps": 12955.04705909352, "speed/FLOPS": 203476816163249.72, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.001, "optim/grad_norm": 2.2677037715911865, "optim/lr": 0.00234, "optim/total_tokens": 1022623744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302343, "loss/out": 4.702587127685547, "created_at": "2025-01-14T18:30:23.443383+00:00"} {"global_step": 3902, "acc_step": 0, "speed/wps": 12960.670173884395, "speed/FLOPS": 203565134908009.84, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.339512348175049, "optim/lr": 0.0023406, "optim/total_tokens": 1022885888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358303, "loss/out": 4.773963928222656, "created_at": "2025-01-14T18:30:28.515147+00:00"} {"global_step": 3903, "acc_step": 0, "speed/wps": 12955.127958240279, "speed/FLOPS": 203478086795516.62, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8035749197006226, "optim/lr": 0.0023412, "optim/total_tokens": 1023148032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 4.850278854370117, "created_at": "2025-01-14T18:30:33.577584+00:00"} {"global_step": 3904, "acc_step": 0, "speed/wps": 12963.778005288947, "speed/FLOPS": 203613947670826.22, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9931601285934448, "optim/lr": 0.0023418, "optim/total_tokens": 1023410176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459743, "loss/out": 4.643960475921631, "created_at": "2025-01-14T18:30:38.642334+00:00"} {"global_step": 3905, "acc_step": 0, "speed/wps": 12956.242145704464, "speed/FLOPS": 203495586640695.66, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.9182615280151367, "optim/lr": 0.0023424, "optim/total_tokens": 1023672320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 4.674731731414795, "created_at": "2025-01-14T18:30:43.705546+00:00"} {"global_step": 3906, "acc_step": 0, "speed/wps": 12962.473408129183, "speed/FLOPS": 203593457179727.75, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5336875915527344, "optim/lr": 0.002343, "optim/total_tokens": 1023934464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426544, "loss/out": 4.609614372253418, "created_at": "2025-01-14T18:30:48.765999+00:00"} {"global_step": 3907, "acc_step": 0, "speed/wps": 12961.439203107522, "speed/FLOPS": 203577213568715.94, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3459596633911133, "optim/lr": 0.0023436, "optim/total_tokens": 1024196608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328073, "loss/out": 4.801612854003906, "created_at": "2025-01-14T18:30:53.827354+00:00"} {"global_step": 3908, "acc_step": 0, "speed/wps": 12962.626787185905, "speed/FLOPS": 203595866208578.5, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.4470553398132324, "optim/lr": 0.0023442, "optim/total_tokens": 1024458752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464007, "loss/out": 4.8024373054504395, "created_at": "2025-01-14T18:30:58.887785+00:00"} {"global_step": 3909, "acc_step": 0, "speed/wps": 12961.14383255271, "speed/FLOPS": 203572574368270.97, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1653337478637695, "optim/lr": 0.0023447999999999997, "optim/total_tokens": 1024720896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503748, "loss/out": 4.72461462020874, "created_at": "2025-01-14T18:31:03.947784+00:00"} {"global_step": 3910, "acc_step": 0, "speed/wps": 12959.114702414694, "speed/FLOPS": 203540704091136.4, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.3729465007781982, "optim/lr": 0.0023454, "optim/total_tokens": 1024983040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449411, "loss/out": 4.735871315002441, "created_at": "2025-01-14T18:31:09.009620+00:00"} {"global_step": 3911, "acc_step": 0, "speed/wps": 12955.047445472257, "speed/FLOPS": 203476822231858.72, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.7955024242401123, "optim/lr": 0.002346, "optim/total_tokens": 1025245184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 4.784719467163086, "created_at": "2025-01-14T18:31:14.086467+00:00"} {"global_step": 3912, "acc_step": 0, "speed/wps": 12952.030968834053, "speed/FLOPS": 203429444321182.44, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.3787906169891357, "optim/lr": 0.0023466, "optim/total_tokens": 1025507328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307440, "loss/out": 4.795860767364502, "created_at": "2025-01-14T18:31:19.153438+00:00"} {"global_step": 3913, "acc_step": 0, "speed/wps": 12957.504366693189, "speed/FLOPS": 203515411555797.0, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0012, "optim/grad_norm": 4.242938995361328, "optim/lr": 0.0023472, "optim/total_tokens": 1025769472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429193, "loss/out": 4.736961841583252, "created_at": "2025-01-14T18:31:24.214840+00:00"} {"global_step": 3914, "acc_step": 0, "speed/wps": 12952.838935281641, "speed/FLOPS": 203442134544504.06, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.074693441390991, "optim/lr": 0.0023478, "optim/total_tokens": 1026031616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457095, "loss/out": 4.816845893859863, "created_at": "2025-01-14T18:31:29.282897+00:00"} {"global_step": 3915, "acc_step": 0, "speed/wps": 12956.178634625992, "speed/FLOPS": 203494589111934.9, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.3349568843841553, "optim/lr": 0.0023484, "optim/total_tokens": 1026293760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 4.810661792755127, "created_at": "2025-01-14T18:31:34.345249+00:00"} {"global_step": 3916, "acc_step": 0, "speed/wps": 12937.868599718726, "speed/FLOPS": 203207004853091.7, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.2294654846191406, "optim/lr": 0.002349, "optim/total_tokens": 1026555904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433179, "loss/out": 4.8006696701049805, "created_at": "2025-01-14T18:31:39.442446+00:00"} {"global_step": 3917, "acc_step": 0, "speed/wps": 12958.727147537309, "speed/FLOPS": 203534617009230.88, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.3588991165161133, "optim/lr": 0.0023496, "optim/total_tokens": 1026818048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445933, "loss/out": 4.7823944091796875, "created_at": "2025-01-14T18:31:44.508662+00:00"} {"global_step": 3918, "acc_step": 0, "speed/wps": 12959.750902443046, "speed/FLOPS": 203550696486812.12, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.8431992530822754, "optim/lr": 0.0023502, "optim/total_tokens": 1027080192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 387505, "loss/out": 4.785894870758057, "created_at": "2025-01-14T18:31:49.570603+00:00"} {"global_step": 3919, "acc_step": 0, "speed/wps": 12958.144489006774, "speed/FLOPS": 203525465556352.1, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.302485942840576, "optim/lr": 0.0023508, "optim/total_tokens": 1027342336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 4.916868209838867, "created_at": "2025-01-14T18:31:54.632877+00:00"} {"global_step": 3920, "acc_step": 0, "speed/wps": 12956.16077224736, "speed/FLOPS": 203494308558733.0, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.8478827476501465, "optim/lr": 0.0023514, "optim/total_tokens": 1027604480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473602, "loss/out": 4.743559837341309, "created_at": "2025-01-14T18:31:59.695264+00:00"} {"global_step": 3921, "acc_step": 0, "speed/wps": 12954.44513763031, "speed/FLOPS": 203467362159543.22, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7196271419525146, "optim/lr": 0.002352, "optim/total_tokens": 1027866624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444641, "loss/out": 4.9014129638671875, "created_at": "2025-01-14T18:32:04.759598+00:00"} {"global_step": 3922, "acc_step": 0, "speed/wps": 12955.549199172992, "speed/FLOPS": 203484702963210.6, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.722066879272461, "optim/lr": 0.0023526000000000003, "optim/total_tokens": 1028128768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437717, "loss/out": 4.801568508148193, "created_at": "2025-01-14T18:32:09.822078+00:00"} {"global_step": 3923, "acc_step": 0, "speed/wps": 12955.352317403262, "speed/FLOPS": 203481610664473.06, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0014, "optim/grad_norm": 8.481451988220215, "optim/lr": 0.0023532, "optim/total_tokens": 1028390912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296240, "loss/out": 4.751798152923584, "created_at": "2025-01-14T18:32:14.896465+00:00"} {"global_step": 3924, "acc_step": 0, "speed/wps": 12962.898388818207, "speed/FLOPS": 203600132085433.53, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.170597076416016, "optim/lr": 0.0023538, "optim/total_tokens": 1028653056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470641, "loss/out": 4.851271629333496, "created_at": "2025-01-14T18:32:19.956524+00:00"} {"global_step": 3925, "acc_step": 0, "speed/wps": 12963.606047374335, "speed/FLOPS": 203611246835482.28, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.6397743225097656, "optim/lr": 0.0023544000000000004, "optim/total_tokens": 1028915200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 4.897963047027588, "created_at": "2025-01-14T18:32:25.022342+00:00"} {"global_step": 3926, "acc_step": 0, "speed/wps": 12953.860252932396, "speed/FLOPS": 203458175741646.7, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.2589447498321533, "optim/lr": 0.0023550000000000003, "optim/total_tokens": 1029177344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 4.726131439208984, "created_at": "2025-01-14T18:32:30.087266+00:00"} {"global_step": 3927, "acc_step": 0, "speed/wps": 12956.41567601196, "speed/FLOPS": 203498312172630.53, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.929266929626465, "optim/lr": 0.0023556000000000002, "optim/total_tokens": 1029439488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426252, "loss/out": 4.819094181060791, "created_at": "2025-01-14T18:32:35.150281+00:00"} {"global_step": 3928, "acc_step": 0, "speed/wps": 12953.834804351009, "speed/FLOPS": 203457776036705.97, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.0163257122039795, "optim/lr": 0.0023562, "optim/total_tokens": 1029701632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317879, "loss/out": 4.757954120635986, "created_at": "2025-01-14T18:32:40.216218+00:00"} {"global_step": 3929, "acc_step": 0, "speed/wps": 12953.458128393831, "speed/FLOPS": 203451859823191.5, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.412465572357178, "optim/lr": 0.0023568, "optim/total_tokens": 1029963776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 397614, "loss/out": 4.726049900054932, "created_at": "2025-01-14T18:32:45.282814+00:00"} {"global_step": 3930, "acc_step": 0, "speed/wps": 12953.034970533583, "speed/FLOPS": 203445213547516.47, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.062543869018555, "optim/lr": 0.0023574000000000004, "optim/total_tokens": 1030225920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 4.857913017272949, "created_at": "2025-01-14T18:32:50.346311+00:00"} {"global_step": 3931, "acc_step": 0, "speed/wps": 12955.652914954395, "speed/FLOPS": 203486331962077.94, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.211745262145996, "optim/lr": 0.0023580000000000003, "optim/total_tokens": 1030488064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424678, "loss/out": 4.989199638366699, "created_at": "2025-01-14T18:32:55.414017+00:00"} {"global_step": 3932, "acc_step": 0, "speed/wps": 12960.090555786177, "speed/FLOPS": 203556031209298.6, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.059927940368652, "optim/lr": 0.0023586, "optim/total_tokens": 1030750208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 4.928894519805908, "created_at": "2025-01-14T18:33:00.477963+00:00"} {"global_step": 3933, "acc_step": 0, "speed/wps": 12961.387606650173, "speed/FLOPS": 203576403175452.3, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.9966137409210205, "optim/lr": 0.0023592, "optim/total_tokens": 1031012352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417752, "loss/out": 4.8351240158081055, "created_at": "2025-01-14T18:33:05.539697+00:00"} {"global_step": 3934, "acc_step": 0, "speed/wps": 12958.967978870567, "speed/FLOPS": 203538399596256.62, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.895473003387451, "optim/lr": 0.0023598, "optim/total_tokens": 1031274496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301637, "loss/out": 4.84690523147583, "created_at": "2025-01-14T18:33:10.600484+00:00"} {"global_step": 3935, "acc_step": 0, "speed/wps": 12957.501262561627, "speed/FLOPS": 203515362801145.3, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3620119094848633, "optim/lr": 0.0023604000000000003, "optim/total_tokens": 1031536640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462672, "loss/out": 4.835807800292969, "created_at": "2025-01-14T18:33:15.663838+00:00"} {"global_step": 3936, "acc_step": 0, "speed/wps": 12953.988157891095, "speed/FLOPS": 203460184664782.88, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.1398415565490723, "optim/lr": 0.0023610000000000003, "optim/total_tokens": 1031798784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 4.8010687828063965, "created_at": "2025-01-14T18:33:20.727798+00:00"} {"global_step": 3937, "acc_step": 0, "speed/wps": 12958.762855474602, "speed/FLOPS": 203535177851453.9, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.001, "optim/grad_norm": 5.078882217407227, "optim/lr": 0.0023616, "optim/total_tokens": 1032060928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430791, "loss/out": 4.839816093444824, "created_at": "2025-01-14T18:33:25.797449+00:00"} {"global_step": 3938, "acc_step": 0, "speed/wps": 12952.720949812217, "speed/FLOPS": 203440281420577.72, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.588252067565918, "optim/lr": 0.0023622, "optim/total_tokens": 1032323072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447263, "loss/out": 5.021234035491943, "created_at": "2025-01-14T18:33:30.861829+00:00"} {"global_step": 3939, "acc_step": 0, "speed/wps": 12955.61029258106, "speed/FLOPS": 203485662519135.44, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.030876636505127, "optim/lr": 0.0023628, "optim/total_tokens": 1032585216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309147, "loss/out": 4.956615447998047, "created_at": "2025-01-14T18:33:35.928096+00:00"} {"global_step": 3940, "acc_step": 0, "speed/wps": 12948.752417738302, "speed/FLOPS": 203377950171022.34, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.198449611663818, "optim/lr": 0.0023634, "optim/total_tokens": 1032847360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468777, "loss/out": 4.896302223205566, "created_at": "2025-01-14T18:33:40.994355+00:00"} {"global_step": 3941, "acc_step": 0, "speed/wps": 12953.46726091591, "speed/FLOPS": 203452003262000.16, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.122300863265991, "optim/lr": 0.002364, "optim/total_tokens": 1033109504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 4.899167060852051, "created_at": "2025-01-14T18:33:46.062999+00:00"} {"global_step": 3942, "acc_step": 0, "speed/wps": 12950.100143329655, "speed/FLOPS": 203399118053403.6, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0009, "optim/grad_norm": 4.243415832519531, "optim/lr": 0.0023646, "optim/total_tokens": 1033371648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463770, "loss/out": 4.926953315734863, "created_at": "2025-01-14T18:33:51.130902+00:00"} {"global_step": 3943, "acc_step": 0, "speed/wps": 12953.315894005136, "speed/FLOPS": 203449625836666.53, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.0846142768859863, "optim/lr": 0.0023652, "optim/total_tokens": 1033633792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462938, "loss/out": 4.868412017822266, "created_at": "2025-01-14T18:33:56.194260+00:00"} {"global_step": 3944, "acc_step": 0, "speed/wps": 12957.19975373155, "speed/FLOPS": 203510627190652.38, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.725179195404053, "optim/lr": 0.0023658, "optim/total_tokens": 1033895936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423342, "loss/out": 4.920733451843262, "created_at": "2025-01-14T18:34:01.256931+00:00"} {"global_step": 3945, "acc_step": 0, "speed/wps": 12957.271875888324, "speed/FLOPS": 203511759968230.75, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.7004284858703613, "optim/lr": 0.0023664, "optim/total_tokens": 1034158080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299119, "loss/out": 5.206490993499756, "created_at": "2025-01-14T18:34:06.318748+00:00"} {"global_step": 3946, "acc_step": 0, "speed/wps": 12956.372127800532, "speed/FLOPS": 203497628188128.53, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.032443046569824, "optim/lr": 0.002367, "optim/total_tokens": 1034420224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465606, "loss/out": 4.968939781188965, "created_at": "2025-01-14T18:34:11.386747+00:00"} {"global_step": 3947, "acc_step": 0, "speed/wps": 12960.06839574001, "speed/FLOPS": 203555683155322.47, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.9571855068206787, "optim/lr": 0.0023676, "optim/total_tokens": 1034682368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 5.246007919311523, "created_at": "2025-01-14T18:34:16.452611+00:00"} {"global_step": 3948, "acc_step": 0, "speed/wps": 12954.38861735315, "speed/FLOPS": 203466474430923.44, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.324844837188721, "optim/lr": 0.0023682, "optim/total_tokens": 1034944512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451559, "loss/out": 5.10391902923584, "created_at": "2025-01-14T18:34:21.515746+00:00"} {"global_step": 3949, "acc_step": 0, "speed/wps": 12961.529762902344, "speed/FLOPS": 203578635934735.84, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.8344781398773193, "optim/lr": 0.0023688, "optim/total_tokens": 1035206656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457889, "loss/out": 5.062748908996582, "created_at": "2025-01-14T18:34:26.600599+00:00"} {"global_step": 3950, "acc_step": 0, "speed/wps": 12959.20109616621, "speed/FLOPS": 203542061023721.44, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.496175765991211, "optim/lr": 0.0023694, "optim/total_tokens": 1035468800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 4.990022659301758, "created_at": "2025-01-14T18:34:31.662345+00:00"} {"global_step": 3951, "acc_step": 0, "speed/wps": 12960.632555294536, "speed/FLOPS": 203564544056362.75, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.591402769088745, "optim/lr": 0.00237, "optim/total_tokens": 1035730944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 4.971715450286865, "created_at": "2025-01-14T18:34:36.732468+00:00"} {"global_step": 3952, "acc_step": 0, "speed/wps": 12950.127266958072, "speed/FLOPS": 203399544067259.66, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.344528675079346, "optim/lr": 0.0023706, "optim/total_tokens": 1035993088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 4.977330684661865, "created_at": "2025-01-14T18:34:41.801574+00:00"} {"global_step": 3953, "acc_step": 0, "speed/wps": 12951.928674628867, "speed/FLOPS": 203427837650123.03, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.8738622665405273, "optim/lr": 0.0023712, "optim/total_tokens": 1036255232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326344, "loss/out": 4.831800937652588, "created_at": "2025-01-14T18:34:46.866354+00:00"} {"global_step": 3954, "acc_step": 0, "speed/wps": 12956.363733084474, "speed/FLOPS": 203497496337577.3, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.6966676712036133, "optim/lr": 0.0023718, "optim/total_tokens": 1036517376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338225, "loss/out": 4.995755195617676, "created_at": "2025-01-14T18:34:51.936045+00:00"} {"global_step": 3955, "acc_step": 0, "speed/wps": 12961.513580844672, "speed/FLOPS": 203578381773283.62, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.5630862712860107, "optim/lr": 0.0023723999999999998, "optim/total_tokens": 1036779520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 384815, "loss/out": 4.805215358734131, "created_at": "2025-01-14T18:34:56.997726+00:00"} {"global_step": 3956, "acc_step": 0, "speed/wps": 12952.601972300052, "speed/FLOPS": 203438412715256.94, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5301218032836914, "optim/lr": 0.002373, "optim/total_tokens": 1037041664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 4.90676212310791, "created_at": "2025-01-14T18:35:02.061051+00:00"} {"global_step": 3957, "acc_step": 0, "speed/wps": 12962.786182612597, "speed/FLOPS": 203598369732786.97, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.2778069972991943, "optim/lr": 0.0023736, "optim/total_tokens": 1037303808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462126, "loss/out": 4.933980464935303, "created_at": "2025-01-14T18:35:07.125415+00:00"} {"global_step": 3958, "acc_step": 0, "speed/wps": 12956.265189616599, "speed/FLOPS": 203495948577001.4, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.371586799621582, "optim/lr": 0.0023742, "optim/total_tokens": 1037565952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463770, "loss/out": 4.833036422729492, "created_at": "2025-01-14T18:35:12.187366+00:00"} {"global_step": 3959, "acc_step": 0, "speed/wps": 12957.748726878652, "speed/FLOPS": 203519249568295.78, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0009, "optim/grad_norm": 2.478297472000122, "optim/lr": 0.0023748, "optim/total_tokens": 1037828096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438240, "loss/out": 4.775008678436279, "created_at": "2025-01-14T18:35:17.250350+00:00"} {"global_step": 3960, "acc_step": 0, "speed/wps": 12957.734761350537, "speed/FLOPS": 203519030220486.1, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3423967361450195, "optim/lr": 0.0023753999999999997, "optim/total_tokens": 1038090240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309920, "loss/out": 4.828168869018555, "created_at": "2025-01-14T18:35:22.313617+00:00"} {"global_step": 3961, "acc_step": 0, "speed/wps": 12960.717966685395, "speed/FLOPS": 203565885559623.75, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.216817855834961, "optim/lr": 0.002376, "optim/total_tokens": 1038352384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459743, "loss/out": 4.905348777770996, "created_at": "2025-01-14T18:35:27.378089+00:00"} {"global_step": 3962, "acc_step": 0, "speed/wps": 12959.067592053201, "speed/FLOPS": 203539964158172.75, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.031362295150757, "optim/lr": 0.0023766, "optim/total_tokens": 1038614528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 4.902512073516846, "created_at": "2025-01-14T18:35:32.439928+00:00"} {"global_step": 3963, "acc_step": 0, "speed/wps": 12958.096951414831, "speed/FLOPS": 203524718913147.12, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.514433860778809, "optim/lr": 0.0023772, "optim/total_tokens": 1038876672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450994, "loss/out": 4.8004560470581055, "created_at": "2025-01-14T18:35:37.504406+00:00"} {"global_step": 3964, "acc_step": 0, "speed/wps": 12960.914395671658, "speed/FLOPS": 203568970746774.72, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0013, "optim/grad_norm": 11.427156448364258, "optim/lr": 0.0023778, "optim/total_tokens": 1039138816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454466, "loss/out": 4.863584041595459, "created_at": "2025-01-14T18:35:42.564401+00:00"} {"global_step": 3965, "acc_step": 0, "speed/wps": 12953.669755844863, "speed/FLOPS": 203455183723116.56, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5087544918060303, "optim/lr": 0.0023783999999999997, "optim/total_tokens": 1039400960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323123, "loss/out": 4.964327812194824, "created_at": "2025-01-14T18:35:47.627965+00:00"} {"global_step": 3966, "acc_step": 0, "speed/wps": 12960.172341429401, "speed/FLOPS": 203557315765210.7, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7750914096832275, "optim/lr": 0.002379, "optim/total_tokens": 1039663104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420635, "loss/out": 4.854614734649658, "created_at": "2025-01-14T18:35:52.689411+00:00"} {"global_step": 3967, "acc_step": 0, "speed/wps": 12954.645838680506, "speed/FLOPS": 203470514445328.7, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.567551612854004, "optim/lr": 0.0023796, "optim/total_tokens": 1039925248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 4.929553985595703, "created_at": "2025-01-14T18:35:57.765452+00:00"} {"global_step": 3968, "acc_step": 0, "speed/wps": 12960.139036521956, "speed/FLOPS": 203556792665873.53, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.848076343536377, "optim/lr": 0.0023802, "optim/total_tokens": 1040187392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 4.976789474487305, "created_at": "2025-01-14T18:36:02.826503+00:00"} {"global_step": 3969, "acc_step": 0, "speed/wps": 12956.536729888427, "speed/FLOPS": 203500213490105.9, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.2110326290130615, "optim/lr": 0.0023807999999999998, "optim/total_tokens": 1040449536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438240, "loss/out": 4.954960823059082, "created_at": "2025-01-14T18:36:07.896348+00:00"} {"global_step": 3970, "acc_step": 0, "speed/wps": 12952.760065073604, "speed/FLOPS": 203440895779507.8, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.0278704166412354, "optim/lr": 0.0023814, "optim/total_tokens": 1040711680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462150, "loss/out": 4.914000988006592, "created_at": "2025-01-14T18:36:12.964853+00:00"} {"global_step": 3971, "acc_step": 0, "speed/wps": 12957.271875888324, "speed/FLOPS": 203511759968230.75, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7495529651641846, "optim/lr": 0.002382, "optim/total_tokens": 1040973824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317346, "loss/out": 4.889099597930908, "created_at": "2025-01-14T18:36:18.026301+00:00"} {"global_step": 3972, "acc_step": 0, "speed/wps": 12960.00163679141, "speed/FLOPS": 203554634614298.28, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.893796682357788, "optim/lr": 0.0023826, "optim/total_tokens": 1041235968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 388560, "loss/out": 4.959505081176758, "created_at": "2025-01-14T18:36:23.088027+00:00"} {"global_step": 3973, "acc_step": 0, "speed/wps": 12955.81562627874, "speed/FLOPS": 203488887567015.56, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.088202953338623, "optim/lr": 0.0023832000000000002, "optim/total_tokens": 1041498112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462150, "loss/out": 4.9298577308654785, "created_at": "2025-01-14T18:36:28.149950+00:00"} {"global_step": 3974, "acc_step": 0, "speed/wps": 12953.08824822779, "speed/FLOPS": 203446050346915.22, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.136402606964111, "optim/lr": 0.0023838, "optim/total_tokens": 1041760256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462963, "loss/out": 4.956737518310547, "created_at": "2025-01-14T18:36:33.213571+00:00"} {"global_step": 3975, "acc_step": 0, "speed/wps": 12955.304051358846, "speed/FLOPS": 203480852579921.03, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.2100112438201904, "optim/lr": 0.0023844, "optim/total_tokens": 1042022400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448890, "loss/out": 4.882644176483154, "created_at": "2025-01-14T18:36:38.279259+00:00"} {"global_step": 3976, "acc_step": 0, "speed/wps": 12958.4324112433, "speed/FLOPS": 203529987770723.34, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.298133134841919, "optim/lr": 0.002385, "optim/total_tokens": 1042284544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447269, "loss/out": 4.942909240722656, "created_at": "2025-01-14T18:36:43.351625+00:00"} {"global_step": 3977, "acc_step": 0, "speed/wps": 12955.74757297589, "speed/FLOPS": 203487818696379.28, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9559892416000366, "optim/lr": 0.0023856000000000003, "optim/total_tokens": 1042546688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371881, "loss/out": 4.908847808837891, "created_at": "2025-01-14T18:36:48.418114+00:00"} {"global_step": 3978, "acc_step": 0, "speed/wps": 12958.163614383722, "speed/FLOPS": 203525765946678.1, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.377037286758423, "optim/lr": 0.0023862, "optim/total_tokens": 1042808832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 4.898265361785889, "created_at": "2025-01-14T18:36:53.485003+00:00"} {"global_step": 3979, "acc_step": 0, "speed/wps": 12963.044764373883, "speed/FLOPS": 203602431114676.53, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.400739908218384, "optim/lr": 0.0023868, "optim/total_tokens": 1043070976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432953, "loss/out": 4.960700988769531, "created_at": "2025-01-14T18:36:58.544232+00:00"} {"global_step": 3980, "acc_step": 0, "speed/wps": 12962.126333181663, "speed/FLOPS": 203588005890746.0, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.986238718032837, "optim/lr": 0.0023874, "optim/total_tokens": 1043333120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451797, "loss/out": 4.74284553527832, "created_at": "2025-01-14T18:37:03.603932+00:00"} {"global_step": 3981, "acc_step": 0, "speed/wps": 12955.619940969964, "speed/FLOPS": 203485814060341.44, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.888695240020752, "optim/lr": 0.0023880000000000004, "optim/total_tokens": 1043595264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 4.89577579498291, "created_at": "2025-01-14T18:37:08.667178+00:00"} {"global_step": 3982, "acc_step": 0, "speed/wps": 12952.854964552907, "speed/FLOPS": 203442386306235.8, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.198138475418091, "optim/lr": 0.0023886000000000003, "optim/total_tokens": 1043857408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433718, "loss/out": 4.896250247955322, "created_at": "2025-01-14T18:37:13.736136+00:00"} {"global_step": 3983, "acc_step": 0, "speed/wps": 12956.728539931028, "speed/FLOPS": 203503226130399.56, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.501340866088867, "optim/lr": 0.0023892, "optim/total_tokens": 1044119552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327499, "loss/out": 5.075247287750244, "created_at": "2025-01-14T18:37:18.798719+00:00"} {"global_step": 3984, "acc_step": 0, "speed/wps": 12958.911834367269, "speed/FLOPS": 203537517769681.7, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.758805274963379, "optim/lr": 0.0023898, "optim/total_tokens": 1044381696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453375, "loss/out": 4.912262439727783, "created_at": "2025-01-14T18:37:23.864627+00:00"} {"global_step": 3985, "acc_step": 0, "speed/wps": 12960.717837779708, "speed/FLOPS": 203565883534982.78, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.850043535232544, "optim/lr": 0.0023904, "optim/total_tokens": 1044643840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 5.0542755126953125, "created_at": "2025-01-14T18:37:28.924849+00:00"} {"global_step": 3986, "acc_step": 0, "speed/wps": 12967.065329451067, "speed/FLOPS": 203665579614049.06, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.1521759033203125, "optim/lr": 0.0023910000000000003, "optim/total_tokens": 1044905984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428688, "loss/out": 4.936147212982178, "created_at": "2025-01-14T18:37:33.983102+00:00"} {"global_step": 3987, "acc_step": 0, "speed/wps": 12964.755564354207, "speed/FLOPS": 203629301579253.06, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9569686651229858, "optim/lr": 0.0023916000000000002, "optim/total_tokens": 1045168128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 4.901426315307617, "created_at": "2025-01-14T18:37:39.041712+00:00"} {"global_step": 3988, "acc_step": 0, "speed/wps": 12968.634897137224, "speed/FLOPS": 203690231831372.2, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.041800498962402, "optim/lr": 0.0023922, "optim/total_tokens": 1045430272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435053, "loss/out": 5.018462181091309, "created_at": "2025-01-14T18:37:44.108359+00:00"} {"global_step": 3989, "acc_step": 0, "speed/wps": 12961.03730497431, "speed/FLOPS": 203570901206267.94, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.9741789102554321, "optim/lr": 0.0023928, "optim/total_tokens": 1045692416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352814, "loss/out": 4.836651802062988, "created_at": "2025-01-14T18:37:49.190288+00:00"} {"global_step": 3990, "acc_step": 0, "speed/wps": 12964.070465065552, "speed/FLOPS": 203618541153499.62, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.3243887424468994, "optim/lr": 0.0023934, "optim/total_tokens": 1045954560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 4.934675693511963, "created_at": "2025-01-14T18:37:54.249281+00:00"} {"global_step": 3991, "acc_step": 0, "speed/wps": 12964.281691833552, "speed/FLOPS": 203621858760147.06, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.5446808338165283, "optim/lr": 0.0023940000000000003, "optim/total_tokens": 1046216704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451523, "loss/out": 5.008495330810547, "created_at": "2025-01-14T18:37:59.308192+00:00"} {"global_step": 3992, "acc_step": 0, "speed/wps": 12962.489126963565, "speed/FLOPS": 203593704065621.66, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.8466458320617676, "optim/lr": 0.0023946, "optim/total_tokens": 1046478848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 4.995146751403809, "created_at": "2025-01-14T18:38:04.372679+00:00"} {"global_step": 3993, "acc_step": 0, "speed/wps": 12967.753773111172, "speed/FLOPS": 203676392567751.75, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.266813278198242, "optim/lr": 0.0023952, "optim/total_tokens": 1046740992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430567, "loss/out": 4.981594085693359, "created_at": "2025-01-14T18:38:09.429953+00:00"} {"global_step": 3994, "acc_step": 0, "speed/wps": 12965.516545293094, "speed/FLOPS": 203641253830595.66, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.202829360961914, "optim/lr": 0.0023958, "optim/total_tokens": 1047003136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438769, "loss/out": 4.875455856323242, "created_at": "2025-01-14T18:38:14.493240+00:00"} {"global_step": 3995, "acc_step": 0, "speed/wps": 12960.242030944573, "speed/FLOPS": 203558410334809.9, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.037496566772461, "optim/lr": 0.0023964, "optim/total_tokens": 1047265280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321253, "loss/out": 4.93448543548584, "created_at": "2025-01-14T18:38:19.554681+00:00"} {"global_step": 3996, "acc_step": 0, "speed/wps": 12961.172081526178, "speed/FLOPS": 203573018057217.4, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.559164047241211, "optim/lr": 0.0023970000000000003, "optim/total_tokens": 1047527424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461359, "loss/out": 4.990113258361816, "created_at": "2025-01-14T18:38:24.623676+00:00"} {"global_step": 3997, "acc_step": 0, "speed/wps": 12970.57660691884, "speed/FLOPS": 203720729051681.3, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.155996799468994, "optim/lr": 0.0023976, "optim/total_tokens": 1047789568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 4.954739093780518, "created_at": "2025-01-14T18:38:29.684218+00:00"} {"global_step": 3998, "acc_step": 0, "speed/wps": 12950.348066243045, "speed/FLOPS": 203403012023440.16, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.9200360774993896, "optim/lr": 0.0023982, "optim/total_tokens": 1048051712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453945, "loss/out": 4.990059852600098, "created_at": "2025-01-14T18:38:34.755628+00:00"} {"global_step": 3999, "acc_step": 0, "speed/wps": 12962.955215009302, "speed/FLOPS": 203601024618852.06, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.7262656688690186, "optim/lr": 0.0023988, "optim/total_tokens": 1048313856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 4.912871837615967, "created_at": "2025-01-14T18:38:39.815220+00:00"} {"global_step": 4000, "acc_step": 0, "speed/wps": 12963.448303487927, "speed/FLOPS": 203608769250983.44, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.586995840072632, "optim/lr": 0.0023994, "optim/total_tokens": 1048576000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427880, "loss/out": 4.971244812011719, "created_at": "2025-01-14T18:38:44.877335+00:00"} {"global_step": 4001, "acc_step": 0, "speed/wps": 12306.757782837594, "speed/FLOPS": 193294542236831.16, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.2665, "optim/grad_norm": 3.920703887939453, "optim/lr": 0.0024000000000000002, "optim/total_tokens": 1048838144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451808, "loss/out": 4.8218913078308105, "created_at": "2025-01-14T18:38:50.207726+00:00"} {"global_step": 4002, "acc_step": 0, "speed/wps": 12975.803266373883, "speed/FLOPS": 203802820920605.5, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.3570046424865723, "optim/lr": 0.0024006, "optim/total_tokens": 1049100288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498247, "loss/out": 4.806610107421875, "created_at": "2025-01-14T18:38:55.262075+00:00"} {"global_step": 4003, "acc_step": 0, "speed/wps": 12974.795064665987, "speed/FLOPS": 203786985727368.66, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.650331020355225, "optim/lr": 0.0024012, "optim/total_tokens": 1049362432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451797, "loss/out": 4.960587501525879, "created_at": "2025-01-14T18:39:00.317990+00:00"} {"global_step": 4004, "acc_step": 0, "speed/wps": 12974.734070089213, "speed/FLOPS": 203786027723725.1, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.275564193725586, "optim/lr": 0.0024018, "optim/total_tokens": 1049624576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330623, "loss/out": 4.883334636688232, "created_at": "2025-01-14T18:39:05.373864+00:00"} {"global_step": 4005, "acc_step": 0, "speed/wps": 12967.971861588443, "speed/FLOPS": 203679817946972.28, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0013, "optim/grad_norm": 2.203629493713379, "optim/lr": 0.0024024, "optim/total_tokens": 1049886720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365596, "loss/out": 5.031694412231445, "created_at": "2025-01-14T18:39:10.433463+00:00"} {"global_step": 4006, "acc_step": 0, "speed/wps": 12975.242883759016, "speed/FLOPS": 203794019341592.6, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.8392564058303833, "optim/lr": 0.002403, "optim/total_tokens": 1050148864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461913, "loss/out": 4.978989601135254, "created_at": "2025-01-14T18:39:15.488110+00:00"} {"global_step": 4007, "acc_step": 0, "speed/wps": 12972.684240875125, "speed/FLOPS": 203753832339157.34, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.8525872230529785, "optim/lr": 0.0024036, "optim/total_tokens": 1050411008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443835, "loss/out": 4.988302230834961, "created_at": "2025-01-14T18:39:20.543661+00:00"} {"global_step": 4008, "acc_step": 0, "speed/wps": 12965.905048697236, "speed/FLOPS": 203647355810417.3, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.4757590293884277, "optim/lr": 0.0024042, "optim/total_tokens": 1050673152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 4.911377906799316, "created_at": "2025-01-14T18:39:25.607891+00:00"} {"global_step": 4009, "acc_step": 0, "speed/wps": 12971.401904865248, "speed/FLOPS": 203733691490007.88, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8808711767196655, "optim/lr": 0.0024048, "optim/total_tokens": 1050935296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 4.792288303375244, "created_at": "2025-01-14T18:39:30.692066+00:00"} {"global_step": 4010, "acc_step": 0, "speed/wps": 12969.558900248705, "speed/FLOPS": 203704744569951.34, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0009, "optim/grad_norm": 3.71409010887146, "optim/lr": 0.0024054, "optim/total_tokens": 1051197440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 4.852436065673828, "created_at": "2025-01-14T18:39:35.754258+00:00"} {"global_step": 4011, "acc_step": 0, "speed/wps": 12971.47717170505, "speed/FLOPS": 203734873659154.44, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.672744631767273, "optim/lr": 0.002406, "optim/total_tokens": 1051459584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328746, "loss/out": 4.895118236541748, "created_at": "2025-01-14T18:39:40.814667+00:00"} {"global_step": 4012, "acc_step": 0, "speed/wps": 12970.356582816159, "speed/FLOPS": 203717273270803.9, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.279143810272217, "optim/lr": 0.0024066, "optim/total_tokens": 1051721728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371206, "loss/out": 5.063565254211426, "created_at": "2025-01-14T18:39:45.870973+00:00"} {"global_step": 4013, "acc_step": 0, "speed/wps": 12966.938088442337, "speed/FLOPS": 203663581119159.5, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.685859203338623, "optim/lr": 0.0024072, "optim/total_tokens": 1051983872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460578, "loss/out": 4.950233459472656, "created_at": "2025-01-14T18:39:50.930742+00:00"} {"global_step": 4014, "acc_step": 0, "speed/wps": 12973.974707394273, "speed/FLOPS": 203774100889127.38, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.189297676086426, "optim/lr": 0.0024078, "optim/total_tokens": 1052246016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 5.12717342376709, "created_at": "2025-01-14T18:39:55.986233+00:00"} {"global_step": 4015, "acc_step": 0, "speed/wps": 12967.07309049028, "speed/FLOPS": 203665701511834.7, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.543213129043579, "optim/lr": 0.0024083999999999998, "optim/total_tokens": 1052508160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432365, "loss/out": 5.105166435241699, "created_at": "2025-01-14T18:40:01.046044+00:00"} {"global_step": 4016, "acc_step": 0, "speed/wps": 12969.180821675542, "speed/FLOPS": 203698806326425.4, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.256899833679199, "optim/lr": 0.002409, "optim/total_tokens": 1052770304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 5.198935031890869, "created_at": "2025-01-14T18:40:06.104054+00:00"} {"global_step": 4017, "acc_step": 0, "speed/wps": 12969.876515716525, "speed/FLOPS": 203709733157322.84, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0009, "optim/grad_norm": 3.015580177307129, "optim/lr": 0.0024096, "optim/total_tokens": 1053032448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444878, "loss/out": 4.998578071594238, "created_at": "2025-01-14T18:40:11.161752+00:00"} {"global_step": 4018, "acc_step": 0, "speed/wps": 12969.67868430753, "speed/FLOPS": 203706625943195.72, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.453350067138672, "optim/lr": 0.0024102, "optim/total_tokens": 1053294592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309025, "loss/out": 4.930527687072754, "created_at": "2025-01-14T18:40:16.219795+00:00"} {"global_step": 4019, "acc_step": 0, "speed/wps": 12976.462037136549, "speed/FLOPS": 203813167820679.06, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.7370705604553223, "optim/lr": 0.0024108, "optim/total_tokens": 1053556736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372985, "loss/out": 5.10041618347168, "created_at": "2025-01-14T18:40:21.279071+00:00"} {"global_step": 4020, "acc_step": 0, "speed/wps": 12975.145232383633, "speed/FLOPS": 203792485592555.16, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.006393909454346, "optim/lr": 0.0024113999999999997, "optim/total_tokens": 1053818880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452596, "loss/out": 4.958183765411377, "created_at": "2025-01-14T18:40:26.334700+00:00"} {"global_step": 4021, "acc_step": 0, "speed/wps": 12957.721182390764, "speed/FLOPS": 203518816944263.4, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.780945777893066, "optim/lr": 0.002412, "optim/total_tokens": 1054081024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461385, "loss/out": 4.9659528732299805, "created_at": "2025-01-14T18:40:31.401208+00:00"} {"global_step": 4022, "acc_step": 0, "speed/wps": 12976.180034642497, "speed/FLOPS": 203808738583997.94, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.999375343322754, "optim/lr": 0.0024126, "optim/total_tokens": 1054343168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 5.0232954025268555, "created_at": "2025-01-14T18:40:36.455357+00:00"} {"global_step": 4023, "acc_step": 0, "speed/wps": 12963.137861205003, "speed/FLOPS": 203603893328337.72, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.343400955200195, "optim/lr": 0.0024132, "optim/total_tokens": 1054605312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 5.0673298835754395, "created_at": "2025-01-14T18:40:41.514728+00:00"} {"global_step": 4024, "acc_step": 0, "speed/wps": 12967.62230557284, "speed/FLOPS": 203674327689406.94, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.001, "optim/grad_norm": 5.241685390472412, "optim/lr": 0.0024138, "optim/total_tokens": 1054867456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434230, "loss/out": 5.1296000480651855, "created_at": "2025-01-14T18:40:46.576416+00:00"} {"global_step": 4025, "acc_step": 0, "speed/wps": 12966.748365364554, "speed/FLOPS": 203660601257516.1, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.454390287399292, "optim/lr": 0.0024144, "optim/total_tokens": 1055129600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 5.038188457489014, "created_at": "2025-01-14T18:40:51.634588+00:00"} {"global_step": 4026, "acc_step": 0, "speed/wps": 12963.375288655165, "speed/FLOPS": 203607622452702.0, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.293043851852417, "optim/lr": 0.002415, "optim/total_tokens": 1055391744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 344503, "loss/out": 5.021997451782227, "created_at": "2025-01-14T18:40:56.694965+00:00"} {"global_step": 4027, "acc_step": 0, "speed/wps": 12971.21462626096, "speed/FLOPS": 203730750022179.78, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 69.38313293457031, "optim/lr": 0.0024156, "optim/total_tokens": 1055653888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335427, "loss/out": 4.883257865905762, "created_at": "2025-01-14T18:41:01.756354+00:00"} {"global_step": 4028, "acc_step": 0, "speed/wps": 12966.80928249284, "speed/FLOPS": 203661558044724.75, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.448424339294434, "optim/lr": 0.0024162, "optim/total_tokens": 1055916032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457652, "loss/out": 4.945126533508301, "created_at": "2025-01-14T18:41:06.814642+00:00"} {"global_step": 4029, "acc_step": 0, "speed/wps": 12971.506108950512, "speed/FLOPS": 203735328158358.38, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.9512758255004883, "optim/lr": 0.0024168, "optim/total_tokens": 1056178176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 5.046773910522461, "created_at": "2025-01-14T18:41:11.872502+00:00"} {"global_step": 4030, "acc_step": 0, "speed/wps": 12960.528115260058, "speed/FLOPS": 203562903682105.62, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7798914909362793, "optim/lr": 0.0024174, "optim/total_tokens": 1056440320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426544, "loss/out": 5.000200271606445, "created_at": "2025-01-14T18:41:16.933681+00:00"} {"global_step": 4031, "acc_step": 0, "speed/wps": 12963.473223862622, "speed/FLOPS": 203609160659712.22, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.361297130584717, "optim/lr": 0.002418, "optim/total_tokens": 1056702464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 4.9781389236450195, "created_at": "2025-01-14T18:41:21.993445+00:00"} {"global_step": 4032, "acc_step": 0, "speed/wps": 12952.510559726179, "speed/FLOPS": 203436976955170.22, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.7024521827697754, "optim/lr": 0.0024186000000000004, "optim/total_tokens": 1056964608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329272, "loss/out": 5.006076812744141, "created_at": "2025-01-14T18:41:27.058020+00:00"} {"global_step": 4033, "acc_step": 0, "speed/wps": 12957.495717598571, "speed/FLOPS": 203515275709880.9, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.358233690261841, "optim/lr": 0.0024192000000000003, "optim/total_tokens": 1057226752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367900, "loss/out": 5.053034782409668, "created_at": "2025-01-14T18:41:32.121213+00:00"} {"global_step": 4034, "acc_step": 0, "speed/wps": 12943.848778128358, "speed/FLOPS": 203300931772641.25, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.401043176651001, "optim/lr": 0.0024198, "optim/total_tokens": 1057488896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465870, "loss/out": 4.857189655303955, "created_at": "2025-01-14T18:41:37.188622+00:00"} {"global_step": 4035, "acc_step": 0, "speed/wps": 12960.277183903625, "speed/FLOPS": 203558962460337.03, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.627650737762451, "optim/lr": 0.0024204, "optim/total_tokens": 1057751040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460287, "loss/out": 4.985018730163574, "created_at": "2025-01-14T18:41:42.250653+00:00"} {"global_step": 4036, "acc_step": 0, "speed/wps": 12960.063440531094, "speed/FLOPS": 203555605326957.25, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.7118964195251465, "optim/lr": 0.0024210000000000004, "optim/total_tokens": 1058013184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481758, "loss/out": 4.904501438140869, "created_at": "2025-01-14T18:41:47.311231+00:00"} {"global_step": 4037, "acc_step": 0, "speed/wps": 12947.794161488937, "speed/FLOPS": 203362899439844.0, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.254764556884766, "optim/lr": 0.0024216000000000003, "optim/total_tokens": 1058275328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430791, "loss/out": 5.035162925720215, "created_at": "2025-01-14T18:41:52.378204+00:00"} {"global_step": 4038, "acc_step": 0, "speed/wps": 12954.348342783167, "speed/FLOPS": 203465841863458.78, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.241182327270508, "optim/lr": 0.0024222000000000002, "optim/total_tokens": 1058537472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452081, "loss/out": 5.033342361450195, "created_at": "2025-01-14T18:41:57.441401+00:00"} {"global_step": 4039, "acc_step": 0, "speed/wps": 12954.389235019336, "speed/FLOPS": 203466484132219.66, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.519723892211914, "optim/lr": 0.0024228, "optim/total_tokens": 1058799616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310371, "loss/out": 5.105373382568359, "created_at": "2025-01-14T18:42:02.505150+00:00"} {"global_step": 4040, "acc_step": 0, "speed/wps": 12956.509629502003, "speed/FLOPS": 203499787841297.28, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.941336154937744, "optim/lr": 0.0024234, "optim/total_tokens": 1059061760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 4.969486236572266, "created_at": "2025-01-14T18:42:07.568254+00:00"} {"global_step": 4041, "acc_step": 0, "speed/wps": 12955.966357467885, "speed/FLOPS": 203491255007467.47, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.506779670715332, "optim/lr": 0.0024240000000000004, "optim/total_tokens": 1059323904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 4.944616317749023, "created_at": "2025-01-14T18:42:12.634511+00:00"} {"global_step": 4042, "acc_step": 0, "speed/wps": 12953.839497260296, "speed/FLOPS": 203457849745295.1, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.25425910949707, "optim/lr": 0.0024246000000000003, "optim/total_tokens": 1059586048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456567, "loss/out": 4.975423336029053, "created_at": "2025-01-14T18:42:17.701543+00:00"} {"global_step": 4043, "acc_step": 0, "speed/wps": 12950.777972575817, "speed/FLOPS": 203409764293149.56, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0012, "optim/grad_norm": 9.248586654663086, "optim/lr": 0.0024252, "optim/total_tokens": 1059848192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437181, "loss/out": 5.055402755737305, "created_at": "2025-01-14T18:42:22.768488+00:00"} {"global_step": 4044, "acc_step": 0, "speed/wps": 12955.895716843903, "speed/FLOPS": 203490145499397.4, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 18.056293487548828, "optim/lr": 0.0024258, "optim/total_tokens": 1060110336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367900, "loss/out": 4.8923845291137695, "created_at": "2025-01-14T18:42:27.833279+00:00"} {"global_step": 4045, "acc_step": 0, "speed/wps": 12957.160998156127, "speed/FLOPS": 203510018481084.62, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.7198052406311035, "optim/lr": 0.0024264, "optim/total_tokens": 1060372480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457665, "loss/out": 5.004156112670898, "created_at": "2025-01-14T18:42:32.895546+00:00"} {"global_step": 4046, "acc_step": 0, "speed/wps": 12959.700040229969, "speed/FLOPS": 203549897625863.75, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.053032398223877, "optim/lr": 0.0024270000000000003, "optim/total_tokens": 1060634624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 4.984524726867676, "created_at": "2025-01-14T18:42:37.957993+00:00"} {"global_step": 4047, "acc_step": 0, "speed/wps": 12949.522135098501, "speed/FLOPS": 203390039639866.78, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.855052471160889, "optim/lr": 0.0024276000000000002, "optim/total_tokens": 1060896768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438014, "loss/out": 5.086448669433594, "created_at": "2025-01-14T18:42:43.023088+00:00"} {"global_step": 4048, "acc_step": 0, "speed/wps": 12952.003992186248, "speed/FLOPS": 203429020615858.88, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.451503276824951, "optim/lr": 0.0024282, "optim/total_tokens": 1061158912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463248, "loss/out": 5.012295722961426, "created_at": "2025-01-14T18:42:48.087800+00:00"} {"global_step": 4049, "acc_step": 0, "speed/wps": 12953.234681289317, "speed/FLOPS": 203448350279366.62, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.483007907867432, "optim/lr": 0.0024288, "optim/total_tokens": 1061421056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466683, "loss/out": 4.9897356033325195, "created_at": "2025-01-14T18:42:53.158973+00:00"} {"global_step": 4050, "acc_step": 0, "speed/wps": 12955.221995615486, "speed/FLOPS": 203479563781715.16, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.775358200073242, "optim/lr": 0.0024294, "optim/total_tokens": 1061683200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335427, "loss/out": 4.823761940002441, "created_at": "2025-01-14T18:42:58.221345+00:00"} {"global_step": 4051, "acc_step": 0, "speed/wps": 12960.792360016983, "speed/FLOPS": 203567054009123.5, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.077972888946533, "optim/lr": 0.0024300000000000003, "optim/total_tokens": 1061945344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 5.033821105957031, "created_at": "2025-01-14T18:43:03.294042+00:00"} {"global_step": 4052, "acc_step": 0, "speed/wps": 12950.868354852097, "speed/FLOPS": 203411183870996.78, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.312270641326904, "optim/lr": 0.0024306, "optim/total_tokens": 1062207488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 4.8887410163879395, "created_at": "2025-01-14T18:43:08.358345+00:00"} {"global_step": 4053, "acc_step": 0, "speed/wps": 12951.42760893746, "speed/FLOPS": 203419967724903.06, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.851893901824951, "optim/lr": 0.0024312, "optim/total_tokens": 1062469632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 5.105534076690674, "created_at": "2025-01-14T18:43:13.422106+00:00"} {"global_step": 4054, "acc_step": 0, "speed/wps": 12951.673453936895, "speed/FLOPS": 203423829058445.62, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.366090297698975, "optim/lr": 0.0024318, "optim/total_tokens": 1062731776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 341043, "loss/out": 4.947338104248047, "created_at": "2025-01-14T18:43:18.486185+00:00"} {"global_step": 4055, "acc_step": 0, "speed/wps": 12956.39971664988, "speed/FLOPS": 203498061508918.6, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.290485382080078, "optim/lr": 0.0024324, "optim/total_tokens": 1062993920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359360, "loss/out": 4.91274881362915, "created_at": "2025-01-14T18:43:23.550451+00:00"} {"global_step": 4056, "acc_step": 0, "speed/wps": 12953.733013409923, "speed/FLOPS": 203456177270100.8, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.618721961975098, "optim/lr": 0.0024330000000000003, "optim/total_tokens": 1063256064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461913, "loss/out": 4.8545732498168945, "created_at": "2025-01-14T18:43:28.613331+00:00"} {"global_step": 4057, "acc_step": 0, "speed/wps": 12954.262190194604, "speed/FLOPS": 203464488718669.34, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.8070058822631836, "optim/lr": 0.0024336, "optim/total_tokens": 1063518208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 5.1422119140625, "created_at": "2025-01-14T18:43:33.684466+00:00"} {"global_step": 4058, "acc_step": 0, "speed/wps": 12954.007578623621, "speed/FLOPS": 203460489694074.1, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.2411651611328125, "optim/lr": 0.0024342, "optim/total_tokens": 1063780352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 4.886345386505127, "created_at": "2025-01-14T18:43:38.749374+00:00"} {"global_step": 4059, "acc_step": 0, "speed/wps": 12947.890319988035, "speed/FLOPS": 203364409741207.97, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.300212860107422, "optim/lr": 0.0024348, "optim/total_tokens": 1064042496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416669, "loss/out": 4.988255500793457, "created_at": "2025-01-14T18:43:43.816632+00:00"} {"global_step": 4060, "acc_step": 0, "speed/wps": 12949.627241430957, "speed/FLOPS": 203391690479249.22, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.3149943351745605, "optim/lr": 0.0024354, "optim/total_tokens": 1064304640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370920, "loss/out": 4.941000461578369, "created_at": "2025-01-14T18:43:48.881708+00:00"} {"global_step": 4061, "acc_step": 0, "speed/wps": 12944.774505600099, "speed/FLOPS": 203315471594667.78, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.539102554321289, "optim/lr": 0.0024360000000000002, "optim/total_tokens": 1064566784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468255, "loss/out": 4.980260848999023, "created_at": "2025-01-14T18:43:53.953349+00:00"} {"global_step": 4062, "acc_step": 0, "speed/wps": 12953.095916246171, "speed/FLOPS": 203446170783680.84, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0012, "optim/grad_norm": 6.976318836212158, "optim/lr": 0.0024366, "optim/total_tokens": 1064828928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 5.003624439239502, "created_at": "2025-01-14T18:43:59.020435+00:00"} {"global_step": 4063, "acc_step": 0, "speed/wps": 12955.34536226221, "speed/FLOPS": 203481501424425.47, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0009, "optim/grad_norm": 8.878594398498535, "optim/lr": 0.0024372, "optim/total_tokens": 1065091072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453945, "loss/out": 5.068641662597656, "created_at": "2025-01-14T18:44:04.083786+00:00"} {"global_step": 4064, "acc_step": 0, "speed/wps": 12947.8005701052, "speed/FLOPS": 203363000095968.78, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.787765026092529, "optim/lr": 0.0024378, "optim/total_tokens": 1065353216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427066, "loss/out": 5.0568437576293945, "created_at": "2025-01-14T18:44:09.150916+00:00"} {"global_step": 4065, "acc_step": 0, "speed/wps": 12952.464498691736, "speed/FLOPS": 203436253503329.8, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.014291286468506, "optim/lr": 0.0024384, "optim/total_tokens": 1065615360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366844, "loss/out": 5.042754650115967, "created_at": "2025-01-14T18:44:14.214596+00:00"} {"global_step": 4066, "acc_step": 0, "speed/wps": 12951.375143443318, "speed/FLOPS": 203419143682222.4, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.1797261238098145, "optim/lr": 0.0024389999999999998, "optim/total_tokens": 1065877504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 4.923297882080078, "created_at": "2025-01-14T18:44:19.291175+00:00"} {"global_step": 4067, "acc_step": 0, "speed/wps": 12944.56460455577, "speed/FLOPS": 203312174810332.3, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.967708110809326, "optim/lr": 0.0024396, "optim/total_tokens": 1066139648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440628, "loss/out": 4.985446929931641, "created_at": "2025-01-14T18:44:24.358402+00:00"} {"global_step": 4068, "acc_step": 0, "speed/wps": 12953.127823502236, "speed/FLOPS": 203446671930982.53, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.930548191070557, "optim/lr": 0.0024402, "optim/total_tokens": 1066401792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 5.024166107177734, "created_at": "2025-01-14T18:44:29.421993+00:00"} {"global_step": 4069, "acc_step": 0, "speed/wps": 12958.976193201099, "speed/FLOPS": 203538528613605.28, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.499526500701904, "optim/lr": 0.0024408, "optim/total_tokens": 1066663936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 340231, "loss/out": 4.88079833984375, "created_at": "2025-01-14T18:44:34.487514+00:00"} {"global_step": 4070, "acc_step": 0, "speed/wps": 12946.743901208365, "speed/FLOPS": 203346403658928.75, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.054960250854492, "optim/lr": 0.0024414, "optim/total_tokens": 1066926080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479416, "loss/out": 5.071707248687744, "created_at": "2025-01-14T18:44:39.553186+00:00"} {"global_step": 4071, "acc_step": 0, "speed/wps": 12950.868891143102, "speed/FLOPS": 203411192294183.97, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.168923854827881, "optim/lr": 0.0024419999999999997, "optim/total_tokens": 1067188224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 5.094953536987305, "created_at": "2025-01-14T18:44:44.617438+00:00"} {"global_step": 4072, "acc_step": 0, "speed/wps": 12952.657985847993, "speed/FLOPS": 203439292484998.03, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0009, "optim/grad_norm": 11.923529624938965, "optim/lr": 0.0024426, "optim/total_tokens": 1067450368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430506, "loss/out": 5.207099437713623, "created_at": "2025-01-14T18:44:49.680687+00:00"} {"global_step": 4073, "acc_step": 0, "speed/wps": 12952.133641324113, "speed/FLOPS": 203431056933725.62, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.6662075519561768, "optim/lr": 0.0024432, "optim/total_tokens": 1067712512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 5.272022247314453, "created_at": "2025-01-14T18:44:54.744852+00:00"} {"global_step": 4074, "acc_step": 0, "speed/wps": 12953.604999032534, "speed/FLOPS": 203454166628393.72, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.9022064208984375, "optim/lr": 0.0024438, "optim/total_tokens": 1067974656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421476, "loss/out": 5.061159610748291, "created_at": "2025-01-14T18:44:59.812424+00:00"} {"global_step": 4075, "acc_step": 0, "speed/wps": 12950.463586986545, "speed/FLOPS": 203404826435458.97, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.940707206726074, "optim/lr": 0.0024444, "optim/total_tokens": 1068236800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309558, "loss/out": 5.041624546051025, "created_at": "2025-01-14T18:45:04.877661+00:00"} {"global_step": 4076, "acc_step": 0, "speed/wps": 12955.378117797969, "speed/FLOPS": 203482015895125.4, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.80838394165039, "optim/lr": 0.0024449999999999997, "optim/total_tokens": 1068498944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411602, "loss/out": 5.031581878662109, "created_at": "2025-01-14T18:45:09.956644+00:00"} {"global_step": 4077, "acc_step": 0, "speed/wps": 12946.304237500106, "speed/FLOPS": 203339498136229.3, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.001, "optim/grad_norm": 7.295024394989014, "optim/lr": 0.0024456, "optim/total_tokens": 1068761088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435866, "loss/out": 5.114927768707275, "created_at": "2025-01-14T18:45:15.022837+00:00"} {"global_step": 4078, "acc_step": 0, "speed/wps": 12957.899157066657, "speed/FLOPS": 203521612281111.12, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.442941188812256, "optim/lr": 0.0024462, "optim/total_tokens": 1069023232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 5.07098913192749, "created_at": "2025-01-14T18:45:20.085571+00:00"} {"global_step": 4079, "acc_step": 0, "speed/wps": 12954.63389746656, "speed/FLOPS": 203470326892154.72, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.507209539413452, "optim/lr": 0.0024468, "optim/total_tokens": 1069285376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430791, "loss/out": 5.052399158477783, "created_at": "2025-01-14T18:45:25.149065+00:00"} {"global_step": 4080, "acc_step": 0, "speed/wps": 12956.249962972717, "speed/FLOPS": 203495709421635.66, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.270434856414795, "optim/lr": 0.0024473999999999997, "optim/total_tokens": 1069547520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356542, "loss/out": 5.098716735839844, "created_at": "2025-01-14T18:45:30.211953+00:00"} {"global_step": 4081, "acc_step": 0, "speed/wps": 12950.80680562434, "speed/FLOPS": 203410217155797.34, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0012, "optim/grad_norm": 2.657099962234497, "optim/lr": 0.002448, "optim/total_tokens": 1069809664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 5.1926727294921875, "created_at": "2025-01-14T18:45:35.276869+00:00"} {"global_step": 4082, "acc_step": 0, "speed/wps": 12956.29990354044, "speed/FLOPS": 203496493806877.25, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.673308849334717, "optim/lr": 0.0024486, "optim/total_tokens": 1070071808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 5.335456848144531, "created_at": "2025-01-14T18:45:40.342560+00:00"} {"global_step": 4083, "acc_step": 0, "speed/wps": 12946.354334846144, "speed/FLOPS": 203340284983889.8, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.183523654937744, "optim/lr": 0.0024492, "optim/total_tokens": 1070333952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 5.363605499267578, "created_at": "2025-01-14T18:45:45.410276+00:00"} {"global_step": 4084, "acc_step": 0, "speed/wps": 12946.407448031387, "speed/FLOPS": 203341119199449.2, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.765763759613037, "optim/lr": 0.0024498000000000002, "optim/total_tokens": 1070596096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322311, "loss/out": 5.210583686828613, "created_at": "2025-01-14T18:45:50.476755+00:00"} {"global_step": 4085, "acc_step": 0, "speed/wps": 12943.116728064842, "speed/FLOPS": 203289433920451.53, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.8571174144744873, "optim/lr": 0.0024504, "optim/total_tokens": 1070858240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 5.178603172302246, "created_at": "2025-01-14T18:45:55.544512+00:00"} {"global_step": 4086, "acc_step": 0, "speed/wps": 12948.034760542409, "speed/FLOPS": 203366678378596.53, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.378597259521484, "optim/lr": 0.002451, "optim/total_tokens": 1071120384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 5.093679428100586, "created_at": "2025-01-14T18:46:00.610121+00:00"} {"global_step": 4087, "acc_step": 0, "speed/wps": 12939.912030334472, "speed/FLOPS": 203239099738879.88, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.0135295391082764, "optim/lr": 0.0024516, "optim/total_tokens": 1071382528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439583, "loss/out": 5.151222229003906, "created_at": "2025-01-14T18:46:05.682817+00:00"} {"global_step": 4088, "acc_step": 0, "speed/wps": 12945.530391964294, "speed/FLOPS": 203327343828714.16, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5057759284973145, "optim/lr": 0.0024522000000000003, "optim/total_tokens": 1071644672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 5.034035682678223, "created_at": "2025-01-14T18:46:10.752876+00:00"} {"global_step": 4089, "acc_step": 0, "speed/wps": 12948.732698239415, "speed/FLOPS": 203377640449194.9, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.391907215118408, "optim/lr": 0.0024528, "optim/total_tokens": 1071906816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322156, "loss/out": 5.187925338745117, "created_at": "2025-01-14T18:46:15.823813+00:00"} {"global_step": 4090, "acc_step": 0, "speed/wps": 12948.126028945697, "speed/FLOPS": 203368111874286.1, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.218050718307495, "optim/lr": 0.0024534, "optim/total_tokens": 1072168960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446459, "loss/out": 5.280490875244141, "created_at": "2025-01-14T18:46:20.888976+00:00"} {"global_step": 4091, "acc_step": 0, "speed/wps": 12948.576674230162, "speed/FLOPS": 203375189877730.12, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.659233331680298, "optim/lr": 0.002454, "optim/total_tokens": 1072431104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 5.126481056213379, "created_at": "2025-01-14T18:46:25.954806+00:00"} {"global_step": 4092, "acc_step": 0, "speed/wps": 12946.750244471003, "speed/FLOPS": 203346503288585.03, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.794276714324951, "optim/lr": 0.0024546000000000004, "optim/total_tokens": 1072693248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435567, "loss/out": 5.025209903717041, "created_at": "2025-01-14T18:46:31.028201+00:00"} {"global_step": 4093, "acc_step": 0, "speed/wps": 12944.698210064478, "speed/FLOPS": 203314273268439.8, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.8584487438201904, "optim/lr": 0.0024552000000000003, "optim/total_tokens": 1072955392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 339985, "loss/out": 5.091176986694336, "created_at": "2025-01-14T18:46:36.099769+00:00"} {"global_step": 4094, "acc_step": 0, "speed/wps": 12950.235476614034, "speed/FLOPS": 203401243648602.2, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.196387767791748, "optim/lr": 0.0024558, "optim/total_tokens": 1073217536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453382, "loss/out": 5.109410285949707, "created_at": "2025-01-14T18:46:41.166629+00:00"} {"global_step": 4095, "acc_step": 0, "speed/wps": 12939.22679179587, "speed/FLOPS": 203228337126014.75, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.706366777420044, "optim/lr": 0.0024564, "optim/total_tokens": 1073479680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 5.090670585632324, "created_at": "2025-01-14T18:46:46.235102+00:00"} {"global_step": 4096, "acc_step": 0, "speed/wps": 12944.934833930974, "speed/FLOPS": 203317989771418.4, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9785816669464111, "optim/lr": 0.002457, "optim/total_tokens": 1073741824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 4.981335163116455, "created_at": "2025-01-14T18:46:51.301683+00:00"} {"global_step": 4097, "acc_step": 0, "speed/wps": 12940.20262424431, "speed/FLOPS": 203243663915551.78, "speed/curr_iter_time": 1.2774, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.588038206100464, "optim/lr": 0.0024576000000000003, "optim/total_tokens": 1074003968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319225, "loss/out": 5.099428176879883, "created_at": "2025-01-14T18:46:56.370376+00:00"} {"global_step": 4098, "acc_step": 0, "speed/wps": 12945.196545220466, "speed/FLOPS": 203322100306848.84, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.876362800598145, "optim/lr": 0.0024582000000000002, "optim/total_tokens": 1074266112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 414480, "loss/out": 5.018455505371094, "created_at": "2025-01-14T18:47:01.436884+00:00"} {"global_step": 4099, "acc_step": 0, "speed/wps": 12941.307691295893, "speed/FLOPS": 203261020512118.16, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.9022161960601807, "optim/lr": 0.0024588, "optim/total_tokens": 1074528256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 5.1809492111206055, "created_at": "2025-01-14T18:47:06.505167+00:00"} {"global_step": 4100, "acc_step": 0, "speed/wps": 12944.771340889121, "speed/FLOPS": 203315421888533.1, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.328989505767822, "optim/lr": 0.0024594, "optim/total_tokens": 1074790400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435567, "loss/out": 5.145750999450684, "created_at": "2025-01-14T18:47:11.574621+00:00"} {"global_step": 4101, "acc_step": 0, "speed/wps": 12945.211152917685, "speed/FLOPS": 203322329740806.75, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.785117149353027, "optim/lr": 0.00246, "optim/total_tokens": 1075052544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 5.09935188293457, "created_at": "2025-01-14T18:47:16.642893+00:00"} {"global_step": 4102, "acc_step": 0, "speed/wps": 12946.730461992025, "speed/FLOPS": 203346192577568.72, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.50494647026062, "optim/lr": 0.0024606000000000003, "optim/total_tokens": 1075314688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448606, "loss/out": 5.2727789878845215, "created_at": "2025-01-14T18:47:21.708823+00:00"} {"global_step": 4103, "acc_step": 0, "speed/wps": 12947.573694658346, "speed/FLOPS": 203359436705316.4, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.840763568878174, "optim/lr": 0.0024612, "optim/total_tokens": 1075576832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364158, "loss/out": 5.076356887817383, "created_at": "2025-01-14T18:47:26.778456+00:00"} {"global_step": 4104, "acc_step": 0, "speed/wps": 12947.986677659164, "speed/FLOPS": 203365923170842.38, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0012, "optim/grad_norm": 4.428947925567627, "optim/lr": 0.0024618, "optim/total_tokens": 1075838976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464799, "loss/out": 5.220416069030762, "created_at": "2025-01-14T18:47:31.844256+00:00"} {"global_step": 4105, "acc_step": 0, "speed/wps": 12944.025631439634, "speed/FLOPS": 203303709496916.75, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.7199482917785645, "optim/lr": 0.0024624, "optim/total_tokens": 1076101120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448890, "loss/out": 5.259453773498535, "created_at": "2025-01-14T18:47:36.913123+00:00"} {"global_step": 4106, "acc_step": 0, "speed/wps": 12941.469197571001, "speed/FLOPS": 203263557190101.6, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.410376787185669, "optim/lr": 0.002463, "optim/total_tokens": 1076363264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462435, "loss/out": 5.062966346740723, "created_at": "2025-01-14T18:47:41.988614+00:00"} {"global_step": 4107, "acc_step": 0, "speed/wps": 12937.48549710656, "speed/FLOPS": 203200987700129.47, "speed/curr_iter_time": 1.2787, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.068126678466797, "optim/lr": 0.0024636000000000003, "optim/total_tokens": 1076625408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355325, "loss/out": 5.357804298400879, "created_at": "2025-01-14T18:47:47.057719+00:00"} {"global_step": 4108, "acc_step": 0, "speed/wps": 12936.662894346147, "speed/FLOPS": 203188067593402.53, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.244356155395508, "optim/lr": 0.0024642, "optim/total_tokens": 1076887552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 5.128830909729004, "created_at": "2025-01-14T18:47:52.152480+00:00"} {"global_step": 4109, "acc_step": 0, "speed/wps": 12944.332796921977, "speed/FLOPS": 203308533952906.53, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.531252384185791, "optim/lr": 0.0024648, "optim/total_tokens": 1077149696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 5.229689121246338, "created_at": "2025-01-14T18:47:57.226915+00:00"} {"global_step": 4110, "acc_step": 0, "speed/wps": 12945.242866546634, "speed/FLOPS": 203322827846928.5, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0009, "optim/grad_norm": 2.604112148284912, "optim/lr": 0.0024654, "optim/total_tokens": 1077411840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356143, "loss/out": 5.068774700164795, "created_at": "2025-01-14T18:48:02.293940+00:00"} {"global_step": 4111, "acc_step": 0, "speed/wps": 12949.440066793435, "speed/FLOPS": 203388750644362.03, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.130455017089844, "optim/lr": 0.002466, "optim/total_tokens": 1077673984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472260, "loss/out": 5.16283655166626, "created_at": "2025-01-14T18:48:07.361789+00:00"} {"global_step": 4112, "acc_step": 0, "speed/wps": 12941.765388168056, "speed/FLOPS": 203268209270436.88, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5553245544433594, "optim/lr": 0.0024666, "optim/total_tokens": 1077936128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447791, "loss/out": 5.104554176330566, "created_at": "2025-01-14T18:48:12.431135+00:00"} {"global_step": 4113, "acc_step": 0, "speed/wps": 12942.673635178011, "speed/FLOPS": 203282474537790.4, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.112260103225708, "optim/lr": 0.0024672, "optim/total_tokens": 1078198272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 5.10517692565918, "created_at": "2025-01-14T18:48:17.498373+00:00"} {"global_step": 4114, "acc_step": 0, "speed/wps": 12937.439207493537, "speed/FLOPS": 203200260658148.34, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.001, "optim/grad_norm": 2.918581247329712, "optim/lr": 0.0024678, "optim/total_tokens": 1078460416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289068, "loss/out": 5.111881256103516, "created_at": "2025-01-14T18:48:22.574438+00:00"} {"global_step": 4115, "acc_step": 0, "speed/wps": 12945.215070377144, "speed/FLOPS": 203322391269890.72, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.556755781173706, "optim/lr": 0.0024684, "optim/total_tokens": 1078722560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295714, "loss/out": 5.108656406402588, "created_at": "2025-01-14T18:48:27.649155+00:00"} {"global_step": 4116, "acc_step": 0, "speed/wps": 12938.473501705108, "speed/FLOPS": 203216505670010.25, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0008, "optim/grad_norm": 15.51995849609375, "optim/lr": 0.002469, "optim/total_tokens": 1078984704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 5.272579669952393, "created_at": "2025-01-14T18:48:32.717930+00:00"} {"global_step": 4117, "acc_step": 0, "speed/wps": 12940.26096256502, "speed/FLOPS": 203244580199043.78, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.979137897491455, "optim/lr": 0.0024696, "optim/total_tokens": 1079246848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413505, "loss/out": 5.545823574066162, "created_at": "2025-01-14T18:48:37.789236+00:00"} {"global_step": 4118, "acc_step": 0, "speed/wps": 12941.64691411051, "speed/FLOPS": 203266348472562.7, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.4403860569000244, "optim/lr": 0.0024702, "optim/total_tokens": 1079508992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452081, "loss/out": 5.189607620239258, "created_at": "2025-01-14T18:48:42.856883+00:00"} {"global_step": 4119, "acc_step": 0, "speed/wps": 12944.173576406327, "speed/FLOPS": 203306033175916.34, "speed/curr_iter_time": 1.2771, "speed/data_load_time": 0.0012, "optim/grad_norm": 4.477667808532715, "optim/lr": 0.0024708, "optim/total_tokens": 1079771136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305807, "loss/out": 5.137317657470703, "created_at": "2025-01-14T18:48:47.927324+00:00"} {"global_step": 4120, "acc_step": 0, "speed/wps": 12938.307386147544, "speed/FLOPS": 203213896596912.0, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.627490758895874, "optim/lr": 0.0024714, "optim/total_tokens": 1080033280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 5.1786651611328125, "created_at": "2025-01-14T18:48:52.996791+00:00"} {"global_step": 4121, "acc_step": 0, "speed/wps": 12941.490192088213, "speed/FLOPS": 203263886937843.9, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.642612338066101, "optim/lr": 0.002472, "optim/total_tokens": 1080295424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427880, "loss/out": 5.172853469848633, "created_at": "2025-01-14T18:48:58.064557+00:00"} {"global_step": 4122, "acc_step": 0, "speed/wps": 12947.675207763867, "speed/FLOPS": 203361031108132.22, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.264960765838623, "optim/lr": 0.0024726, "optim/total_tokens": 1080557568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 5.210125923156738, "created_at": "2025-01-14T18:49:03.130917+00:00"} {"global_step": 4123, "acc_step": 0, "speed/wps": 12941.128368098898, "speed/FLOPS": 203258203994893.44, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.924402952194214, "optim/lr": 0.0024732, "optim/total_tokens": 1080819712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319909, "loss/out": 5.134415626525879, "created_at": "2025-01-14T18:49:08.201128+00:00"} {"global_step": 4124, "acc_step": 0, "speed/wps": 12943.017029044053, "speed/FLOPS": 203287868010329.22, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.9642937183380127, "optim/lr": 0.0024738, "optim/total_tokens": 1081081856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443835, "loss/out": 5.0313262939453125, "created_at": "2025-01-14T18:49:13.273124+00:00"} {"global_step": 4125, "acc_step": 0, "speed/wps": 12944.671440164904, "speed/FLOPS": 203313852810381.28, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0008, "optim/grad_norm": 24.415142059326172, "optim/lr": 0.0024744, "optim/total_tokens": 1081344000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444644, "loss/out": 5.207792282104492, "created_at": "2025-01-14T18:49:18.341719+00:00"} {"global_step": 4126, "acc_step": 0, "speed/wps": 12949.590599624575, "speed/FLOPS": 203391114969328.88, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.463042974472046, "optim/lr": 0.0024749999999999998, "optim/total_tokens": 1081606144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 5.331851482391357, "created_at": "2025-01-14T18:49:23.406419+00:00"} {"global_step": 4127, "acc_step": 0, "speed/wps": 12937.770373069876, "speed/FLOPS": 203205462068593.94, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.0592517852783203, "optim/lr": 0.0024756, "optim/total_tokens": 1081868288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432365, "loss/out": 5.110426902770996, "created_at": "2025-01-14T18:49:28.477195+00:00"} {"global_step": 4128, "acc_step": 0, "speed/wps": 12947.241772137544, "speed/FLOPS": 203354223406017.88, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8724346160888672, "optim/lr": 0.0024762, "optim/total_tokens": 1082130432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323417, "loss/out": 5.145068168640137, "created_at": "2025-01-14T18:49:33.542583+00:00"} {"global_step": 4129, "acc_step": 0, "speed/wps": 12945.73387012179, "speed/FLOPS": 203330539732785.94, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.722508907318115, "optim/lr": 0.0024768, "optim/total_tokens": 1082392576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 394230, "loss/out": 5.071969032287598, "created_at": "2025-01-14T18:49:38.612624+00:00"} {"global_step": 4130, "acc_step": 0, "speed/wps": 12945.188862737568, "speed/FLOPS": 203321979642898.1, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.4023592472076416, "optim/lr": 0.0024774, "optim/total_tokens": 1082654720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 5.099691390991211, "created_at": "2025-01-14T18:49:43.684850+00:00"} {"global_step": 4131, "acc_step": 0, "speed/wps": 12947.951512720421, "speed/FLOPS": 203365370857157.72, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0009, "optim/grad_norm": 4.720544815063477, "optim/lr": 0.0024779999999999997, "optim/total_tokens": 1082916864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 5.202551364898682, "created_at": "2025-01-14T18:49:48.759165+00:00"} {"global_step": 4132, "acc_step": 0, "speed/wps": 12943.655825483203, "speed/FLOPS": 203297901186204.38, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.723893642425537, "optim/lr": 0.0024786, "optim/total_tokens": 1083179008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287957, "loss/out": 5.115530490875244, "created_at": "2025-01-14T18:49:53.825945+00:00"} {"global_step": 4133, "acc_step": 0, "speed/wps": 12947.996879339062, "speed/FLOPS": 203366083402243.72, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.936224460601807, "optim/lr": 0.0024792, "optim/total_tokens": 1083441152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 5.11764669418335, "created_at": "2025-01-14T18:49:58.891464+00:00"} {"global_step": 4134, "acc_step": 0, "speed/wps": 12939.397539655993, "speed/FLOPS": 203231018955790.1, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.646265029907227, "optim/lr": 0.0024798, "optim/total_tokens": 1083703296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471976, "loss/out": 5.128468990325928, "created_at": "2025-01-14T18:50:03.961176+00:00"} {"global_step": 4135, "acc_step": 0, "speed/wps": 12940.356678432605, "speed/FLOPS": 203246083548271.6, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.046552658081055, "optim/lr": 0.0024804, "optim/total_tokens": 1083965440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 5.285314559936523, "created_at": "2025-01-14T18:50:09.029436+00:00"} {"global_step": 4136, "acc_step": 0, "speed/wps": 12927.57781785062, "speed/FLOPS": 203045373982836.94, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.952041149139404, "optim/lr": 0.002481, "optim/total_tokens": 1084227584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326344, "loss/out": 5.284566879272461, "created_at": "2025-01-14T18:50:14.103422+00:00"} {"global_step": 4137, "acc_step": 0, "speed/wps": 12940.243710396311, "speed/FLOPS": 203244309230027.25, "speed/curr_iter_time": 1.2774, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.8390772342681885, "optim/lr": 0.0024816, "optim/total_tokens": 1084489728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 5.167750358581543, "created_at": "2025-01-14T18:50:19.173690+00:00"} {"global_step": 4138, "acc_step": 0, "speed/wps": 12943.979702162042, "speed/FLOPS": 203302988114498.72, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.569030284881592, "optim/lr": 0.0024822, "optim/total_tokens": 1084751872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 5.061453342437744, "created_at": "2025-01-14T18:50:24.240389+00:00"} {"global_step": 4139, "acc_step": 0, "speed/wps": 12946.969980545531, "speed/FLOPS": 203349954545584.5, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.923160076141357, "optim/lr": 0.0024828, "optim/total_tokens": 1085014016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431320, "loss/out": 5.092921257019043, "created_at": "2025-01-14T18:50:29.305892+00:00"} {"global_step": 4140, "acc_step": 0, "speed/wps": 12942.775559471896, "speed/FLOPS": 203284075398884.53, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.603630542755127, "optim/lr": 0.0024834, "optim/total_tokens": 1085276160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432953, "loss/out": 5.315462589263916, "created_at": "2025-01-14T18:50:34.373264+00:00"} {"global_step": 4141, "acc_step": 0, "speed/wps": 12942.161766969391, "speed/FLOPS": 203274434944193.4, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.127839088439941, "optim/lr": 0.002484, "optim/total_tokens": 1085538304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326583, "loss/out": 5.193852424621582, "created_at": "2025-01-14T18:50:39.442576+00:00"} {"global_step": 4142, "acc_step": 0, "speed/wps": 12942.023696076873, "speed/FLOPS": 203272266351097.1, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.3491458892822266, "optim/lr": 0.0024846, "optim/total_tokens": 1085800448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390139, "loss/out": 5.177433013916016, "created_at": "2025-01-14T18:50:44.510484+00:00"} {"global_step": 4143, "acc_step": 0, "speed/wps": 12939.938985188992, "speed/FLOPS": 203239523101909.97, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0006, "optim/grad_norm": 16.01950454711914, "optim/lr": 0.0024852000000000003, "optim/total_tokens": 1086062592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 5.100780487060547, "created_at": "2025-01-14T18:50:49.583470+00:00"} {"global_step": 4144, "acc_step": 0, "speed/wps": 12945.12568859288, "speed/FLOPS": 203320987406145.4, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0011, "optim/grad_norm": 7.1022796630859375, "optim/lr": 0.0024858000000000002, "optim/total_tokens": 1086324736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 5.1706438064575195, "created_at": "2025-01-14T18:50:54.649730+00:00"} {"global_step": 4145, "acc_step": 0, "speed/wps": 12947.604831432362, "speed/FLOPS": 203359925751138.72, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0006, "optim/grad_norm": 21.06827735900879, "optim/lr": 0.0024864, "optim/total_tokens": 1086586880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480954, "loss/out": 5.1382904052734375, "created_at": "2025-01-14T18:50:59.716022+00:00"} {"global_step": 4146, "acc_step": 0, "speed/wps": 12950.930135868091, "speed/FLOPS": 203412154226749.0, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 22.735002517700195, "optim/lr": 0.002487, "optim/total_tokens": 1086849024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320435, "loss/out": 5.099660873413086, "created_at": "2025-01-14T18:51:04.782578+00:00"} {"global_step": 4147, "acc_step": 0, "speed/wps": 12944.314843374146, "speed/FLOPS": 203308251967767.1, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.466361045837402, "optim/lr": 0.0024876000000000004, "optim/total_tokens": 1087111168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 5.193075180053711, "created_at": "2025-01-14T18:51:09.856727+00:00"} {"global_step": 4148, "acc_step": 0, "speed/wps": 12951.542376602049, "speed/FLOPS": 203421770308782.72, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.704592704772949, "optim/lr": 0.0024882000000000003, "optim/total_tokens": 1087373312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457665, "loss/out": 5.249354362487793, "created_at": "2025-01-14T18:51:14.920722+00:00"} {"global_step": 4149, "acc_step": 0, "speed/wps": 12945.709911009328, "speed/FLOPS": 203330163422001.44, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0006, "optim/grad_norm": 70.66398620605469, "optim/lr": 0.0024888, "optim/total_tokens": 1087635456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 5.226093292236328, "created_at": "2025-01-14T18:51:19.988512+00:00"} {"global_step": 4150, "acc_step": 0, "speed/wps": 12942.277432113446, "speed/FLOPS": 203276251624221.34, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0008, "optim/grad_norm": 19.136920928955078, "optim/lr": 0.0024894, "optim/total_tokens": 1087897600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317059, "loss/out": 5.1833109855651855, "created_at": "2025-01-14T18:51:25.055971+00:00"} {"global_step": 4151, "acc_step": 0, "speed/wps": 12945.71465042911, "speed/FLOPS": 203330237861101.78, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0006, "optim/grad_norm": 19.5074462890625, "optim/lr": 0.00249, "optim/total_tokens": 1088159744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 5.230666160583496, "created_at": "2025-01-14T18:51:30.126809+00:00"} {"global_step": 4152, "acc_step": 0, "speed/wps": 12944.175190759692, "speed/FLOPS": 203306058531554.28, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.05399227142334, "optim/lr": 0.0024906000000000004, "optim/total_tokens": 1088421888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 5.366004943847656, "created_at": "2025-01-14T18:51:35.193782+00:00"} {"global_step": 4153, "acc_step": 0, "speed/wps": 12940.548547524491, "speed/FLOPS": 203249097116015.44, "speed/curr_iter_time": 1.2781, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.34975528717041, "optim/lr": 0.0024912000000000003, "optim/total_tokens": 1088684032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459528, "loss/out": 5.2343244552612305, "created_at": "2025-01-14T18:51:40.264486+00:00"} {"global_step": 4154, "acc_step": 0, "speed/wps": 12935.149090450957, "speed/FLOPS": 203164291223044.03, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.67346477508545, "optim/lr": 0.0024918, "optim/total_tokens": 1088946176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460271, "loss/out": 5.29649543762207, "created_at": "2025-01-14T18:51:45.335400+00:00"} {"global_step": 4155, "acc_step": 0, "speed/wps": 12937.448367392215, "speed/FLOPS": 203200404526944.1, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 15.556896209716797, "optim/lr": 0.0024924, "optim/total_tokens": 1089208320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451523, "loss/out": 5.276863098144531, "created_at": "2025-01-14T18:51:50.413070+00:00"} {"global_step": 4156, "acc_step": 0, "speed/wps": 12936.447115737126, "speed/FLOPS": 203184678493836.97, "speed/curr_iter_time": 1.2778, "speed/data_load_time": 0.0006, "optim/grad_norm": 30.88677215576172, "optim/lr": 0.002493, "optim/total_tokens": 1089470464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 5.256300926208496, "created_at": "2025-01-14T18:51:55.484474+00:00"} {"global_step": 4157, "acc_step": 0, "speed/wps": 12946.237048642994, "speed/FLOPS": 203338442842901.2, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 22.355224609375, "optim/lr": 0.0024936000000000003, "optim/total_tokens": 1089732608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457380, "loss/out": 5.328534126281738, "created_at": "2025-01-14T18:52:00.552416+00:00"} {"global_step": 4158, "acc_step": 0, "speed/wps": 12945.45155834125, "speed/FLOPS": 203326105638335.53, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.497221946716309, "optim/lr": 0.0024942000000000002, "optim/total_tokens": 1089994752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 5.367190361022949, "created_at": "2025-01-14T18:52:05.622305+00:00"} {"global_step": 4159, "acc_step": 0, "speed/wps": 12941.668525658066, "speed/FLOPS": 203266687911614.44, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.260051250457764, "optim/lr": 0.0024948, "optim/total_tokens": 1090256896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338354, "loss/out": 5.486021995544434, "created_at": "2025-01-14T18:52:10.694378+00:00"} {"global_step": 4160, "acc_step": 0, "speed/wps": 12942.048277078902, "speed/FLOPS": 203272652429511.8, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.046412467956543, "optim/lr": 0.0024954, "optim/total_tokens": 1090519040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456312, "loss/out": 5.390398979187012, "created_at": "2025-01-14T18:52:15.763387+00:00"} {"global_step": 4161, "acc_step": 0, "speed/wps": 12941.389723194376, "speed/FLOPS": 203262308935807.47, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.949774742126465, "optim/lr": 0.002496, "optim/total_tokens": 1090781184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 5.353643894195557, "created_at": "2025-01-14T18:52:20.831304+00:00"} {"global_step": 4162, "acc_step": 0, "speed/wps": 12939.095902033465, "speed/FLOPS": 203226281322434.78, "speed/curr_iter_time": 1.2784, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.416817665100098, "optim/lr": 0.0024966000000000003, "optim/total_tokens": 1091043328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 5.350144863128662, "created_at": "2025-01-14T18:52:25.907995+00:00"} {"global_step": 4163, "acc_step": 0, "speed/wps": 12942.595605236473, "speed/FLOPS": 203281248970334.16, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0012, "optim/grad_norm": 7.91076135635376, "optim/lr": 0.0024972, "optim/total_tokens": 1091305472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312776, "loss/out": 5.33045768737793, "created_at": "2025-01-14T18:52:30.980554+00:00"} {"global_step": 4164, "acc_step": 0, "speed/wps": 12928.723303233763, "speed/FLOPS": 203063365404841.03, "speed/curr_iter_time": 1.2784, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.244240760803223, "optim/lr": 0.0024978, "optim/total_tokens": 1091567616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 5.175739288330078, "created_at": "2025-01-14T18:52:36.053693+00:00"} {"global_step": 4165, "acc_step": 0, "speed/wps": 12931.481847264235, "speed/FLOPS": 203106692129476.34, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.180420875549316, "optim/lr": 0.0024984, "optim/total_tokens": 1091829760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 5.335251808166504, "created_at": "2025-01-14T18:52:41.126918+00:00"} {"global_step": 4166, "acc_step": 0, "speed/wps": 12932.799948446664, "speed/FLOPS": 203127394719807.22, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.9583516120910645, "optim/lr": 0.002499, "optim/total_tokens": 1092091904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424364, "loss/out": 5.35560417175293, "created_at": "2025-01-14T18:52:46.200943+00:00"} {"global_step": 4167, "acc_step": 0, "speed/wps": 12943.611363049464, "speed/FLOPS": 203297202842584.7, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.538299560546875, "optim/lr": 0.0024996000000000003, "optim/total_tokens": 1092354048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362816, "loss/out": 5.312544822692871, "created_at": "2025-01-14T18:52:51.268261+00:00"} {"global_step": 4168, "acc_step": 0, "speed/wps": 12939.461095699058, "speed/FLOPS": 203232017190781.6, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0006, "optim/grad_norm": 17.028728485107422, "optim/lr": 0.0025002, "optim/total_tokens": 1092616192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453666, "loss/out": 5.323101043701172, "created_at": "2025-01-14T18:52:56.340454+00:00"} {"global_step": 4169, "acc_step": 0, "speed/wps": 12938.511172571883, "speed/FLOPS": 203217097342738.16, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 21.808176040649414, "optim/lr": 0.0025008, "optim/total_tokens": 1092878336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 5.349016189575195, "created_at": "2025-01-14T18:53:01.409237+00:00"} {"global_step": 4170, "acc_step": 0, "speed/wps": 12938.393757236166, "speed/FLOPS": 203215253173545.12, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.760019302368164, "optim/lr": 0.0025014, "optim/total_tokens": 1093140480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 5.363227844238281, "created_at": "2025-01-14T18:53:06.478476+00:00"} {"global_step": 4171, "acc_step": 0, "speed/wps": 12944.647923130466, "speed/FLOPS": 203313483443039.66, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 18.515798568725586, "optim/lr": 0.002502, "optim/total_tokens": 1093402624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427351, "loss/out": 5.455958843231201, "created_at": "2025-01-14T18:53:11.545215+00:00"} {"global_step": 4172, "acc_step": 0, "speed/wps": 12935.573658115936, "speed/FLOPS": 203170959641639.62, "speed/curr_iter_time": 1.2771, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.594120502471924, "optim/lr": 0.0025026000000000002, "optim/total_tokens": 1093664768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425702, "loss/out": 5.48699426651001, "created_at": "2025-01-14T18:53:16.619429+00:00"} {"global_step": 4173, "acc_step": 0, "speed/wps": 12937.382481107916, "speed/FLOPS": 203199369692311.94, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.972189903259277, "optim/lr": 0.0025032, "optim/total_tokens": 1093926912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 5.419898986816406, "created_at": "2025-01-14T18:53:21.688654+00:00"} {"global_step": 4174, "acc_step": 0, "speed/wps": 12939.188962112696, "speed/FLOPS": 203227742958857.03, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.646195650100708, "optim/lr": 0.0025038, "optim/total_tokens": 1094189056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 5.3780436515808105, "created_at": "2025-01-14T18:53:26.759163+00:00"} {"global_step": 4175, "acc_step": 0, "speed/wps": 12941.210707119062, "speed/FLOPS": 203259497242350.78, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.876653671264648, "optim/lr": 0.0025044, "optim/total_tokens": 1094451200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337583, "loss/out": 5.328064918518066, "created_at": "2025-01-14T18:53:31.832934+00:00"} {"global_step": 4176, "acc_step": 0, "speed/wps": 12939.294928609648, "speed/FLOPS": 203229407308304.9, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0006, "optim/grad_norm": 15.3460054397583, "optim/lr": 0.002505, "optim/total_tokens": 1094713344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456567, "loss/out": 5.442129135131836, "created_at": "2025-01-14T18:53:36.906561+00:00"} {"global_step": 4177, "acc_step": 0, "speed/wps": 12940.208842119087, "speed/FLOPS": 203243761575819.4, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.310418128967285, "optim/lr": 0.0025056, "optim/total_tokens": 1094975488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441158, "loss/out": 5.465730667114258, "created_at": "2025-01-14T18:53:41.975822+00:00"} {"global_step": 4178, "acc_step": 0, "speed/wps": 12937.047546546997, "speed/FLOPS": 203194109084784.4, "speed/curr_iter_time": 1.2779, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.348109722137451, "optim/lr": 0.0025062, "optim/total_tokens": 1095237632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 5.614498615264893, "created_at": "2025-01-14T18:53:47.045499+00:00"} {"global_step": 4179, "acc_step": 0, "speed/wps": 12941.15126070221, "speed/FLOPS": 203258563554686.2, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0009, "optim/grad_norm": 8.416741371154785, "optim/lr": 0.0025068, "optim/total_tokens": 1095499776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333550, "loss/out": 5.387088775634766, "created_at": "2025-01-14T18:53:52.116654+00:00"} {"global_step": 4180, "acc_step": 0, "speed/wps": 12938.689067193724, "speed/FLOPS": 203219891422226.97, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.109192371368408, "optim/lr": 0.0025074, "optim/total_tokens": 1095761920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390139, "loss/out": 5.550741672515869, "created_at": "2025-01-14T18:53:57.185637+00:00"} {"global_step": 4181, "acc_step": 0, "speed/wps": 12934.108421446419, "speed/FLOPS": 203147946086298.7, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.387475490570068, "optim/lr": 0.002508, "optim/total_tokens": 1096024064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 5.382241725921631, "created_at": "2025-01-14T18:54:02.261334+00:00"} {"global_step": 4182, "acc_step": 0, "speed/wps": 12940.89820899975, "speed/FLOPS": 203254589029969.6, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.539365768432617, "optim/lr": 0.0025086, "optim/total_tokens": 1096286208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419618, "loss/out": 5.442636013031006, "created_at": "2025-01-14T18:54:07.331458+00:00"} {"global_step": 4183, "acc_step": 0, "speed/wps": 12938.026887297054, "speed/FLOPS": 203209490977019.25, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.578742027282715, "optim/lr": 0.0025092, "optim/total_tokens": 1096548352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441679, "loss/out": 5.403132438659668, "created_at": "2025-01-14T18:54:12.403289+00:00"} {"global_step": 4184, "acc_step": 0, "speed/wps": 12938.147804755628, "speed/FLOPS": 203211390151864.16, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.806108474731445, "optim/lr": 0.0025098, "optim/total_tokens": 1096810496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 387743, "loss/out": 5.385088920593262, "created_at": "2025-01-14T18:54:17.477753+00:00"} {"global_step": 4185, "acc_step": 0, "speed/wps": 12939.981402192228, "speed/FLOPS": 203240189319232.88, "speed/curr_iter_time": 1.2771, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.418623447418213, "optim/lr": 0.0025104, "optim/total_tokens": 1097072640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480663, "loss/out": 5.34981632232666, "created_at": "2025-01-14T18:54:22.546062+00:00"} {"global_step": 4186, "acc_step": 0, "speed/wps": 12945.051394034812, "speed/FLOPS": 203319820508019.53, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.89776611328125, "optim/lr": 0.0025109999999999998, "optim/total_tokens": 1097334784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442208, "loss/out": 5.183530330657959, "created_at": "2025-01-14T18:54:27.612779+00:00"} {"global_step": 4187, "acc_step": 0, "speed/wps": 12945.586218143382, "speed/FLOPS": 203328220655569.97, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.830156326293945, "optim/lr": 0.0025116, "optim/total_tokens": 1097596928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443835, "loss/out": 5.311899185180664, "created_at": "2025-01-14T18:54:32.678917+00:00"} {"global_step": 4188, "acc_step": 0, "speed/wps": 12942.823803626094, "speed/FLOPS": 203284833139620.62, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.5157408714294434, "optim/lr": 0.0025122, "optim/total_tokens": 1097859072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 341575, "loss/out": 5.252145767211914, "created_at": "2025-01-14T18:54:37.753914+00:00"} {"global_step": 4189, "acc_step": 0, "speed/wps": 12947.38433293352, "speed/FLOPS": 203356462519217.72, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0009, "optim/grad_norm": 6.664470672607422, "optim/lr": 0.0025128, "optim/total_tokens": 1098121216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 5.3029351234436035, "created_at": "2025-01-14T18:54:42.819971+00:00"} {"global_step": 4190, "acc_step": 0, "speed/wps": 12942.70631238505, "speed/FLOPS": 203282987778230.12, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.634846687316895, "optim/lr": 0.0025134, "optim/total_tokens": 1098383360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 5.368427753448486, "created_at": "2025-01-14T18:54:47.887789+00:00"} {"global_step": 4191, "acc_step": 0, "speed/wps": 12939.687291943786, "speed/FLOPS": 203235569913631.28, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.70120906829834, "optim/lr": 0.0025139999999999997, "optim/total_tokens": 1098645504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436679, "loss/out": 5.367112159729004, "created_at": "2025-01-14T18:54:52.965454+00:00"} {"global_step": 4192, "acc_step": 0, "speed/wps": 12942.022748733065, "speed/FLOPS": 203272251471760.8, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.656330585479736, "optim/lr": 0.0025146, "optim/total_tokens": 1098907648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427332, "loss/out": 5.2596869468688965, "created_at": "2025-01-14T18:54:58.033540+00:00"} {"global_step": 4193, "acc_step": 0, "speed/wps": 12951.969461001396, "speed/FLOPS": 203428478256149.94, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.468377590179443, "optim/lr": 0.0025152, "optim/total_tokens": 1099169792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460287, "loss/out": 5.289801597595215, "created_at": "2025-01-14T18:55:03.101316+00:00"} {"global_step": 4194, "acc_step": 0, "speed/wps": 12943.650497117605, "speed/FLOPS": 203297817496900.72, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.30361008644104, "optim/lr": 0.0025158, "optim/total_tokens": 1099431936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 5.403078556060791, "created_at": "2025-01-14T18:55:08.169198+00:00"} {"global_step": 4195, "acc_step": 0, "speed/wps": 12942.895558722228, "speed/FLOPS": 203285960152007.25, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.857519149780273, "optim/lr": 0.0025164000000000002, "optim/total_tokens": 1099694080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451275, "loss/out": 5.33255672454834, "created_at": "2025-01-14T18:55:13.238695+00:00"} {"global_step": 4196, "acc_step": 0, "speed/wps": 12935.582717879543, "speed/FLOPS": 203171101937676.5, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.319748878479004, "optim/lr": 0.002517, "optim/total_tokens": 1099956224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325819, "loss/out": 5.281479358673096, "created_at": "2025-01-14T18:55:18.310131+00:00"} {"global_step": 4197, "acc_step": 0, "speed/wps": 12937.280589409564, "speed/FLOPS": 203197769343175.44, "speed/curr_iter_time": 1.2771, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.62872314453125, "optim/lr": 0.0025176, "optim/total_tokens": 1100218368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463795, "loss/out": 5.412344932556152, "created_at": "2025-01-14T18:55:23.379449+00:00"} {"global_step": 4198, "acc_step": 0, "speed/wps": 12947.882853489831, "speed/FLOPS": 203364292469593.62, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.090030670166016, "optim/lr": 0.0025182, "optim/total_tokens": 1100480512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 5.409135818481445, "created_at": "2025-01-14T18:55:28.444554+00:00"} {"global_step": 4199, "acc_step": 0, "speed/wps": 12939.352285621333, "speed/FLOPS": 203230308178990.22, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.540093421936035, "optim/lr": 0.0025188000000000003, "optim/total_tokens": 1100742656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465633, "loss/out": 5.382602214813232, "created_at": "2025-01-14T18:55:33.513186+00:00"} {"global_step": 4200, "acc_step": 0, "speed/wps": 12944.66001500931, "speed/FLOPS": 203313673362612.9, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.512732028961182, "optim/lr": 0.0025194, "optim/total_tokens": 1101004800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430791, "loss/out": 5.311488628387451, "created_at": "2025-01-14T18:55:38.579608+00:00"} {"global_step": 4201, "acc_step": 0, "speed/wps": 12929.18839388562, "speed/FLOPS": 203070670292630.0, "speed/curr_iter_time": 1.2779, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.697762489318848, "optim/lr": 0.00252, "optim/total_tokens": 1101266944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355857, "loss/out": 5.265041828155518, "created_at": "2025-01-14T18:55:43.652416+00:00"} {"global_step": 4202, "acc_step": 0, "speed/wps": 12945.492063395051, "speed/FLOPS": 203326741825865.25, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.8143880367279053, "optim/lr": 0.0025206, "optim/total_tokens": 1101529088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 5.218436241149902, "created_at": "2025-01-14T18:55:48.719264+00:00"} {"global_step": 4203, "acc_step": 0, "speed/wps": 12937.300138467377, "speed/FLOPS": 203198076387991.62, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.475529432296753, "optim/lr": 0.0025212000000000004, "optim/total_tokens": 1101791232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443545, "loss/out": 5.455547332763672, "created_at": "2025-01-14T18:55:53.790621+00:00"} {"global_step": 4204, "acc_step": 0, "speed/wps": 12947.221168688813, "speed/FLOPS": 203353899800543.44, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.930027484893799, "optim/lr": 0.0025218000000000003, "optim/total_tokens": 1102053376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437717, "loss/out": 5.470118045806885, "created_at": "2025-01-14T18:55:58.856845+00:00"} {"global_step": 4205, "acc_step": 0, "speed/wps": 12945.122688026859, "speed/FLOPS": 203320940278132.84, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.496034622192383, "optim/lr": 0.0025224, "optim/total_tokens": 1102315520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328460, "loss/out": 5.442714691162109, "created_at": "2025-01-14T18:56:03.922940+00:00"} {"global_step": 4206, "acc_step": 0, "speed/wps": 12939.479064337173, "speed/FLOPS": 203232299412934.7, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.733399391174316, "optim/lr": 0.002523, "optim/total_tokens": 1102577664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 5.585431098937988, "created_at": "2025-01-14T18:56:08.992265+00:00"} {"global_step": 4207, "acc_step": 0, "speed/wps": 12945.145492363505, "speed/FLOPS": 203321298451576.3, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.001, "optim/grad_norm": 4.169675827026367, "optim/lr": 0.0025236, "optim/total_tokens": 1102839808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478284, "loss/out": 5.3732123374938965, "created_at": "2025-01-14T18:56:14.066333+00:00"} {"global_step": 4208, "acc_step": 0, "speed/wps": 12945.435804592047, "speed/FLOPS": 203325858204056.38, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.229336738586426, "optim/lr": 0.0025242000000000003, "optim/total_tokens": 1103101952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447263, "loss/out": 5.357174873352051, "created_at": "2025-01-14T18:56:19.133696+00:00"} {"global_step": 4209, "acc_step": 0, "speed/wps": 12945.531211217909, "speed/FLOPS": 203327356696217.97, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.770811557769775, "optim/lr": 0.0025248000000000002, "optim/total_tokens": 1103364096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432365, "loss/out": 5.4434814453125, "created_at": "2025-01-14T18:56:24.200795+00:00"} {"global_step": 4210, "acc_step": 0, "speed/wps": 12950.601911905218, "speed/FLOPS": 203406999018385.1, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.495867729187012, "optim/lr": 0.0025254, "optim/total_tokens": 1103626240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367993, "loss/out": 5.302624702453613, "created_at": "2025-01-14T18:56:29.266807+00:00"} {"global_step": 4211, "acc_step": 0, "speed/wps": 12947.748634289854, "speed/FLOPS": 203362184372621.47, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.612368583679199, "optim/lr": 0.002526, "optim/total_tokens": 1103888384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457380, "loss/out": 5.433750629425049, "created_at": "2025-01-14T18:56:34.334565+00:00"} {"global_step": 4212, "acc_step": 0, "speed/wps": 12943.407620681108, "speed/FLOPS": 203294002788718.3, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.842892169952393, "optim/lr": 0.0025266, "optim/total_tokens": 1104150528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443545, "loss/out": 5.439526081085205, "created_at": "2025-01-14T18:56:39.407751+00:00"} {"global_step": 4213, "acc_step": 0, "speed/wps": 12942.598349920547, "speed/FLOPS": 203281292079369.12, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.871903896331787, "optim/lr": 0.0025272000000000003, "optim/total_tokens": 1104412672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444881, "loss/out": 5.336308479309082, "created_at": "2025-01-14T18:56:44.475416+00:00"} {"global_step": 4214, "acc_step": 0, "speed/wps": 12942.565951772533, "speed/FLOPS": 203280783221934.84, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.308560848236084, "optim/lr": 0.0025278, "optim/total_tokens": 1104674816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 339062, "loss/out": 5.319110870361328, "created_at": "2025-01-14T18:56:49.546950+00:00"} {"global_step": 4215, "acc_step": 0, "speed/wps": 12933.81252892338, "speed/FLOPS": 203143298687625.4, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.747430801391602, "optim/lr": 0.0025284, "optim/total_tokens": 1104936960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 5.236179351806641, "created_at": "2025-01-14T18:56:54.620620+00:00"} {"global_step": 4216, "acc_step": 0, "speed/wps": 12932.908596059513, "speed/FLOPS": 203129101179864.47, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.887373924255371, "optim/lr": 0.002529, "optim/total_tokens": 1105199104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464535, "loss/out": 5.3793721199035645, "created_at": "2025-01-14T18:56:59.691719+00:00"} {"global_step": 4217, "acc_step": 0, "speed/wps": 12945.250653877347, "speed/FLOPS": 203322950157658.25, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.782258987426758, "optim/lr": 0.0025296, "optim/total_tokens": 1105461248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419618, "loss/out": 5.411484718322754, "created_at": "2025-01-14T18:57:04.759157+00:00"} {"global_step": 4218, "acc_step": 0, "speed/wps": 12941.105582673128, "speed/FLOPS": 203257846118471.28, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.79442024230957, "optim/lr": 0.0025302000000000002, "optim/total_tokens": 1105723392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 384005, "loss/out": 5.267304420471191, "created_at": "2025-01-14T18:57:09.827832+00:00"} {"global_step": 4219, "acc_step": 0, "speed/wps": 12941.960962222061, "speed/FLOPS": 203271281029701.94, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0009, "optim/grad_norm": 10.611288070678711, "optim/lr": 0.0025308, "optim/total_tokens": 1105985536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 5.3611979484558105, "created_at": "2025-01-14T18:57:14.901214+00:00"} {"global_step": 4220, "acc_step": 0, "speed/wps": 12944.691599732068, "speed/FLOPS": 203314169444085.75, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 20.299991607666016, "optim/lr": 0.0025314, "optim/total_tokens": 1106247680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439820, "loss/out": 5.439289569854736, "created_at": "2025-01-14T18:57:19.970851+00:00"} {"global_step": 4221, "acc_step": 0, "speed/wps": 12945.553835995363, "speed/FLOPS": 203327712049437.6, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.885597229003906, "optim/lr": 0.002532, "optim/total_tokens": 1106509824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469069, "loss/out": 5.345252513885498, "created_at": "2025-01-14T18:57:25.037231+00:00"} {"global_step": 4222, "acc_step": 0, "speed/wps": 12947.356630002509, "speed/FLOPS": 203356027406618.38, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.720163345336914, "optim/lr": 0.0025326, "optim/total_tokens": 1106771968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437717, "loss/out": 5.391109943389893, "created_at": "2025-01-14T18:57:30.103786+00:00"} {"global_step": 4223, "acc_step": 0, "speed/wps": 12949.110095458334, "speed/FLOPS": 203383567991116.16, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.453479766845703, "optim/lr": 0.0025332, "optim/total_tokens": 1107034112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353744, "loss/out": 5.331836700439453, "created_at": "2025-01-14T18:57:35.169744+00:00"} {"global_step": 4224, "acc_step": 0, "speed/wps": 12946.588475350829, "speed/FLOPS": 203343962482258.66, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 18.52442741394043, "optim/lr": 0.0025338, "optim/total_tokens": 1107296256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453131, "loss/out": 5.365963935852051, "created_at": "2025-01-14T18:57:40.240588+00:00"} {"global_step": 4225, "acc_step": 0, "speed/wps": 12945.26236108696, "speed/FLOPS": 203323134035472.62, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0009, "optim/grad_norm": 11.242268562316895, "optim/lr": 0.0025344, "optim/total_tokens": 1107558400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434807, "loss/out": 5.371485233306885, "created_at": "2025-01-14T18:57:45.309171+00:00"} {"global_step": 4226, "acc_step": 0, "speed/wps": 12942.855048490039, "speed/FLOPS": 203285323883143.8, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.658621788024902, "optim/lr": 0.002535, "optim/total_tokens": 1107820544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429739, "loss/out": 5.335018157958984, "created_at": "2025-01-14T18:57:50.379503+00:00"} {"global_step": 4227, "acc_step": 0, "speed/wps": 12941.567104338103, "speed/FLOPS": 203265094950416.94, "speed/curr_iter_time": 1.2786, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.594177484512329, "optim/lr": 0.0025356, "optim/total_tokens": 1108082688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354800, "loss/out": 5.424726486206055, "created_at": "2025-01-14T18:57:55.448498+00:00"} {"global_step": 4228, "acc_step": 0, "speed/wps": 12948.570798563625, "speed/FLOPS": 203375097592313.16, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 39.18235397338867, "optim/lr": 0.0025362, "optim/total_tokens": 1108344832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 5.403896808624268, "created_at": "2025-01-14T18:58:00.515966+00:00"} {"global_step": 4229, "acc_step": 0, "speed/wps": 12947.21537286403, "speed/FLOPS": 203353808769150.9, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.107415199279785, "optim/lr": 0.0025368, "optim/total_tokens": 1108606976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428688, "loss/out": 5.706811904907227, "created_at": "2025-01-14T18:58:05.583400+00:00"} {"global_step": 4230, "acc_step": 0, "speed/wps": 12951.900964167298, "speed/FLOPS": 203427402419245.94, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.241147994995117, "optim/lr": 0.0025374, "optim/total_tokens": 1108869120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 5.4412665367126465, "created_at": "2025-01-14T18:58:10.650641+00:00"} {"global_step": 4231, "acc_step": 0, "speed/wps": 12945.090153407455, "speed/FLOPS": 203320429277227.84, "speed/curr_iter_time": 1.2771, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.362811088562012, "optim/lr": 0.002538, "optim/total_tokens": 1109131264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298593, "loss/out": 5.57082462310791, "created_at": "2025-01-14T18:58:15.722598+00:00"} {"global_step": 4232, "acc_step": 0, "speed/wps": 12943.322589655476, "speed/FLOPS": 203292667259615.88, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.517906188964844, "optim/lr": 0.0025386, "optim/total_tokens": 1109393408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441450, "loss/out": 5.540152549743652, "created_at": "2025-01-14T18:58:20.796189+00:00"} {"global_step": 4233, "acc_step": 0, "speed/wps": 12945.031773917397, "speed/FLOPS": 203319512347114.66, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0006, "optim/grad_norm": 20.242172241210938, "optim/lr": 0.0025392, "optim/total_tokens": 1109655552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440399, "loss/out": 5.437625408172607, "created_at": "2025-01-14T18:58:25.862428+00:00"} {"global_step": 4234, "acc_step": 0, "speed/wps": 12942.714227592609, "speed/FLOPS": 203283112097441.5, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.8425889015197754, "optim/lr": 0.0025398, "optim/total_tokens": 1109917696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461622, "loss/out": 5.576164245605469, "created_at": "2025-01-14T18:58:30.931151+00:00"} {"global_step": 4235, "acc_step": 0, "speed/wps": 12948.931428072945, "speed/FLOPS": 203380761774313.9, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.465795516967773, "optim/lr": 0.0025404, "optim/total_tokens": 1110179840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442778, "loss/out": 5.576667785644531, "created_at": "2025-01-14T18:58:35.997076+00:00"} {"global_step": 4236, "acc_step": 0, "speed/wps": 12944.016205077643, "speed/FLOPS": 203303561442948.5, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0006, "optim/grad_norm": 20.179080963134766, "optim/lr": 0.002541, "optim/total_tokens": 1110441984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287957, "loss/out": 5.494193077087402, "created_at": "2025-01-14T18:58:41.066887+00:00"} {"global_step": 4237, "acc_step": 0, "speed/wps": 12948.757814639173, "speed/FLOPS": 203378034936766.66, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.340997219085693, "optim/lr": 0.0025415999999999998, "optim/total_tokens": 1110704128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447839, "loss/out": 5.581316947937012, "created_at": "2025-01-14T18:58:46.132969+00:00"} {"global_step": 4238, "acc_step": 0, "speed/wps": 12942.790361671427, "speed/FLOPS": 203284307887768.5, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.4870405197143555, "optim/lr": 0.0025422, "optim/total_tokens": 1110966272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452325, "loss/out": 5.562206268310547, "created_at": "2025-01-14T18:58:51.202898+00:00"} {"global_step": 4239, "acc_step": 0, "speed/wps": 12944.422117753224, "speed/FLOPS": 203309936859300.97, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0012, "optim/grad_norm": 4.169341087341309, "optim/lr": 0.0025428, "optim/total_tokens": 1111228416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 5.486944198608398, "created_at": "2025-01-14T18:58:56.291996+00:00"} {"global_step": 4240, "acc_step": 0, "speed/wps": 12955.589268922598, "speed/FLOPS": 203485332313689.75, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.188906192779541, "optim/lr": 0.0025434, "optim/total_tokens": 1111490560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463485, "loss/out": 5.460180282592773, "created_at": "2025-01-14T18:59:01.354251+00:00"} {"global_step": 4241, "acc_step": 0, "speed/wps": 12945.740807801827, "speed/FLOPS": 203330648698584.28, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0009, "optim/grad_norm": 9.30300521850586, "optim/lr": 0.002544, "optim/total_tokens": 1111752704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 5.439235210418701, "created_at": "2025-01-14T18:59:06.423197+00:00"} {"global_step": 4242, "acc_step": 0, "speed/wps": 12942.981072162866, "speed/FLOPS": 203287303258100.22, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.918271541595459, "optim/lr": 0.0025445999999999997, "optim/total_tokens": 1112014848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 5.396125793457031, "created_at": "2025-01-14T18:59:11.491854+00:00"} {"global_step": 4243, "acc_step": 0, "speed/wps": 12936.440820617834, "speed/FLOPS": 203184579620338.1, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0011, "optim/grad_norm": 8.734699249267578, "optim/lr": 0.0025452, "optim/total_tokens": 1112276992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332731, "loss/out": 5.392729759216309, "created_at": "2025-01-14T18:59:16.562490+00:00"} {"global_step": 4244, "acc_step": 0, "speed/wps": 12942.724649495703, "speed/FLOPS": 203283275787750.78, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.284034729003906, "optim/lr": 0.0025458, "optim/total_tokens": 1112539136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462417, "loss/out": 5.337693214416504, "created_at": "2025-01-14T18:59:21.630703+00:00"} {"global_step": 4245, "acc_step": 0, "speed/wps": 12947.975146587743, "speed/FLOPS": 203365742059520.34, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.41811466217041, "optim/lr": 0.0025464, "optim/total_tokens": 1112801280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480663, "loss/out": 5.406623840332031, "created_at": "2025-01-14T18:59:26.701991+00:00"} {"global_step": 4246, "acc_step": 0, "speed/wps": 12946.850846636677, "speed/FLOPS": 203348083383838.8, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 23.637351989746094, "optim/lr": 0.002547, "optim/total_tokens": 1113063424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 5.356875419616699, "created_at": "2025-01-14T18:59:31.773251+00:00"} {"global_step": 4247, "acc_step": 0, "speed/wps": 12945.402727797282, "speed/FLOPS": 203325338687542.3, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.541885375976562, "optim/lr": 0.0025476, "optim/total_tokens": 1113325568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428688, "loss/out": 5.463922500610352, "created_at": "2025-01-14T18:59:36.839606+00:00"} {"global_step": 4248, "acc_step": 0, "speed/wps": 12946.464253588116, "speed/FLOPS": 203342011408776.62, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0008, "optim/grad_norm": 30.058137893676758, "optim/lr": 0.0025482, "optim/total_tokens": 1113587712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326344, "loss/out": 5.5001726150512695, "created_at": "2025-01-14T18:59:41.908599+00:00"} {"global_step": 4249, "acc_step": 0, "speed/wps": 12943.949018538553, "speed/FLOPS": 203302506186028.1, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.319458961486816, "optim/lr": 0.0025488, "optim/total_tokens": 1113849856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460556, "loss/out": 5.475644111633301, "created_at": "2025-01-14T18:59:46.975869+00:00"} {"global_step": 4250, "acc_step": 0, "speed/wps": 12944.07595877814, "speed/FLOPS": 203304499956924.97, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0006, "optim/grad_norm": 17.770469665527344, "optim/lr": 0.0025494, "optim/total_tokens": 1114112000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 5.373712062835693, "created_at": "2025-01-14T18:59:52.043336+00:00"} {"global_step": 4251, "acc_step": 0, "speed/wps": 12944.521764605262, "speed/FLOPS": 203311501950041.34, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.367514610290527, "optim/lr": 0.00255, "optim/total_tokens": 1114374144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469590, "loss/out": 5.384546756744385, "created_at": "2025-01-14T18:59:57.110653+00:00"} {"global_step": 4252, "acc_step": 0, "speed/wps": 12944.752897965422, "speed/FLOPS": 203315132217073.16, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 216.8415069580078, "optim/lr": 0.0025506, "optim/total_tokens": 1114636288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438769, "loss/out": 5.30786657333374, "created_at": "2025-01-14T19:00:02.179467+00:00"} {"global_step": 4253, "acc_step": 0, "speed/wps": 12941.282777528606, "speed/FLOPS": 203260629207167.8, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.632325172424316, "optim/lr": 0.0025512, "optim/total_tokens": 1114898432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424656, "loss/out": 5.689328193664551, "created_at": "2025-01-14T19:00:07.255210+00:00"} {"global_step": 4254, "acc_step": 0, "speed/wps": 12939.06869374042, "speed/FLOPS": 203225853978804.4, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.057591915130615, "optim/lr": 0.0025518000000000003, "optim/total_tokens": 1115160576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463222, "loss/out": 5.40920877456665, "created_at": "2025-01-14T19:00:12.331780+00:00"} {"global_step": 4255, "acc_step": 0, "speed/wps": 12930.734415224797, "speed/FLOPS": 203094952682217.72, "speed/curr_iter_time": 1.2785, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.984951496124268, "optim/lr": 0.0025524000000000002, "optim/total_tokens": 1115422720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450746, "loss/out": 5.3448991775512695, "created_at": "2025-01-14T19:00:17.408420+00:00"} {"global_step": 4256, "acc_step": 0, "speed/wps": 12940.305454335701, "speed/FLOPS": 203245279003440.44, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.990516662597656, "optim/lr": 0.002553, "optim/total_tokens": 1115684864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363445, "loss/out": 5.254671573638916, "created_at": "2025-01-14T19:00:22.482332+00:00"} {"global_step": 4257, "acc_step": 0, "speed/wps": 12946.085009900824, "speed/FLOPS": 203336054865532.28, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 15.282024383544922, "optim/lr": 0.0025536, "optim/total_tokens": 1115947008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501085, "loss/out": 5.4524970054626465, "created_at": "2025-01-14T19:00:27.548437+00:00"} {"global_step": 4258, "acc_step": 0, "speed/wps": 12941.55604643523, "speed/FLOPS": 203264921270857.16, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0006, "optim/grad_norm": 23.386249542236328, "optim/lr": 0.0025542000000000004, "optim/total_tokens": 1116209152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435567, "loss/out": 5.302866458892822, "created_at": "2025-01-14T19:00:32.615931+00:00"} {"global_step": 4259, "acc_step": 0, "speed/wps": 12943.460250198536, "speed/FLOPS": 203294829407610.06, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.805872917175293, "optim/lr": 0.0025548000000000003, "optim/total_tokens": 1116471296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 5.281503677368164, "created_at": "2025-01-14T19:00:37.682994+00:00"} {"global_step": 4260, "acc_step": 0, "speed/wps": 12940.638225454893, "speed/FLOPS": 203250505631142.78, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.313382148742676, "optim/lr": 0.0025554, "optim/total_tokens": 1116733440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338354, "loss/out": 5.356696128845215, "created_at": "2025-01-14T19:00:42.752590+00:00"} {"global_step": 4261, "acc_step": 0, "speed/wps": 12952.825793359249, "speed/FLOPS": 203441928132554.0, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 22.03824234008789, "optim/lr": 0.002556, "optim/total_tokens": 1116995584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298880, "loss/out": 5.286186218261719, "created_at": "2025-01-14T19:00:47.816915+00:00"} {"global_step": 4262, "acc_step": 0, "speed/wps": 12941.343688860261, "speed/FLOPS": 203261585903332.38, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 35.35481262207031, "optim/lr": 0.0025566, "optim/total_tokens": 1117257728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449411, "loss/out": 5.4336748123168945, "created_at": "2025-01-14T19:00:52.885348+00:00"} {"global_step": 4263, "acc_step": 0, "speed/wps": 12946.098307253345, "speed/FLOPS": 203336263718726.12, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0006, "optim/grad_norm": 24.534700393676758, "optim/lr": 0.0025572000000000004, "optim/total_tokens": 1117519872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 5.423933982849121, "created_at": "2025-01-14T19:00:57.956922+00:00"} {"global_step": 4264, "acc_step": 0, "speed/wps": 12943.58211205348, "speed/FLOPS": 203296743415497.7, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0009, "optim/grad_norm": 16.42714500427246, "optim/lr": 0.0025578000000000003, "optim/total_tokens": 1117782016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 5.408156871795654, "created_at": "2025-01-14T19:01:03.024615+00:00"} {"global_step": 4265, "acc_step": 0, "speed/wps": 12946.998553667318, "speed/FLOPS": 203350403325725.75, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.568927764892578, "optim/lr": 0.0025584, "optim/total_tokens": 1118044160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334076, "loss/out": 5.3333353996276855, "created_at": "2025-01-14T19:01:08.094289+00:00"} {"global_step": 4266, "acc_step": 0, "speed/wps": 12940.58945739318, "speed/FLOPS": 203249739661719.22, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.519815444946289, "optim/lr": 0.002559, "optim/total_tokens": 1118306304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455517, "loss/out": 5.373606204986572, "created_at": "2025-01-14T19:01:13.166166+00:00"} {"global_step": 4267, "acc_step": 0, "speed/wps": 12944.260944932119, "speed/FLOPS": 203307405418670.84, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.943840503692627, "optim/lr": 0.0025596, "optim/total_tokens": 1118568448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 5.3563408851623535, "created_at": "2025-01-14T19:01:18.235629+00:00"} {"global_step": 4268, "acc_step": 0, "speed/wps": 12942.068737912525, "speed/FLOPS": 203272973795020.0, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.595102310180664, "optim/lr": 0.0025602000000000003, "optim/total_tokens": 1118830592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 5.3177361488342285, "created_at": "2025-01-14T19:01:23.304368+00:00"} {"global_step": 4269, "acc_step": 0, "speed/wps": 12942.138901734763, "speed/FLOPS": 203274075814263.44, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0007, "optim/grad_norm": 38.68895721435547, "optim/lr": 0.0025608000000000002, "optim/total_tokens": 1119092736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440390, "loss/out": 5.441223621368408, "created_at": "2025-01-14T19:01:28.373424+00:00"} {"global_step": 4270, "acc_step": 0, "speed/wps": 12939.497159129773, "speed/FLOPS": 203232583616517.3, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.470314025878906, "optim/lr": 0.0025614, "optim/total_tokens": 1119354880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429217, "loss/out": 5.529829025268555, "created_at": "2025-01-14T19:01:33.441794+00:00"} {"global_step": 4271, "acc_step": 0, "speed/wps": 12947.504960819446, "speed/FLOPS": 203358357145926.28, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.699738502502441, "optim/lr": 0.002562, "optim/total_tokens": 1119617024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 5.44329309463501, "created_at": "2025-01-14T19:01:38.514161+00:00"} {"global_step": 4272, "acc_step": 0, "speed/wps": 12951.840108366736, "speed/FLOPS": 203426446595273.47, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.381391525268555, "optim/lr": 0.0025626, "optim/total_tokens": 1119879168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415127, "loss/out": 5.401036739349365, "created_at": "2025-01-14T19:01:43.579356+00:00"} {"global_step": 4273, "acc_step": 0, "speed/wps": 12943.800792995165, "speed/FLOPS": 203300178100185.97, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.02240753173828, "optim/lr": 0.0025632000000000003, "optim/total_tokens": 1120141312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466683, "loss/out": 5.457921504974365, "created_at": "2025-01-14T19:01:48.647072+00:00"} {"global_step": 4274, "acc_step": 0, "speed/wps": 12943.4459869047, "speed/FLOPS": 203294605382980.56, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.949646949768066, "optim/lr": 0.0025638, "optim/total_tokens": 1120403456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287670, "loss/out": 5.529659748077393, "created_at": "2025-01-14T19:01:53.722777+00:00"} {"global_step": 4275, "acc_step": 0, "speed/wps": 12946.568567123757, "speed/FLOPS": 203343649796195.72, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 58.18553161621094, "optim/lr": 0.0025644, "optim/total_tokens": 1120665600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461385, "loss/out": 5.53738260269165, "created_at": "2025-01-14T19:01:58.790012+00:00"} {"global_step": 4276, "acc_step": 0, "speed/wps": 12949.668635292317, "speed/FLOPS": 203392340626723.0, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.055704116821289, "optim/lr": 0.002565, "optim/total_tokens": 1120927744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 5.716134071350098, "created_at": "2025-01-14T19:02:03.856365+00:00"} {"global_step": 4277, "acc_step": 0, "speed/wps": 12945.024215461039, "speed/FLOPS": 203319393631171.16, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.191078186035156, "optim/lr": 0.0025656, "optim/total_tokens": 1121189888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435329, "loss/out": 5.53756856918335, "created_at": "2025-01-14T19:02:08.925315+00:00"} {"global_step": 4278, "acc_step": 0, "speed/wps": 12945.117925226506, "speed/FLOPS": 203320865471808.53, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 22.037626266479492, "optim/lr": 0.0025662000000000003, "optim/total_tokens": 1121452032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445170, "loss/out": 5.631679534912109, "created_at": "2025-01-14T19:02:13.992532+00:00"} {"global_step": 4279, "acc_step": 0, "speed/wps": 12938.069750875216, "speed/FLOPS": 203210164208415.03, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0006, "optim/grad_norm": 15.646329879760742, "optim/lr": 0.0025668, "optim/total_tokens": 1121714176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 344264, "loss/out": 5.4482316970825195, "created_at": "2025-01-14T19:02:19.065361+00:00"} {"global_step": 4280, "acc_step": 0, "speed/wps": 12938.955912351023, "speed/FLOPS": 203224082592106.03, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0012, "optim/grad_norm": 17.568819046020508, "optim/lr": 0.0025674, "optim/total_tokens": 1121976320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 5.474756240844727, "created_at": "2025-01-14T19:02:24.136126+00:00"} {"global_step": 4281, "acc_step": 0, "speed/wps": 12941.42102539582, "speed/FLOPS": 203262800579894.88, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0006, "optim/grad_norm": 66.00908660888672, "optim/lr": 0.002568, "optim/total_tokens": 1122238464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451797, "loss/out": 5.511550426483154, "created_at": "2025-01-14T19:02:29.210691+00:00"} {"global_step": 4282, "acc_step": 0, "speed/wps": 12947.178075319187, "speed/FLOPS": 203353222959956.97, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.939093589782715, "optim/lr": 0.0025686, "optim/total_tokens": 1122500608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441971, "loss/out": 5.561670303344727, "created_at": "2025-01-14T19:02:34.276158+00:00"} {"global_step": 4283, "acc_step": 0, "speed/wps": 12940.97538075942, "speed/FLOPS": 203255801118500.3, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.476809024810791, "optim/lr": 0.0025692, "optim/total_tokens": 1122762752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363872, "loss/out": 5.572010517120361, "created_at": "2025-01-14T19:02:39.344133+00:00"} {"global_step": 4284, "acc_step": 0, "speed/wps": 12938.981818778828, "speed/FLOPS": 203224489488153.88, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.415478706359863, "optim/lr": 0.0025698, "optim/total_tokens": 1123024896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 5.628111362457275, "created_at": "2025-01-14T19:02:44.413858+00:00"} {"global_step": 4285, "acc_step": 0, "speed/wps": 12939.435154041677, "speed/FLOPS": 203231609741404.84, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 15.906102180480957, "optim/lr": 0.0025704, "optim/total_tokens": 1123287040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429739, "loss/out": 5.5556135177612305, "created_at": "2025-01-14T19:02:49.483359+00:00"} {"global_step": 4286, "acc_step": 0, "speed/wps": 12940.849705519338, "speed/FLOPS": 203253827216158.94, "speed/curr_iter_time": 1.2774, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.112419128417969, "optim/lr": 0.002571, "optim/total_tokens": 1123549184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 5.5112504959106445, "created_at": "2025-01-14T19:02:54.582587+00:00"} {"global_step": 4287, "acc_step": 0, "speed/wps": 12952.861609452093, "speed/FLOPS": 203442490673508.6, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 59.41968536376953, "optim/lr": 0.0025716, "optim/total_tokens": 1123811328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352926, "loss/out": 5.536801338195801, "created_at": "2025-01-14T19:02:59.651150+00:00"} {"global_step": 4288, "acc_step": 0, "speed/wps": 12948.373785738644, "speed/FLOPS": 203372003235172.84, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.069376945495605, "optim/lr": 0.0025722, "optim/total_tokens": 1124073472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463485, "loss/out": 5.65231466293335, "created_at": "2025-01-14T19:03:04.719680+00:00"} {"global_step": 4289, "acc_step": 0, "speed/wps": 12932.718267845732, "speed/FLOPS": 203126111813728.7, "speed/curr_iter_time": 1.2783, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.717024803161621, "optim/lr": 0.0025728, "optim/total_tokens": 1124335616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447791, "loss/out": 5.60256290435791, "created_at": "2025-01-14T19:03:09.795080+00:00"} {"global_step": 4290, "acc_step": 0, "speed/wps": 12941.38886638442, "speed/FLOPS": 203262295478429.75, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0009, "optim/grad_norm": 20.555221557617188, "optim/lr": 0.0025734, "optim/total_tokens": 1124597760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 5.582995891571045, "created_at": "2025-01-14T19:03:14.876595+00:00"} {"global_step": 4291, "acc_step": 0, "speed/wps": 12939.527733614914, "speed/FLOPS": 203233063830820.2, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0006, "optim/grad_norm": 17.6827449798584, "optim/lr": 0.002574, "optim/total_tokens": 1124859904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 5.483090877532959, "created_at": "2025-01-14T19:03:19.948617+00:00"} {"global_step": 4292, "acc_step": 0, "speed/wps": 12936.767830087547, "speed/FLOPS": 203189715753418.03, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.001, "optim/grad_norm": 23.8356990814209, "optim/lr": 0.0025746, "optim/total_tokens": 1125122048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440921, "loss/out": 5.593237400054932, "created_at": "2025-01-14T19:03:25.019919+00:00"} {"global_step": 4293, "acc_step": 0, "speed/wps": 12926.479325702525, "speed/FLOPS": 203028120654164.7, "speed/curr_iter_time": 1.2774, "speed/data_load_time": 0.0007, "optim/grad_norm": 34.37356185913086, "optim/lr": 0.0025752, "optim/total_tokens": 1125384192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 5.609842777252197, "created_at": "2025-01-14T19:03:30.093862+00:00"} {"global_step": 4294, "acc_step": 0, "speed/wps": 12941.62450053661, "speed/FLOPS": 203265996436585.12, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0006, "optim/grad_norm": 34.169132232666016, "optim/lr": 0.0025758, "optim/total_tokens": 1125646336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327253, "loss/out": 5.573139190673828, "created_at": "2025-01-14T19:03:35.165455+00:00"} {"global_step": 4295, "acc_step": 0, "speed/wps": 12941.373481904995, "speed/FLOPS": 203262053844039.88, "speed/curr_iter_time": 1.2778, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.19049072265625, "optim/lr": 0.0025764, "optim/total_tokens": 1125908480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467734, "loss/out": 5.608678817749023, "created_at": "2025-01-14T19:03:40.236545+00:00"} {"global_step": 4296, "acc_step": 0, "speed/wps": 12948.317937897542, "speed/FLOPS": 203371126068085.34, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 25.706764221191406, "optim/lr": 0.002577, "optim/total_tokens": 1126170624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433196, "loss/out": 5.456090927124023, "created_at": "2025-01-14T19:03:45.311494+00:00"} {"global_step": 4297, "acc_step": 0, "speed/wps": 12940.09980768495, "speed/FLOPS": 203242049040202.28, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 203.3535614013672, "optim/lr": 0.0025775999999999998, "optim/total_tokens": 1126432768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428688, "loss/out": 5.5300068855285645, "created_at": "2025-01-14T19:03:50.381208+00:00"} {"global_step": 4298, "acc_step": 0, "speed/wps": 12938.684090255325, "speed/FLOPS": 203219813252570.4, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0006, "optim/grad_norm": 433.0382080078125, "optim/lr": 0.0025782, "optim/total_tokens": 1126694912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416669, "loss/out": 5.605433464050293, "created_at": "2025-01-14T19:03:55.452411+00:00"} {"global_step": 4299, "acc_step": 0, "speed/wps": 12943.615934273543, "speed/FLOPS": 203297274639940.38, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.001, "optim/grad_norm": 42.61789321899414, "optim/lr": 0.0025788, "optim/total_tokens": 1126957056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 5.7648444175720215, "created_at": "2025-01-14T19:04:00.519599+00:00"} {"global_step": 4300, "acc_step": 0, "speed/wps": 12941.60127542428, "speed/FLOPS": 203265631654281.34, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0007, "optim/grad_norm": 19.28985595703125, "optim/lr": 0.0025794, "optim/total_tokens": 1127219200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448597, "loss/out": 5.726552963256836, "created_at": "2025-01-14T19:04:05.587604+00:00"} {"global_step": 4301, "acc_step": 0, "speed/wps": 12943.27284192869, "speed/FLOPS": 203291885903206.2, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 53.9228515625, "optim/lr": 0.00258, "optim/total_tokens": 1127481344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457902, "loss/out": 5.625667572021484, "created_at": "2025-01-14T19:04:10.657162+00:00"} {"global_step": 4302, "acc_step": 0, "speed/wps": 12942.502484246874, "speed/FLOPS": 203279786377230.88, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0008, "optim/grad_norm": 34.32269287109375, "optim/lr": 0.0025805999999999997, "optim/total_tokens": 1127743488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435866, "loss/out": 5.63852596282959, "created_at": "2025-01-14T19:04:15.732235+00:00"} {"global_step": 4303, "acc_step": 0, "speed/wps": 12943.013434299193, "speed/FLOPS": 203287811549921.5, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 18.591106414794922, "optim/lr": 0.0025812, "optim/total_tokens": 1128005632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453144, "loss/out": 5.477165222167969, "created_at": "2025-01-14T19:04:20.804002+00:00"} {"global_step": 4304, "acc_step": 0, "speed/wps": 12942.302385416031, "speed/FLOPS": 203276643550127.88, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.652724266052246, "optim/lr": 0.0025818, "optim/total_tokens": 1128267776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 5.588406085968018, "created_at": "2025-01-14T19:04:25.871368+00:00"} {"global_step": 4305, "acc_step": 0, "speed/wps": 12941.84961803016, "speed/FLOPS": 203269532216165.2, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0008, "optim/grad_norm": 18.4403076171875, "optim/lr": 0.0025824, "optim/total_tokens": 1128529920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 5.526180267333984, "created_at": "2025-01-14T19:04:30.945314+00:00"} {"global_step": 4306, "acc_step": 0, "speed/wps": 12939.778683027898, "speed/FLOPS": 203237005336191.75, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.864278793334961, "optim/lr": 0.0025830000000000002, "optim/total_tokens": 1128792064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306094, "loss/out": 5.723276138305664, "created_at": "2025-01-14T19:04:36.014640+00:00"} {"global_step": 4307, "acc_step": 0, "speed/wps": 12943.872754161785, "speed/FLOPS": 203301308349192.72, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 60.46977615356445, "optim/lr": 0.0025836, "optim/total_tokens": 1129054208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451275, "loss/out": 5.645127773284912, "created_at": "2025-01-14T19:04:41.084340+00:00"} {"global_step": 4308, "acc_step": 0, "speed/wps": 12947.078878283455, "speed/FLOPS": 203351664934199.78, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0008, "optim/grad_norm": 26.653474807739258, "optim/lr": 0.0025842, "optim/total_tokens": 1129316352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 5.614219665527344, "created_at": "2025-01-14T19:04:46.152688+00:00"} {"global_step": 4309, "acc_step": 0, "speed/wps": 12941.340759067536, "speed/FLOPS": 203261539886911.7, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0006, "optim/grad_norm": 27.366817474365234, "optim/lr": 0.0025848, "optim/total_tokens": 1129578496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 5.76848030090332, "created_at": "2025-01-14T19:04:51.221039+00:00"} {"global_step": 4310, "acc_step": 0, "speed/wps": 12943.053895529423, "speed/FLOPS": 203288447049141.8, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 19.650739669799805, "optim/lr": 0.0025854000000000003, "optim/total_tokens": 1129840640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426544, "loss/out": 5.572261810302734, "created_at": "2025-01-14T19:04:56.291894+00:00"} {"global_step": 4311, "acc_step": 0, "speed/wps": 12939.104926285345, "speed/FLOPS": 203226423060711.16, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0007, "optim/grad_norm": 15.24917221069336, "optim/lr": 0.002586, "optim/total_tokens": 1130102784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 409401, "loss/out": 5.367412567138672, "created_at": "2025-01-14T19:05:01.361827+00:00"} {"global_step": 4312, "acc_step": 0, "speed/wps": 12941.509254054972, "speed/FLOPS": 203264186332225.75, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0006, "optim/grad_norm": 22.870759963989258, "optim/lr": 0.0025866, "optim/total_tokens": 1130364928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480954, "loss/out": 5.432540416717529, "created_at": "2025-01-14T19:05:06.430382+00:00"} {"global_step": 4313, "acc_step": 0, "speed/wps": 12942.68719690115, "speed/FLOPS": 203282687543288.06, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0008, "optim/grad_norm": 17.353797912597656, "optim/lr": 0.0025872000000000004, "optim/total_tokens": 1130627072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435338, "loss/out": 5.563551902770996, "created_at": "2025-01-14T19:05:11.501290+00:00"} {"global_step": 4314, "acc_step": 0, "speed/wps": 12936.457822448232, "speed/FLOPS": 203184846657447.47, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0007, "optim/grad_norm": 20.3869571685791, "optim/lr": 0.0025878000000000003, "optim/total_tokens": 1130889216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 5.5161027908325195, "created_at": "2025-01-14T19:05:16.576064+00:00"} {"global_step": 4315, "acc_step": 0, "speed/wps": 12936.649754309894, "speed/FLOPS": 203187861211076.88, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.19802188873291, "optim/lr": 0.0025884000000000002, "optim/total_tokens": 1131151360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300465, "loss/out": 5.590615749359131, "created_at": "2025-01-14T19:05:21.648110+00:00"} {"global_step": 4316, "acc_step": 0, "speed/wps": 12940.702395769682, "speed/FLOPS": 203251513514115.78, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.685040473937988, "optim/lr": 0.002589, "optim/total_tokens": 1131413504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449411, "loss/out": 5.686498641967773, "created_at": "2025-01-14T19:05:26.718674+00:00"} {"global_step": 4317, "acc_step": 0, "speed/wps": 12944.818773615614, "speed/FLOPS": 203316166884682.06, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.756749153137207, "optim/lr": 0.0025896, "optim/total_tokens": 1131675648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 5.660069465637207, "created_at": "2025-01-14T19:05:31.792031+00:00"} {"global_step": 4318, "acc_step": 0, "speed/wps": 12944.831939777194, "speed/FLOPS": 203316373677341.88, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0008, "optim/grad_norm": 8.238286018371582, "optim/lr": 0.0025902, "optim/total_tokens": 1131937792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 5.7974677085876465, "created_at": "2025-01-14T19:05:36.858365+00:00"} {"global_step": 4319, "acc_step": 0, "speed/wps": 12938.759206271017, "speed/FLOPS": 203220993052817.06, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 21.535987854003906, "optim/lr": 0.0025908000000000003, "optim/total_tokens": 1132199936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331010, "loss/out": 5.886399745941162, "created_at": "2025-01-14T19:05:41.929516+00:00"} {"global_step": 4320, "acc_step": 0, "speed/wps": 12939.756340169217, "speed/FLOPS": 203236654410893.84, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.345786094665527, "optim/lr": 0.0025914, "optim/total_tokens": 1132462080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 405668, "loss/out": 5.941833019256592, "created_at": "2025-01-14T19:05:47.004026+00:00"} {"global_step": 4321, "acc_step": 0, "speed/wps": 12937.315403851328, "speed/FLOPS": 203198316151823.4, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0007, "optim/grad_norm": 15.094377517700195, "optim/lr": 0.002592, "optim/total_tokens": 1132724224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 5.860189437866211, "created_at": "2025-01-14T19:05:52.073513+00:00"} {"global_step": 4322, "acc_step": 0, "speed/wps": 12946.45460692189, "speed/FLOPS": 203341859894627.66, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0008, "optim/grad_norm": 12.445295333862305, "optim/lr": 0.0025926, "optim/total_tokens": 1132986368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 5.876277446746826, "created_at": "2025-01-14T19:05:57.140309+00:00"} {"global_step": 4323, "acc_step": 0, "speed/wps": 12941.920824394434, "speed/FLOPS": 203270650609963.9, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0008, "optim/grad_norm": 10.374900817871094, "optim/lr": 0.0025932, "optim/total_tokens": 1133248512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328221, "loss/out": 5.856917381286621, "created_at": "2025-01-14T19:06:02.214481+00:00"} {"global_step": 4324, "acc_step": 0, "speed/wps": 12944.562761510542, "speed/FLOPS": 203312145862774.4, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 28.555606842041016, "optim/lr": 0.0025938000000000003, "optim/total_tokens": 1133510656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 5.79676628112793, "created_at": "2025-01-14T19:06:07.286283+00:00"} {"global_step": 4325, "acc_step": 0, "speed/wps": 12944.103047653318, "speed/FLOPS": 203304925424933.66, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 75.84349822998047, "optim/lr": 0.0025944, "optim/total_tokens": 1133772800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 5.865175247192383, "created_at": "2025-01-14T19:06:12.360921+00:00"} {"global_step": 4326, "acc_step": 0, "speed/wps": 12935.094281560861, "speed/FLOPS": 203163430374109.78, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0008, "optim/grad_norm": 45.931640625, "optim/lr": 0.002595, "optim/total_tokens": 1134034944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 5.877960205078125, "created_at": "2025-01-14T19:06:17.431420+00:00"} {"global_step": 4327, "acc_step": 0, "speed/wps": 12938.164987107455, "speed/FLOPS": 203211660024310.62, "speed/curr_iter_time": 1.2771, "speed/data_load_time": 0.0008, "optim/grad_norm": 16.359195709228516, "optim/lr": 0.0025956, "optim/total_tokens": 1134297088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439302, "loss/out": 5.955227851867676, "created_at": "2025-01-14T19:06:22.503441+00:00"} {"global_step": 4328, "acc_step": 0, "speed/wps": 12948.706280971997, "speed/FLOPS": 203377225529708.97, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 50.130741119384766, "optim/lr": 0.0025962, "optim/total_tokens": 1134559232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307679, "loss/out": 5.910369873046875, "created_at": "2025-01-14T19:06:27.568367+00:00"} {"global_step": 4329, "acc_step": 0, "speed/wps": 12937.256143061719, "speed/FLOPS": 203197385379689.75, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.07441520690918, "optim/lr": 0.0025968000000000002, "optim/total_tokens": 1134821376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453945, "loss/out": 6.047320365905762, "created_at": "2025-01-14T19:06:32.641927+00:00"} {"global_step": 4330, "acc_step": 0, "speed/wps": 12944.525362569188, "speed/FLOPS": 203311558461008.9, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 47.4586067199707, "optim/lr": 0.0025974, "optim/total_tokens": 1135083520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453660, "loss/out": 5.998445987701416, "created_at": "2025-01-14T19:06:37.709826+00:00"} {"global_step": 4331, "acc_step": 0, "speed/wps": 12939.467277171423, "speed/FLOPS": 203232114279299.3, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0008, "optim/grad_norm": 13.244027137756348, "optim/lr": 0.002598, "optim/total_tokens": 1135345664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493719, "loss/out": 6.087179183959961, "created_at": "2025-01-14T19:06:42.782393+00:00"} {"global_step": 4332, "acc_step": 0, "speed/wps": 12939.56517982321, "speed/FLOPS": 203233651974977.97, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 122.29845428466797, "optim/lr": 0.0025986, "optim/total_tokens": 1135607808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316156, "loss/out": 5.944378852844238, "created_at": "2025-01-14T19:06:47.851345+00:00"} {"global_step": 4333, "acc_step": 0, "speed/wps": 12939.131877777443, "speed/FLOPS": 203226846370929.8, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 26.874704360961914, "optim/lr": 0.0025992, "optim/total_tokens": 1135869952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448605, "loss/out": 6.161609649658203, "created_at": "2025-01-14T19:06:52.924384+00:00"} {"global_step": 4334, "acc_step": 0, "speed/wps": 12940.755412614439, "speed/FLOPS": 203252346216515.47, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0006, "optim/grad_norm": 25.833112716674805, "optim/lr": 0.0025998, "optim/total_tokens": 1136132096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461605, "loss/out": 6.131955146789551, "created_at": "2025-01-14T19:06:57.992493+00:00"} {"global_step": 4335, "acc_step": 0, "speed/wps": 12941.197122310436, "speed/FLOPS": 203259283874263.84, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.155630111694336, "optim/lr": 0.0026004, "optim/total_tokens": 1136394240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495573, "loss/out": 6.0687665939331055, "created_at": "2025-01-14T19:07:03.060923+00:00"} {"global_step": 4336, "acc_step": 0, "speed/wps": 12939.62203242339, "speed/FLOPS": 203234544923187.44, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0008, "optim/grad_norm": 23.36821937561035, "optim/lr": 0.002601, "optim/total_tokens": 1136656384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326583, "loss/out": 6.044795513153076, "created_at": "2025-01-14T19:07:08.133317+00:00"} {"global_step": 4337, "acc_step": 0, "speed/wps": 12942.09300250754, "speed/FLOPS": 203273354903827.75, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.978312492370605, "optim/lr": 0.0026016, "optim/total_tokens": 1136918528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466392, "loss/out": 5.947072982788086, "created_at": "2025-01-14T19:07:13.201917+00:00"} {"global_step": 4338, "acc_step": 0, "speed/wps": 12943.308055261416, "speed/FLOPS": 203292438976984.8, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 23.79752540588379, "optim/lr": 0.0026022, "optim/total_tokens": 1137180672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460840, "loss/out": 5.992375373840332, "created_at": "2025-01-14T19:07:18.269988+00:00"} {"global_step": 4339, "acc_step": 0, "speed/wps": 12946.516974659426, "speed/FLOPS": 203342839465647.9, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 64.54387664794922, "optim/lr": 0.0026028, "optim/total_tokens": 1137442816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464820, "loss/out": 5.957621097564697, "created_at": "2025-01-14T19:07:23.337950+00:00"} {"global_step": 4340, "acc_step": 0, "speed/wps": 12938.979308784268, "speed/FLOPS": 203224450065240.22, "speed/curr_iter_time": 1.2784, "speed/data_load_time": 0.0008, "optim/grad_norm": 28.6566162109375, "optim/lr": 0.0026034, "optim/total_tokens": 1137704960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451275, "loss/out": 6.012565612792969, "created_at": "2025-01-14T19:07:28.407830+00:00"} {"global_step": 4341, "acc_step": 0, "speed/wps": 12950.34271091629, "speed/FLOPS": 203402927910674.47, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 54.18690872192383, "optim/lr": 0.002604, "optim/total_tokens": 1137967104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333311, "loss/out": 5.999885559082031, "created_at": "2025-01-14T19:07:33.472481+00:00"} {"global_step": 4342, "acc_step": 0, "speed/wps": 12946.494579942088, "speed/FLOPS": 203342487725838.53, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 212.20455932617188, "optim/lr": 0.0026046, "optim/total_tokens": 1138229248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 387743, "loss/out": 6.062028408050537, "created_at": "2025-01-14T19:07:38.538410+00:00"} {"global_step": 4343, "acc_step": 0, "speed/wps": 12953.715570303582, "speed/FLOPS": 203455903302146.4, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0011, "optim/grad_norm": 13.812850952148438, "optim/lr": 0.0026052, "optim/total_tokens": 1138491392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449411, "loss/out": 6.11458158493042, "created_at": "2025-01-14T19:07:43.601445+00:00"} {"global_step": 4344, "acc_step": 0, "speed/wps": 12942.113715793752, "speed/FLOPS": 203273680234450.75, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 25.072460174560547, "optim/lr": 0.0026058, "optim/total_tokens": 1138753536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461100, "loss/out": 6.092864036560059, "created_at": "2025-01-14T19:07:48.672360+00:00"} {"global_step": 4345, "acc_step": 0, "speed/wps": 12938.79091446519, "speed/FLOPS": 203221491073578.16, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 28.398483276367188, "optim/lr": 0.0026064, "optim/total_tokens": 1139015680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443545, "loss/out": 6.059849739074707, "created_at": "2025-01-14T19:07:53.745262+00:00"} {"global_step": 4346, "acc_step": 0, "speed/wps": 12949.023006972146, "speed/FLOPS": 203382200146768.44, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 34.27822494506836, "optim/lr": 0.002607, "optim/total_tokens": 1139277824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292485, "loss/out": 6.011870861053467, "created_at": "2025-01-14T19:07:58.811572+00:00"} {"global_step": 4347, "acc_step": 0, "speed/wps": 12947.324532008837, "speed/FLOPS": 203355523263520.5, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0006, "optim/grad_norm": 44.78086471557617, "optim/lr": 0.0026076, "optim/total_tokens": 1139539968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 5.995232582092285, "created_at": "2025-01-14T19:08:03.882047+00:00"} {"global_step": 4348, "acc_step": 0, "speed/wps": 12942.106291660675, "speed/FLOPS": 203273563628238.94, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.880515098571777, "optim/lr": 0.0026081999999999998, "optim/total_tokens": 1139802112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 6.109795093536377, "created_at": "2025-01-14T19:08:08.955717+00:00"} {"global_step": 4349, "acc_step": 0, "speed/wps": 12942.761564306662, "speed/FLOPS": 203283855585583.16, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0006, "optim/grad_norm": 58.241275787353516, "optim/lr": 0.0026088, "optim/total_tokens": 1140064256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460271, "loss/out": 6.009881019592285, "created_at": "2025-01-14T19:08:14.023579+00:00"} {"global_step": 4350, "acc_step": 0, "speed/wps": 12939.073264135619, "speed/FLOPS": 203225925763141.38, "speed/curr_iter_time": 1.2784, "speed/data_load_time": 0.0007, "optim/grad_norm": 42.144901275634766, "optim/lr": 0.0026094, "optim/total_tokens": 1140326400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305035, "loss/out": 6.118010520935059, "created_at": "2025-01-14T19:08:19.092321+00:00"} {"global_step": 4351, "acc_step": 0, "speed/wps": 12946.488965792483, "speed/FLOPS": 203342399547904.34, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 16.4718074798584, "optim/lr": 0.00261, "optim/total_tokens": 1140588544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461100, "loss/out": 6.204918384552002, "created_at": "2025-01-14T19:08:24.159604+00:00"} {"global_step": 4352, "acc_step": 0, "speed/wps": 12946.440229843056, "speed/FLOPS": 203341634082848.4, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 18.230363845825195, "optim/lr": 0.0026106, "optim/total_tokens": 1140850688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 6.04006290435791, "created_at": "2025-01-14T19:08:29.232735+00:00"} {"global_step": 4353, "acc_step": 0, "speed/wps": 12945.425852243265, "speed/FLOPS": 203325701888742.78, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0006, "optim/grad_norm": 19.986507415771484, "optim/lr": 0.0026111999999999997, "optim/total_tokens": 1141112832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424963, "loss/out": 6.038211345672607, "created_at": "2025-01-14T19:08:34.300148+00:00"} {"global_step": 4354, "acc_step": 0, "speed/wps": 12937.866194820253, "speed/FLOPS": 203206967080856.5, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 124.85041809082031, "optim/lr": 0.0026118, "optim/total_tokens": 1141374976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 6.057336807250977, "created_at": "2025-01-14T19:08:39.372317+00:00"} {"global_step": 4355, "acc_step": 0, "speed/wps": 12939.784905254142, "speed/FLOPS": 203237103064805.16, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.111945152282715, "optim/lr": 0.0026124, "optim/total_tokens": 1141637120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314942, "loss/out": 6.1707963943481445, "created_at": "2025-01-14T19:08:44.445214+00:00"} {"global_step": 4356, "acc_step": 0, "speed/wps": 12944.003971542192, "speed/FLOPS": 203303369298463.6, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.707011222839355, "optim/lr": 0.002613, "optim/total_tokens": 1141899264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475696, "loss/out": 6.019782066345215, "created_at": "2025-01-14T19:08:49.516807+00:00"} {"global_step": 4357, "acc_step": 0, "speed/wps": 12944.307426239944, "speed/FLOPS": 203308135471482.25, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.58858871459961, "optim/lr": 0.0026135999999999998, "optim/total_tokens": 1142161408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422290, "loss/out": 6.188445091247559, "created_at": "2025-01-14T19:08:54.587604+00:00"} {"global_step": 4358, "acc_step": 0, "speed/wps": 12934.38562168669, "speed/FLOPS": 203152299896985.2, "speed/curr_iter_time": 1.2787, "speed/data_load_time": 0.0007, "optim/grad_norm": 38.23379135131836, "optim/lr": 0.0026142, "optim/total_tokens": 1142423552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432365, "loss/out": 6.0268354415893555, "created_at": "2025-01-14T19:08:59.665925+00:00"} {"global_step": 4359, "acc_step": 0, "speed/wps": 12937.034305835437, "speed/FLOPS": 203193901121214.7, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0007, "optim/grad_norm": 225.54437255859375, "optim/lr": 0.0026148, "optim/total_tokens": 1142685696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453131, "loss/out": 6.060495376586914, "created_at": "2025-01-14T19:09:04.743793+00:00"} {"global_step": 4360, "acc_step": 0, "speed/wps": 12944.210066129026, "speed/FLOPS": 203306606297153.47, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0008, "optim/grad_norm": 64.64209747314453, "optim/lr": 0.0026154, "optim/total_tokens": 1142947840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 6.217087268829346, "created_at": "2025-01-14T19:09:09.811722+00:00"} {"global_step": 4361, "acc_step": 0, "speed/wps": 12940.934239984022, "speed/FLOPS": 203255154946089.34, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0008, "optim/grad_norm": 18.29026222229004, "optim/lr": 0.002616, "optim/total_tokens": 1143209984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480143, "loss/out": 6.25142765045166, "created_at": "2025-01-14T19:09:14.879580+00:00"} {"global_step": 4362, "acc_step": 0, "speed/wps": 12944.991769446666, "speed/FLOPS": 203318884021929.9, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0013, "optim/grad_norm": 20.18337631225586, "optim/lr": 0.0026166, "optim/total_tokens": 1143472128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332731, "loss/out": 6.121894836425781, "created_at": "2025-01-14T19:09:19.946256+00:00"} {"global_step": 4363, "acc_step": 0, "speed/wps": 12943.737620493168, "speed/FLOPS": 203299185889233.97, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0006, "optim/grad_norm": 56.53289031982422, "optim/lr": 0.0026172, "optim/total_tokens": 1143734272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457889, "loss/out": 6.276674270629883, "created_at": "2025-01-14T19:09:25.016280+00:00"} {"global_step": 4364, "acc_step": 0, "speed/wps": 12936.151513344346, "speed/FLOPS": 203180035652058.12, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.87585163116455, "optim/lr": 0.0026178, "optim/total_tokens": 1143996416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454995, "loss/out": 6.169386863708496, "created_at": "2025-01-14T19:09:30.092701+00:00"} {"global_step": 4365, "acc_step": 0, "speed/wps": 12945.305748914403, "speed/FLOPS": 203323815500923.75, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 16.920509338378906, "optim/lr": 0.0026184000000000003, "optim/total_tokens": 1144258560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 5.966717720031738, "created_at": "2025-01-14T19:09:35.160474+00:00"} {"global_step": 4366, "acc_step": 0, "speed/wps": 12939.128987053238, "speed/FLOPS": 203226800968133.9, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0009, "optim/grad_norm": 24.550750732421875, "optim/lr": 0.0026190000000000002, "optim/total_tokens": 1144520704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297588, "loss/out": 6.068819046020508, "created_at": "2025-01-14T19:09:40.235308+00:00"} {"global_step": 4367, "acc_step": 0, "speed/wps": 12933.71186303377, "speed/FLOPS": 203141717591499.7, "speed/curr_iter_time": 1.2785, "speed/data_load_time": 0.0007, "optim/grad_norm": 90.961181640625, "optim/lr": 0.0026196, "optim/total_tokens": 1144782848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355857, "loss/out": 6.104503631591797, "created_at": "2025-01-14T19:09:45.314070+00:00"} {"global_step": 4368, "acc_step": 0, "speed/wps": 12944.339602139198, "speed/FLOPS": 203308640838194.34, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0007, "optim/grad_norm": 41.93701934814453, "optim/lr": 0.0026202, "optim/total_tokens": 1145044992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 6.1539154052734375, "created_at": "2025-01-14T19:09:50.380615+00:00"} {"global_step": 4369, "acc_step": 0, "speed/wps": 12943.088472036534, "speed/FLOPS": 203288990120699.1, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0006, "optim/grad_norm": 31.37312889099121, "optim/lr": 0.0026208000000000004, "optim/total_tokens": 1145307136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430268, "loss/out": 6.232296466827393, "created_at": "2025-01-14T19:09:55.448022+00:00"} {"global_step": 4370, "acc_step": 0, "speed/wps": 12942.740437115915, "speed/FLOPS": 203283523754020.7, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.02059555053711, "optim/lr": 0.0026214000000000003, "optim/total_tokens": 1145569280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450756, "loss/out": 6.114310264587402, "created_at": "2025-01-14T19:10:00.517377+00:00"} {"global_step": 4371, "acc_step": 0, "speed/wps": 12945.747819319862, "speed/FLOPS": 203330758824109.88, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0009, "optim/grad_norm": 25.44756317138672, "optim/lr": 0.002622, "optim/total_tokens": 1145831424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 389016, "loss/out": 6.195497512817383, "created_at": "2025-01-14T19:10:05.592510+00:00"} {"global_step": 4372, "acc_step": 0, "speed/wps": 12941.087595237845, "speed/FLOPS": 203257563601082.8, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0008, "optim/grad_norm": 29.47781753540039, "optim/lr": 0.0026226, "optim/total_tokens": 1146093568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 6.063558578491211, "created_at": "2025-01-14T19:10:10.665476+00:00"} {"global_step": 4373, "acc_step": 0, "speed/wps": 12933.564996246738, "speed/FLOPS": 203139410846793.22, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 17.201128005981445, "optim/lr": 0.0026232, "optim/total_tokens": 1146355712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483617, "loss/out": 6.0794782638549805, "created_at": "2025-01-14T19:10:15.736831+00:00"} {"global_step": 4374, "acc_step": 0, "speed/wps": 12941.724128122234, "speed/FLOPS": 203267561224720.0, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 17.19784927368164, "optim/lr": 0.0026238000000000004, "optim/total_tokens": 1146617856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334369, "loss/out": 6.110030174255371, "created_at": "2025-01-14T19:10:20.806603+00:00"} {"global_step": 4375, "acc_step": 0, "speed/wps": 12936.27046945327, "speed/FLOPS": 203181904021211.56, "speed/curr_iter_time": 1.2774, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.043259620666504, "optim/lr": 0.0026244000000000003, "optim/total_tokens": 1146880000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439809, "loss/out": 6.008562088012695, "created_at": "2025-01-14T19:10:25.876198+00:00"} {"global_step": 4376, "acc_step": 0, "speed/wps": 12935.644429236872, "speed/FLOPS": 203172071199343.66, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.655009746551514, "optim/lr": 0.002625, "optim/total_tokens": 1147142144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 6.157838344573975, "created_at": "2025-01-14T19:10:30.951952+00:00"} {"global_step": 4377, "acc_step": 0, "speed/wps": 12943.71733294865, "speed/FLOPS": 203298867245469.3, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.4364914894104, "optim/lr": 0.0026256, "optim/total_tokens": 1147404288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479906, "loss/out": 6.900594234466553, "created_at": "2025-01-14T19:10:36.022476+00:00"} {"global_step": 4378, "acc_step": 0, "speed/wps": 12940.909560880016, "speed/FLOPS": 203254767326848.28, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.001, "optim/grad_norm": 10.528414726257324, "optim/lr": 0.0026262, "optim/total_tokens": 1147666432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319909, "loss/out": 6.9249958992004395, "created_at": "2025-01-14T19:10:41.090501+00:00"} {"global_step": 4379, "acc_step": 0, "speed/wps": 12949.345739213191, "speed/FLOPS": 203387269100094.62, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.2603759765625, "optim/lr": 0.0026268000000000003, "optim/total_tokens": 1147928576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355086, "loss/out": 7.01566743850708, "created_at": "2025-01-14T19:10:46.160475+00:00"} {"global_step": 4380, "acc_step": 0, "speed/wps": 12949.942065432844, "speed/FLOPS": 203396635222812.38, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.2792177200317383, "optim/lr": 0.0026274000000000002, "optim/total_tokens": 1148190720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452081, "loss/out": 7.228512763977051, "created_at": "2025-01-14T19:10:51.225902+00:00"} {"global_step": 4381, "acc_step": 0, "speed/wps": 12946.99146933174, "speed/FLOPS": 203350292056500.62, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 22.897218704223633, "optim/lr": 0.002628, "optim/total_tokens": 1148452864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 7.097332000732422, "created_at": "2025-01-14T19:10:56.292650+00:00"} {"global_step": 4382, "acc_step": 0, "speed/wps": 12949.04850813627, "speed/FLOPS": 203382600677593.28, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.846163272857666, "optim/lr": 0.0026286, "optim/total_tokens": 1148715008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456851, "loss/out": 7.211112976074219, "created_at": "2025-01-14T19:11:01.359038+00:00"} {"global_step": 4383, "acc_step": 0, "speed/wps": 12956.164519812299, "speed/FLOPS": 203494367419390.06, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.756202697753906, "optim/lr": 0.0026292, "optim/total_tokens": 1148977152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451559, "loss/out": 7.167259216308594, "created_at": "2025-01-14T19:11:06.422949+00:00"} {"global_step": 4384, "acc_step": 0, "speed/wps": 12954.119840374377, "speed/FLOPS": 203462252919133.5, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0009, "optim/grad_norm": 2.648099422454834, "optim/lr": 0.0026298000000000003, "optim/total_tokens": 1149239296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292773, "loss/out": 7.0682172775268555, "created_at": "2025-01-14T19:11:11.486377+00:00"} {"global_step": 4385, "acc_step": 0, "speed/wps": 12945.040725484405, "speed/FLOPS": 203319652943775.25, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.800068736076355, "optim/lr": 0.0026304, "optim/total_tokens": 1149501440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449696, "loss/out": 7.006474018096924, "created_at": "2025-01-14T19:11:16.556693+00:00"} {"global_step": 4386, "acc_step": 0, "speed/wps": 12948.116889658453, "speed/FLOPS": 203367968329221.22, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7075958251953125, "optim/lr": 0.002631, "optim/total_tokens": 1149763584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474645, "loss/out": 6.877985954284668, "created_at": "2025-01-14T19:11:21.621761+00:00"} {"global_step": 4387, "acc_step": 0, "speed/wps": 12951.805756845271, "speed/FLOPS": 203425907057424.9, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1078970432281494, "optim/lr": 0.0026316, "optim/total_tokens": 1150025728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459528, "loss/out": 6.845795631408691, "created_at": "2025-01-14T19:11:26.689320+00:00"} {"global_step": 4388, "acc_step": 0, "speed/wps": 12952.183619299634, "speed/FLOPS": 203431841906508.12, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7353476285934448, "optim/lr": 0.0026322, "optim/total_tokens": 1150287872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415893, "loss/out": 6.801354885101318, "created_at": "2025-01-14T19:11:31.752626+00:00"} {"global_step": 4389, "acc_step": 0, "speed/wps": 12945.919094201694, "speed/FLOPS": 203333448931485.38, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.010600090026855, "optim/lr": 0.0026328000000000002, "optim/total_tokens": 1150550016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359462, "loss/out": 6.593752861022949, "created_at": "2025-01-14T19:11:36.819645+00:00"} {"global_step": 4390, "acc_step": 0, "speed/wps": 12946.046894785775, "speed/FLOPS": 203335456215274.78, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.385251045227051, "optim/lr": 0.0026334, "optim/total_tokens": 1150812160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 6.920592308044434, "created_at": "2025-01-14T19:11:41.885564+00:00"} {"global_step": 4391, "acc_step": 0, "speed/wps": 12948.906604083262, "speed/FLOPS": 203380371879444.2, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.586066722869873, "optim/lr": 0.002634, "optim/total_tokens": 1151074304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439586, "loss/out": 6.734225749969482, "created_at": "2025-01-14T19:11:46.952956+00:00"} {"global_step": 4392, "acc_step": 0, "speed/wps": 12950.409677844222, "speed/FLOPS": 203403979718299.84, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.806146621704102, "optim/lr": 0.0026346, "optim/total_tokens": 1151336448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 6.514763832092285, "created_at": "2025-01-14T19:11:52.017682+00:00"} {"global_step": 4393, "acc_step": 0, "speed/wps": 12941.613901896919, "speed/FLOPS": 203265829970384.72, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.001, "optim/grad_norm": 12.21484661102295, "optim/lr": 0.0026352, "optim/total_tokens": 1151598592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435329, "loss/out": 6.578868865966797, "created_at": "2025-01-14T19:11:57.085877+00:00"} {"global_step": 4394, "acc_step": 0, "speed/wps": 12950.235617227705, "speed/FLOPS": 203401245857133.1, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.106200218200684, "optim/lr": 0.0026358, "optim/total_tokens": 1151860736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297240, "loss/out": 6.612403869628906, "created_at": "2025-01-14T19:12:02.151452+00:00"} {"global_step": 4395, "acc_step": 0, "speed/wps": 12950.96894713727, "speed/FLOPS": 203412763811063.72, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.110801696777344, "optim/lr": 0.0026364, "optim/total_tokens": 1152122880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462981, "loss/out": 6.588696479797363, "created_at": "2025-01-14T19:12:07.216616+00:00"} {"global_step": 4396, "acc_step": 0, "speed/wps": 12952.840809295243, "speed/FLOPS": 203442163978462.84, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 98.74288177490234, "optim/lr": 0.002637, "optim/total_tokens": 1152385024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 6.594686508178711, "created_at": "2025-01-14T19:12:12.280960+00:00"} {"global_step": 4397, "acc_step": 0, "speed/wps": 12949.299238395115, "speed/FLOPS": 203386538740847.94, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0013, "optim/grad_norm": 9.867003440856934, "optim/lr": 0.0026376, "optim/total_tokens": 1152647168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445976, "loss/out": 6.70296573638916, "created_at": "2025-01-14T19:12:17.348460+00:00"} {"global_step": 4398, "acc_step": 0, "speed/wps": 12950.094047023222, "speed/FLOPS": 203399022302533.84, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 11.23628044128418, "optim/lr": 0.0026382, "optim/total_tokens": 1152909312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432668, "loss/out": 6.537106990814209, "created_at": "2025-01-14T19:12:22.413105+00:00"} {"global_step": 4399, "acc_step": 0, "speed/wps": 12956.036135340804, "speed/FLOPS": 203492350964846.5, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 21.27471351623535, "optim/lr": 0.0026388, "optim/total_tokens": 1153171456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290845, "loss/out": 6.494502544403076, "created_at": "2025-01-14T19:12:27.480572+00:00"} {"global_step": 4400, "acc_step": 0, "speed/wps": 12956.599943888874, "speed/FLOPS": 203501206352848.06, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 59.26849365234375, "optim/lr": 0.0026394, "optim/total_tokens": 1153433600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448361, "loss/out": 6.545838356018066, "created_at": "2025-01-14T19:12:32.546260+00:00"} {"global_step": 4401, "acc_step": 0, "speed/wps": 12955.760021930088, "speed/FLOPS": 203488014224311.6, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 18.650272369384766, "optim/lr": 0.00264, "optim/total_tokens": 1153695744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 6.505706787109375, "created_at": "2025-01-14T19:12:37.608964+00:00"} {"global_step": 4402, "acc_step": 0, "speed/wps": 12948.78679357191, "speed/FLOPS": 203378490090726.5, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0009, "optim/grad_norm": 16.108396530151367, "optim/lr": 0.0026406, "optim/total_tokens": 1153957888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 6.522647380828857, "created_at": "2025-01-14T19:12:42.685856+00:00"} {"global_step": 4403, "acc_step": 0, "speed/wps": 12946.926812621912, "speed/FLOPS": 203349276534025.12, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 29.778379440307617, "optim/lr": 0.0026412, "optim/total_tokens": 1154220032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302050, "loss/out": 6.566615104675293, "created_at": "2025-01-14T19:12:47.755719+00:00"} {"global_step": 4404, "acc_step": 0, "speed/wps": 12947.93876190364, "speed/FLOPS": 203365170588058.72, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.518277168273926, "optim/lr": 0.0026418, "optim/total_tokens": 1154482176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464820, "loss/out": 6.556768417358398, "created_at": "2025-01-14T19:12:52.829749+00:00"} {"global_step": 4405, "acc_step": 0, "speed/wps": 12950.703076847372, "speed/FLOPS": 203408587952816.8, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 24.0937442779541, "optim/lr": 0.0026424, "optim/total_tokens": 1154744320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435582, "loss/out": 6.500990867614746, "created_at": "2025-01-14T19:12:57.898888+00:00"} {"global_step": 4406, "acc_step": 0, "speed/wps": 12952.4579114184, "speed/FLOPS": 203436150041150.2, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.3624677658081055, "optim/lr": 0.002643, "optim/total_tokens": 1155006464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 6.50363302230835, "created_at": "2025-01-14T19:13:02.963656+00:00"} {"global_step": 4407, "acc_step": 0, "speed/wps": 12949.64560292402, "speed/FLOPS": 203391978871729.12, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 24.352252960205078, "optim/lr": 0.0026436, "optim/total_tokens": 1155268608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440105, "loss/out": 6.363034725189209, "created_at": "2025-01-14T19:13:08.028094+00:00"} {"global_step": 4408, "acc_step": 0, "speed/wps": 12947.6086907828, "speed/FLOPS": 203359986367540.66, "speed/curr_iter_time": 1.2778, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.283149242401123, "optim/lr": 0.0026441999999999998, "optim/total_tokens": 1155530752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330862, "loss/out": 6.37432861328125, "created_at": "2025-01-14T19:13:13.098129+00:00"} {"global_step": 4409, "acc_step": 0, "speed/wps": 12936.18927057204, "speed/FLOPS": 203180628681202.9, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.250887393951416, "optim/lr": 0.0026448, "optim/total_tokens": 1155792896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 6.380334377288818, "created_at": "2025-01-14T19:13:18.169224+00:00"} {"global_step": 4410, "acc_step": 0, "speed/wps": 12945.05883345403, "speed/FLOPS": 203319937354321.2, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0009, "optim/grad_norm": 6.263243198394775, "optim/lr": 0.0026454, "optim/total_tokens": 1156055040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 6.243356704711914, "created_at": "2025-01-14T19:13:23.239323+00:00"} {"global_step": 4411, "acc_step": 0, "speed/wps": 12946.453220661408, "speed/FLOPS": 203341838121501.84, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.970499038696289, "optim/lr": 0.002646, "optim/total_tokens": 1156317184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 6.325821876525879, "created_at": "2025-01-14T19:13:28.307375+00:00"} {"global_step": 4412, "acc_step": 0, "speed/wps": 12945.379403388288, "speed/FLOPS": 203324972345647.94, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 31.333786010742188, "optim/lr": 0.0026466, "optim/total_tokens": 1156579328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439583, "loss/out": 6.378814697265625, "created_at": "2025-01-14T19:13:33.373494+00:00"} {"global_step": 4413, "acc_step": 0, "speed/wps": 12947.393340163646, "speed/FLOPS": 203356603990144.1, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 20.810604095458984, "optim/lr": 0.0026471999999999997, "optim/total_tokens": 1156841472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 397678, "loss/out": 6.253831386566162, "created_at": "2025-01-14T19:13:38.440748+00:00"} {"global_step": 4414, "acc_step": 0, "speed/wps": 12950.633537548041, "speed/FLOPS": 203407495742563.8, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 23.046266555786133, "optim/lr": 0.0026478, "optim/total_tokens": 1157103616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 6.317582130432129, "created_at": "2025-01-14T19:13:43.504661+00:00"} {"global_step": 4415, "acc_step": 0, "speed/wps": 12941.066447332769, "speed/FLOPS": 203257231444173.34, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0008, "optim/grad_norm": 20.665023803710938, "optim/lr": 0.0026484, "optim/total_tokens": 1157365760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423864, "loss/out": 6.240765571594238, "created_at": "2025-01-14T19:13:48.575446+00:00"} {"global_step": 4416, "acc_step": 0, "speed/wps": 12943.923216026204, "speed/FLOPS": 203302100922115.28, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0006, "optim/grad_norm": 85.84825134277344, "optim/lr": 0.002649, "optim/total_tokens": 1157627904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458734, "loss/out": 6.283924102783203, "created_at": "2025-01-14T19:13:53.643558+00:00"} {"global_step": 4417, "acc_step": 0, "speed/wps": 12939.908465867944, "speed/FLOPS": 203239043754034.97, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.171189308166504, "optim/lr": 0.0026496000000000002, "optim/total_tokens": 1157890048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 342387, "loss/out": 6.358572483062744, "created_at": "2025-01-14T19:13:58.712162+00:00"} {"global_step": 4418, "acc_step": 0, "speed/wps": 12943.242647321022, "speed/FLOPS": 203291411655401.0, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.86279582977295, "optim/lr": 0.0026502, "optim/total_tokens": 1158152192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452081, "loss/out": 6.301063537597656, "created_at": "2025-01-14T19:14:03.779366+00:00"} {"global_step": 4419, "acc_step": 0, "speed/wps": 12944.793005736083, "speed/FLOPS": 203315762164725.25, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0009, "optim/grad_norm": 105.74767303466797, "optim/lr": 0.0026508, "optim/total_tokens": 1158414336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 6.317838668823242, "created_at": "2025-01-14T19:14:08.850520+00:00"} {"global_step": 4420, "acc_step": 0, "speed/wps": 12936.240897284873, "speed/FLOPS": 203181439549670.5, "speed/curr_iter_time": 1.2778, "speed/data_load_time": 0.0007, "optim/grad_norm": 31.136884689331055, "optim/lr": 0.0026514, "optim/total_tokens": 1158676480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449135, "loss/out": 6.450286388397217, "created_at": "2025-01-14T19:14:13.920335+00:00"} {"global_step": 4421, "acc_step": 0, "speed/wps": 12939.255861321577, "speed/FLOPS": 203228793702861.66, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.721408367156982, "optim/lr": 0.0026520000000000003, "optim/total_tokens": 1158938624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 6.316678047180176, "created_at": "2025-01-14T19:14:18.989206+00:00"} {"global_step": 4422, "acc_step": 0, "speed/wps": 12936.188735496102, "speed/FLOPS": 203180620277100.03, "speed/curr_iter_time": 1.2771, "speed/data_load_time": 0.001, "optim/grad_norm": 10.313376426696777, "optim/lr": 0.0026526, "optim/total_tokens": 1159200768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290558, "loss/out": 6.256186008453369, "created_at": "2025-01-14T19:14:24.063468+00:00"} {"global_step": 4423, "acc_step": 0, "speed/wps": 12942.689963048626, "speed/FLOPS": 203282730989435.25, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 25.483591079711914, "optim/lr": 0.0026532, "optim/total_tokens": 1159462912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450471, "loss/out": 6.385008335113525, "created_at": "2025-01-14T19:14:29.134163+00:00"} {"global_step": 4424, "acc_step": 0, "speed/wps": 12941.645007632742, "speed/FLOPS": 203266318528709.6, "speed/curr_iter_time": 1.2774, "speed/data_load_time": 0.0007, "optim/grad_norm": 44.13283920288086, "optim/lr": 0.0026538000000000004, "optim/total_tokens": 1159725056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 6.239345073699951, "created_at": "2025-01-14T19:14:34.202823+00:00"} {"global_step": 4425, "acc_step": 0, "speed/wps": 12938.817679080774, "speed/FLOPS": 203221911448644.34, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0007, "optim/grad_norm": 44.424617767333984, "optim/lr": 0.0026544000000000003, "optim/total_tokens": 1159987200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 6.4415130615234375, "created_at": "2025-01-14T19:14:39.274920+00:00"} {"global_step": 4426, "acc_step": 0, "speed/wps": 12940.074620161453, "speed/FLOPS": 203241653435534.47, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0007, "optim/grad_norm": 20.465436935424805, "optim/lr": 0.0026550000000000002, "optim/total_tokens": 1160249344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309845, "loss/out": 6.388813018798828, "created_at": "2025-01-14T19:14:44.346022+00:00"} {"global_step": 4427, "acc_step": 0, "speed/wps": 12937.464991297547, "speed/FLOPS": 203200665628220.97, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.993338584899902, "optim/lr": 0.0026556, "optim/total_tokens": 1160511488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461913, "loss/out": 6.469208717346191, "created_at": "2025-01-14T19:14:49.418703+00:00"} {"global_step": 4428, "acc_step": 0, "speed/wps": 12937.157121789049, "speed/FLOPS": 203195830114532.97, "speed/curr_iter_time": 1.2774, "speed/data_load_time": 0.0006, "optim/grad_norm": 176.69229125976562, "optim/lr": 0.0026562, "optim/total_tokens": 1160773632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441679, "loss/out": 6.35713005065918, "created_at": "2025-01-14T19:14:54.488458+00:00"} {"global_step": 4429, "acc_step": 0, "speed/wps": 12939.920363293926, "speed/FLOPS": 203239230619458.8, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.570784568786621, "optim/lr": 0.0026568000000000004, "optim/total_tokens": 1161035776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466369, "loss/out": 6.458688735961914, "created_at": "2025-01-14T19:14:59.556811+00:00"} {"global_step": 4430, "acc_step": 0, "speed/wps": 12937.108839003024, "speed/FLOPS": 203195071767030.94, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0006, "optim/grad_norm": 99.86434936523438, "optim/lr": 0.0026574000000000003, "optim/total_tokens": 1161297920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324322, "loss/out": 6.428030014038086, "created_at": "2025-01-14T19:15:04.627136+00:00"} {"global_step": 4431, "acc_step": 0, "speed/wps": 12936.08282462539, "speed/FLOPS": 203178956801338.72, "speed/curr_iter_time": 1.2774, "speed/data_load_time": 0.0006, "optim/grad_norm": 19.818439483642578, "optim/lr": 0.002658, "optim/total_tokens": 1161560064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451559, "loss/out": 6.572972297668457, "created_at": "2025-01-14T19:15:09.699600+00:00"} {"global_step": 4432, "acc_step": 0, "speed/wps": 12923.003010300983, "speed/FLOPS": 202973520343825.53, "speed/curr_iter_time": 1.2785, "speed/data_load_time": 0.0008, "optim/grad_norm": 53.06034851074219, "optim/lr": 0.0026586, "optim/total_tokens": 1161822208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445695, "loss/out": 6.464656829833984, "created_at": "2025-01-14T19:15:14.774566+00:00"} {"global_step": 4433, "acc_step": 0, "speed/wps": 12928.173074295391, "speed/FLOPS": 203054723303270.12, "speed/curr_iter_time": 1.2782, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.260801315307617, "optim/lr": 0.0026592, "optim/total_tokens": 1162084352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 6.475515365600586, "created_at": "2025-01-14T19:15:19.849376+00:00"} {"global_step": 4434, "acc_step": 0, "speed/wps": 12938.137040533826, "speed/FLOPS": 203211221084969.16, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.019660949707031, "optim/lr": 0.0026598000000000004, "optim/total_tokens": 1162346496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331435, "loss/out": 6.39979887008667, "created_at": "2025-01-14T19:15:24.923333+00:00"} {"global_step": 4435, "acc_step": 0, "speed/wps": 12934.364036833445, "speed/FLOPS": 203151960877204.28, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.929439544677734, "optim/lr": 0.0026604000000000003, "optim/total_tokens": 1162608640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449135, "loss/out": 6.309632778167725, "created_at": "2025-01-14T19:15:29.993824+00:00"} {"global_step": 4436, "acc_step": 0, "speed/wps": 12929.854180961493, "speed/FLOPS": 203081127393543.88, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.12643814086914, "optim/lr": 0.002661, "optim/total_tokens": 1162870784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465083, "loss/out": 6.463934898376465, "created_at": "2025-01-14T19:15:35.066896+00:00"} {"global_step": 4437, "acc_step": 0, "speed/wps": 12934.664505229786, "speed/FLOPS": 203156680146255.66, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0009, "optim/grad_norm": 18.593475341796875, "optim/lr": 0.0026616, "optim/total_tokens": 1163132928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 6.520981788635254, "created_at": "2025-01-14T19:15:40.141321+00:00"} {"global_step": 4438, "acc_step": 0, "speed/wps": 12933.517203810225, "speed/FLOPS": 203138660200904.1, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 60.85667419433594, "optim/lr": 0.0026622, "optim/total_tokens": 1163395072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306381, "loss/out": 6.564211368560791, "created_at": "2025-01-14T19:15:45.216272+00:00"} {"global_step": 4439, "acc_step": 0, "speed/wps": 12930.146594654452, "speed/FLOPS": 203085720152410.34, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0008, "optim/grad_norm": 29.339130401611328, "optim/lr": 0.0026628, "optim/total_tokens": 1163657216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 6.617002010345459, "created_at": "2025-01-14T19:15:50.292545+00:00"} {"global_step": 4440, "acc_step": 0, "speed/wps": 12924.840851624393, "speed/FLOPS": 203002386167268.3, "speed/curr_iter_time": 1.2783, "speed/data_load_time": 0.0007, "optim/grad_norm": 75.37628936767578, "optim/lr": 0.0026634000000000002, "optim/total_tokens": 1163919360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454182, "loss/out": 6.487491607666016, "created_at": "2025-01-14T19:15:55.372278+00:00"} {"global_step": 4441, "acc_step": 0, "speed/wps": 12929.865045443372, "speed/FLOPS": 203081298035161.16, "speed/curr_iter_time": 1.2781, "speed/data_load_time": 0.0007, "optim/grad_norm": 20.822608947753906, "optim/lr": 0.002664, "optim/total_tokens": 1164181504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326966, "loss/out": 6.592087745666504, "created_at": "2025-01-14T19:16:00.448711+00:00"} {"global_step": 4442, "acc_step": 0, "speed/wps": 12936.02659287896, "speed/FLOPS": 203178073604490.5, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0006, "optim/grad_norm": 28.128826141357422, "optim/lr": 0.0026646, "optim/total_tokens": 1164443648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462150, "loss/out": 6.564863204956055, "created_at": "2025-01-14T19:16:05.518771+00:00"} {"global_step": 4443, "acc_step": 0, "speed/wps": 12931.38888375433, "speed/FLOPS": 203105232009808.03, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0007, "optim/grad_norm": 35.827659606933594, "optim/lr": 0.0026652, "optim/total_tokens": 1164705792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 6.7004547119140625, "created_at": "2025-01-14T19:16:10.590692+00:00"} {"global_step": 4444, "acc_step": 0, "speed/wps": 12932.888254501438, "speed/FLOPS": 203128781687742.7, "speed/curr_iter_time": 1.2774, "speed/data_load_time": 0.0007, "optim/grad_norm": 35.24607849121094, "optim/lr": 0.0026658, "optim/total_tokens": 1164967936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 6.619501113891602, "created_at": "2025-01-14T19:16:15.668698+00:00"} {"global_step": 4445, "acc_step": 0, "speed/wps": 12928.458129361305, "speed/FLOPS": 203059200484787.0, "speed/curr_iter_time": 1.2784, "speed/data_load_time": 0.0006, "optim/grad_norm": 303.7909851074219, "optim/lr": 0.0026664, "optim/total_tokens": 1165230080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294941, "loss/out": 6.627067565917969, "created_at": "2025-01-14T19:16:20.742093+00:00"} {"global_step": 4446, "acc_step": 0, "speed/wps": 12931.42674388543, "speed/FLOPS": 203105826655192.22, "speed/curr_iter_time": 1.2778, "speed/data_load_time": 0.0007, "optim/grad_norm": 19.066139221191406, "optim/lr": 0.002667, "optim/total_tokens": 1165492224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452325, "loss/out": 6.670530319213867, "created_at": "2025-01-14T19:16:25.814895+00:00"} {"global_step": 4447, "acc_step": 0, "speed/wps": 12937.584587629064, "speed/FLOPS": 203202544052950.03, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.983797073364258, "optim/lr": 0.0026676, "optim/total_tokens": 1165754368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 6.499975681304932, "created_at": "2025-01-14T19:16:30.887532+00:00"} {"global_step": 4448, "acc_step": 0, "speed/wps": 12929.762504752887, "speed/FLOPS": 203079687492711.44, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0008, "optim/grad_norm": 22.15235137939453, "optim/lr": 0.0026682, "optim/total_tokens": 1166016512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451559, "loss/out": 6.459436416625977, "created_at": "2025-01-14T19:16:35.959752+00:00"} {"global_step": 4449, "acc_step": 0, "speed/wps": 12934.55672443591, "speed/FLOPS": 203154987300781.5, "speed/curr_iter_time": 1.2778, "speed/data_load_time": 0.0007, "optim/grad_norm": 20.460325241088867, "optim/lr": 0.0026688, "optim/total_tokens": 1166278656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443545, "loss/out": 6.5309906005859375, "created_at": "2025-01-14T19:16:41.030028+00:00"} {"global_step": 4450, "acc_step": 0, "speed/wps": 12935.37548074548, "speed/FLOPS": 203167846993712.38, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0009, "optim/grad_norm": 8.08424186706543, "optim/lr": 0.0026694, "optim/total_tokens": 1166540800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375182, "loss/out": 6.572614669799805, "created_at": "2025-01-14T19:16:46.103393+00:00"} {"global_step": 4451, "acc_step": 0, "speed/wps": 12929.429064533075, "speed/FLOPS": 203074450355864.4, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0008, "optim/grad_norm": 71.41576385498047, "optim/lr": 0.00267, "optim/total_tokens": 1166802944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 6.658280372619629, "created_at": "2025-01-14T19:16:51.179348+00:00"} {"global_step": 4452, "acc_step": 0, "speed/wps": 12932.55176288402, "speed/FLOPS": 203123496624503.2, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.206541061401367, "optim/lr": 0.0026706, "optim/total_tokens": 1167065088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 6.765135288238525, "created_at": "2025-01-14T19:16:56.279012+00:00"} {"global_step": 4453, "acc_step": 0, "speed/wps": 12931.977853467919, "speed/FLOPS": 203114482588489.16, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.52661657333374, "optim/lr": 0.0026712, "optim/total_tokens": 1167327232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459256, "loss/out": 6.7049946784973145, "created_at": "2025-01-14T19:17:01.350580+00:00"} {"global_step": 4454, "acc_step": 0, "speed/wps": 12934.087897813162, "speed/FLOPS": 203147623734435.97, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.751111030578613, "optim/lr": 0.0026717999999999998, "optim/total_tokens": 1167589376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 6.8115949630737305, "created_at": "2025-01-14T19:17:06.423555+00:00"} {"global_step": 4455, "acc_step": 0, "speed/wps": 12934.557031135377, "speed/FLOPS": 203154992117918.1, "speed/curr_iter_time": 1.2771, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.745141983032227, "optim/lr": 0.0026724, "optim/total_tokens": 1167851520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 6.674949645996094, "created_at": "2025-01-14T19:17:11.496484+00:00"} {"global_step": 4456, "acc_step": 0, "speed/wps": 12931.796640379645, "speed/FLOPS": 203111636387924.94, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.188358306884766, "optim/lr": 0.002673, "optim/total_tokens": 1168113664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326344, "loss/out": 6.84898042678833, "created_at": "2025-01-14T19:17:16.569620+00:00"} {"global_step": 4457, "acc_step": 0, "speed/wps": 12935.886762344915, "speed/FLOPS": 203175877373806.03, "speed/curr_iter_time": 1.2784, "speed/data_load_time": 0.001, "optim/grad_norm": 6.332953929901123, "optim/lr": 0.0026736, "optim/total_tokens": 1168375808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451797, "loss/out": 6.547520160675049, "created_at": "2025-01-14T19:17:21.640032+00:00"} {"global_step": 4458, "acc_step": 0, "speed/wps": 12932.556621020429, "speed/FLOPS": 203123572928211.28, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0006, "optim/grad_norm": 47.99391555786133, "optim/lr": 0.0026742, "optim/total_tokens": 1168637952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 6.610527515411377, "created_at": "2025-01-14T19:17:26.711161+00:00"} {"global_step": 4459, "acc_step": 0, "speed/wps": 12932.485206307423, "speed/FLOPS": 203122451262009.4, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.2797627449035645, "optim/lr": 0.0026747999999999997, "optim/total_tokens": 1168900096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 6.719846725463867, "created_at": "2025-01-14T19:17:31.782680+00:00"} {"global_step": 4460, "acc_step": 0, "speed/wps": 12939.20809581537, "speed/FLOPS": 203228043479950.0, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.266935348510742, "optim/lr": 0.0026754, "optim/total_tokens": 1169162240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 6.945986747741699, "created_at": "2025-01-14T19:17:36.853939+00:00"} {"global_step": 4461, "acc_step": 0, "speed/wps": 12942.5973096542, "speed/FLOPS": 203281275740556.7, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.308567762374878, "optim/lr": 0.002676, "optim/total_tokens": 1169424384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327115, "loss/out": 6.924065589904785, "created_at": "2025-01-14T19:17:41.925673+00:00"} {"global_step": 4462, "acc_step": 0, "speed/wps": 12946.40738133892, "speed/FLOPS": 203341118151952.38, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.85469388961792, "optim/lr": 0.0026766, "optim/total_tokens": 1169686528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450756, "loss/out": 6.857194900512695, "created_at": "2025-01-14T19:17:46.995259+00:00"} {"global_step": 4463, "acc_step": 0, "speed/wps": 12941.26069841322, "speed/FLOPS": 203260282424320.78, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.4901041984558105, "optim/lr": 0.0026772, "optim/total_tokens": 1169948672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 6.836121559143066, "created_at": "2025-01-14T19:17:52.063813+00:00"} {"global_step": 4464, "acc_step": 0, "speed/wps": 12939.974020956937, "speed/FLOPS": 203240073386789.75, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0008, "optim/grad_norm": 11.817061424255371, "optim/lr": 0.0026777999999999997, "optim/total_tokens": 1170210816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 6.626131534576416, "created_at": "2025-01-14T19:17:57.133047+00:00"} {"global_step": 4465, "acc_step": 0, "speed/wps": 12941.542489462572, "speed/FLOPS": 203264708339972.34, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 66.73738861083984, "optim/lr": 0.0026784, "optim/total_tokens": 1170472960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435567, "loss/out": 6.850266456604004, "created_at": "2025-01-14T19:18:02.201179+00:00"} {"global_step": 4466, "acc_step": 0, "speed/wps": 12930.018238244642, "speed/FLOPS": 203083704138614.06, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.354941368103027, "optim/lr": 0.002679, "optim/total_tokens": 1170735104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455808, "loss/out": 6.940529823303223, "created_at": "2025-01-14T19:18:07.276643+00:00"} {"global_step": 4467, "acc_step": 0, "speed/wps": 12945.19243725808, "speed/FLOPS": 203322035785654.66, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0009, "optim/grad_norm": 3.4083735942840576, "optim/lr": 0.0026796, "optim/total_tokens": 1170997248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475904, "loss/out": 6.784001350402832, "created_at": "2025-01-14T19:18:12.347397+00:00"} {"global_step": 4468, "acc_step": 0, "speed/wps": 12937.302659703128, "speed/FLOPS": 203198115987463.62, "speed/curr_iter_time": 1.2774, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.0125815868377686, "optim/lr": 0.0026801999999999998, "optim/total_tokens": 1171259392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 6.718837738037109, "created_at": "2025-01-14T19:18:17.423204+00:00"} {"global_step": 4469, "acc_step": 0, "speed/wps": 12945.975950252394, "speed/FLOPS": 203334341933890.03, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.343736171722412, "optim/lr": 0.0026808, "optim/total_tokens": 1171521536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306381, "loss/out": 6.690021514892578, "created_at": "2025-01-14T19:18:22.495414+00:00"} {"global_step": 4470, "acc_step": 0, "speed/wps": 12941.149554285385, "speed/FLOPS": 203258536753065.1, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.895511627197266, "optim/lr": 0.0026814, "optim/total_tokens": 1171783680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450746, "loss/out": 6.608983516693115, "created_at": "2025-01-14T19:18:27.565894+00:00"} {"global_step": 4471, "acc_step": 0, "speed/wps": 12935.818383196774, "speed/FLOPS": 203174803385321.7, "speed/curr_iter_time": 1.2779, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.689367294311523, "optim/lr": 0.002682, "optim/total_tokens": 1172045824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431320, "loss/out": 6.7232232093811035, "created_at": "2025-01-14T19:18:32.637490+00:00"} {"global_step": 4472, "acc_step": 0, "speed/wps": 12929.36021912735, "speed/FLOPS": 203073369044163.97, "speed/curr_iter_time": 1.2781, "speed/data_load_time": 0.001, "optim/grad_norm": 10.312296867370605, "optim/lr": 0.0026826, "optim/total_tokens": 1172307968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 6.652951240539551, "created_at": "2025-01-14T19:18:37.712122+00:00"} {"global_step": 4473, "acc_step": 0, "speed/wps": 12911.378990633706, "speed/FLOPS": 202790949141874.06, "speed/curr_iter_time": 1.2778, "speed/data_load_time": 0.0007, "optim/grad_norm": 18.079011917114258, "optim/lr": 0.0026832, "optim/total_tokens": 1172570112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294835, "loss/out": 6.583208084106445, "created_at": "2025-01-14T19:18:42.792846+00:00"} {"global_step": 4474, "acc_step": 0, "speed/wps": 12942.718891023189, "speed/FLOPS": 203283185343026.97, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0006, "optim/grad_norm": 11.978992462158203, "optim/lr": 0.0026838, "optim/total_tokens": 1172832256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458430, "loss/out": 6.781959533691406, "created_at": "2025-01-14T19:18:47.861155+00:00"} {"global_step": 4475, "acc_step": 0, "speed/wps": 12938.118254928206, "speed/FLOPS": 203210926031218.78, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.298783302307129, "optim/lr": 0.0026844, "optim/total_tokens": 1173094400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462938, "loss/out": 6.628671169281006, "created_at": "2025-01-14T19:18:52.931572+00:00"} {"global_step": 4476, "acc_step": 0, "speed/wps": 12938.446360261085, "speed/FLOPS": 203216079376335.62, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0012, "optim/grad_norm": 11.089678764343262, "optim/lr": 0.0026850000000000003, "optim/total_tokens": 1173356544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 6.620466709136963, "created_at": "2025-01-14T19:18:58.000628+00:00"} {"global_step": 4477, "acc_step": 0, "speed/wps": 12940.748092370684, "speed/FLOPS": 203252231242028.25, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.118202209472656, "optim/lr": 0.0026856000000000002, "optim/total_tokens": 1173618688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324181, "loss/out": 6.610785961151123, "created_at": "2025-01-14T19:19:03.069887+00:00"} {"global_step": 4478, "acc_step": 0, "speed/wps": 12945.438326618252, "speed/FLOPS": 203325897815943.53, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.501622200012207, "optim/lr": 0.0026862, "optim/total_tokens": 1173880832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 409734, "loss/out": 6.561521053314209, "created_at": "2025-01-14T19:19:08.138162+00:00"} {"global_step": 4479, "acc_step": 0, "speed/wps": 12943.194626210025, "speed/FLOPS": 203290657417864.88, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.393556594848633, "optim/lr": 0.0026868, "optim/total_tokens": 1174142976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 6.548103332519531, "created_at": "2025-01-14T19:19:13.207043+00:00"} {"global_step": 4480, "acc_step": 0, "speed/wps": 12950.075074153183, "speed/FLOPS": 203398724307538.44, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.309301853179932, "optim/lr": 0.0026874000000000004, "optim/total_tokens": 1174405120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449657, "loss/out": 6.463399410247803, "created_at": "2025-01-14T19:19:18.272460+00:00"} {"global_step": 4481, "acc_step": 0, "speed/wps": 12943.03488376665, "speed/FLOPS": 203288148443282.56, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 24.876873016357422, "optim/lr": 0.0026880000000000003, "optim/total_tokens": 1174667264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464799, "loss/out": 6.655076026916504, "created_at": "2025-01-14T19:19:23.339621+00:00"} {"global_step": 4482, "acc_step": 0, "speed/wps": 12945.73781647715, "speed/FLOPS": 203330601715719.75, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.365306854248047, "optim/lr": 0.0026886, "optim/total_tokens": 1174929408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312011, "loss/out": 6.720712661743164, "created_at": "2025-01-14T19:19:28.407307+00:00"} {"global_step": 4483, "acc_step": 0, "speed/wps": 12945.777732675064, "speed/FLOPS": 203331228654458.56, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.676827430725098, "optim/lr": 0.0026892, "optim/total_tokens": 1175191552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446220, "loss/out": 6.588340759277344, "created_at": "2025-01-14T19:19:33.479425+00:00"} {"global_step": 4484, "acc_step": 0, "speed/wps": 12948.85794285843, "speed/FLOPS": 203379607588040.7, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.875857830047607, "optim/lr": 0.0026898, "optim/total_tokens": 1175453696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443306, "loss/out": 6.604076385498047, "created_at": "2025-01-14T19:19:38.544127+00:00"} {"global_step": 4485, "acc_step": 0, "speed/wps": 12940.866588049601, "speed/FLOPS": 203254092379495.75, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0008, "optim/grad_norm": 13.86555004119873, "optim/lr": 0.0026904000000000003, "optim/total_tokens": 1175715840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 6.526576042175293, "created_at": "2025-01-14T19:19:43.614705+00:00"} {"global_step": 4486, "acc_step": 0, "speed/wps": 12938.751693202765, "speed/FLOPS": 203220875049756.12, "speed/curr_iter_time": 1.2779, "speed/data_load_time": 0.0007, "optim/grad_norm": 20.544878005981445, "optim/lr": 0.0026910000000000002, "optim/total_tokens": 1175977984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 277255, "loss/out": 6.755448818206787, "created_at": "2025-01-14T19:19:48.690606+00:00"} {"global_step": 4487, "acc_step": 0, "speed/wps": 12940.38100085902, "speed/FLOPS": 203246465565400.97, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.91547203063965, "optim/lr": 0.0026916, "optim/total_tokens": 1176240128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 400357, "loss/out": 6.613156795501709, "created_at": "2025-01-14T19:19:53.762211+00:00"} {"global_step": 4488, "acc_step": 0, "speed/wps": 12944.719548435018, "speed/FLOPS": 203314608416871.12, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.817927360534668, "optim/lr": 0.0026922, "optim/total_tokens": 1176502272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 6.551544189453125, "created_at": "2025-01-14T19:19:58.829962+00:00"} {"global_step": 4489, "acc_step": 0, "speed/wps": 12939.310082117589, "speed/FLOPS": 203229645314970.22, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0009, "optim/grad_norm": 40.46426010131836, "optim/lr": 0.0026928, "optim/total_tokens": 1176764416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 6.741339206695557, "created_at": "2025-01-14T19:20:03.902047+00:00"} {"global_step": 4490, "acc_step": 0, "speed/wps": 12943.662615703452, "speed/FLOPS": 203298007835944.1, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 23.428308486938477, "optim/lr": 0.0026934000000000003, "optim/total_tokens": 1177026560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446220, "loss/out": 6.915704727172852, "created_at": "2025-01-14T19:20:08.968856+00:00"} {"global_step": 4491, "acc_step": 0, "speed/wps": 12934.681554697769, "speed/FLOPS": 203156947931578.75, "speed/curr_iter_time": 1.2792, "speed/data_load_time": 0.0011, "optim/grad_norm": 3.9873170852661133, "optim/lr": 0.002694, "optim/total_tokens": 1177288704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369341, "loss/out": 6.743102073669434, "created_at": "2025-01-14T19:20:14.040951+00:00"} {"global_step": 4492, "acc_step": 0, "speed/wps": 12928.155816340548, "speed/FLOPS": 203054452243374.4, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.796196222305298, "optim/lr": 0.0026946, "optim/total_tokens": 1177550848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 6.625269412994385, "created_at": "2025-01-14T19:20:19.119028+00:00"} {"global_step": 4493, "acc_step": 0, "speed/wps": 12943.791600289953, "speed/FLOPS": 203300033716118.5, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.34132981300354, "optim/lr": 0.0026952, "optim/total_tokens": 1177812992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461385, "loss/out": 6.535776138305664, "created_at": "2025-01-14T19:20:24.187621+00:00"} {"global_step": 4494, "acc_step": 0, "speed/wps": 12945.286699576665, "speed/FLOPS": 203323516304898.03, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0007, "optim/grad_norm": 18.283403396606445, "optim/lr": 0.0026958, "optim/total_tokens": 1178075136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436151, "loss/out": 6.675905227661133, "created_at": "2025-01-14T19:20:29.260641+00:00"} {"global_step": 4495, "acc_step": 0, "speed/wps": 12944.054446334763, "speed/FLOPS": 203304162074440.56, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 15.420247077941895, "optim/lr": 0.0026964000000000003, "optim/total_tokens": 1178337280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 407300, "loss/out": 6.817370414733887, "created_at": "2025-01-14T19:20:34.331072+00:00"} {"global_step": 4496, "acc_step": 0, "speed/wps": 12948.266329107148, "speed/FLOPS": 203370315481114.3, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.105227947235107, "optim/lr": 0.002697, "optim/total_tokens": 1178599424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 6.824676036834717, "created_at": "2025-01-14T19:20:39.397142+00:00"} {"global_step": 4497, "acc_step": 0, "speed/wps": 12949.05698392386, "speed/FLOPS": 203382733801484.34, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.482993125915527, "optim/lr": 0.0026976, "optim/total_tokens": 1178861568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437432, "loss/out": 6.605096817016602, "created_at": "2025-01-14T19:20:44.466528+00:00"} {"global_step": 4498, "acc_step": 0, "speed/wps": 12947.304326124728, "speed/FLOPS": 203355205902344.75, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.389194488525391, "optim/lr": 0.0026982, "optim/total_tokens": 1179123712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443016, "loss/out": 6.635856628417969, "created_at": "2025-01-14T19:20:49.533426+00:00"} {"global_step": 4499, "acc_step": 0, "speed/wps": 12952.7276351053, "speed/FLOPS": 203440386422292.16, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.388566493988037, "optim/lr": 0.0026988, "optim/total_tokens": 1179385856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334369, "loss/out": 6.650206565856934, "created_at": "2025-01-14T19:20:54.600221+00:00"} {"global_step": 4500, "acc_step": 0, "speed/wps": 12950.12633511024, "speed/FLOPS": 203399529431308.97, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.268345832824707, "optim/lr": 0.0026994000000000002, "optim/total_tokens": 1179648000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304502, "loss/out": 6.522500991821289, "created_at": "2025-01-14T19:20:59.665366+00:00"} {"global_step": 4501, "acc_step": 0, "speed/wps": 12946.581505797034, "speed/FLOPS": 203343853015839.12, "speed/curr_iter_time": 1.2771, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.7193764448165894, "optim/lr": 0.0027, "optim/total_tokens": 1179910144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450990, "loss/out": 6.585474967956543, "created_at": "2025-01-14T19:21:04.731407+00:00"} {"global_step": 4502, "acc_step": 0, "speed/wps": 12956.649104663476, "speed/FLOPS": 203501978490367.2, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.356642246246338, "optim/lr": 0.0027006, "optim/total_tokens": 1180172288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 6.450335502624512, "created_at": "2025-01-14T19:21:09.796803+00:00"} {"global_step": 4503, "acc_step": 0, "speed/wps": 12946.173254529704, "speed/FLOPS": 203337440868689.66, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.339261531829834, "optim/lr": 0.0027012, "optim/total_tokens": 1180434432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441442, "loss/out": 6.423029899597168, "created_at": "2025-01-14T19:21:14.865495+00:00"} {"global_step": 4504, "acc_step": 0, "speed/wps": 12944.7657091838, "speed/FLOPS": 203315333434862.25, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.6959526538848877, "optim/lr": 0.0027018, "optim/total_tokens": 1180696576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452615, "loss/out": 6.654764175415039, "created_at": "2025-01-14T19:21:19.934447+00:00"} {"global_step": 4505, "acc_step": 0, "speed/wps": 12950.352856726005, "speed/FLOPS": 203403087264557.88, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.428115367889404, "optim/lr": 0.0027024, "optim/total_tokens": 1180958720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322976, "loss/out": 6.465778827667236, "created_at": "2025-01-14T19:21:25.001938+00:00"} {"global_step": 4506, "acc_step": 0, "speed/wps": 12948.335711944877, "speed/FLOPS": 203371405233922.84, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.023167610168457, "optim/lr": 0.002703, "optim/total_tokens": 1181220864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456851, "loss/out": 6.52670955657959, "created_at": "2025-01-14T19:21:30.068972+00:00"} {"global_step": 4507, "acc_step": 0, "speed/wps": 12949.815944298189, "speed/FLOPS": 203394654317084.0, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 11.229286193847656, "optim/lr": 0.0027036, "optim/total_tokens": 1181483008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447839, "loss/out": 6.483759880065918, "created_at": "2025-01-14T19:21:35.133798+00:00"} {"global_step": 4508, "acc_step": 0, "speed/wps": 12949.447987856302, "speed/FLOPS": 203388875055539.03, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.4193124771118164, "optim/lr": 0.0027042, "optim/total_tokens": 1181745152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 6.475277900695801, "created_at": "2025-01-14T19:21:40.199409+00:00"} {"global_step": 4509, "acc_step": 0, "speed/wps": 12949.106971518702, "speed/FLOPS": 203383518925351.5, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.4049072265625, "optim/lr": 0.0027048, "optim/total_tokens": 1182007296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429977, "loss/out": 6.520604610443115, "created_at": "2025-01-14T19:21:45.267146+00:00"} {"global_step": 4510, "acc_step": 0, "speed/wps": 12944.67497629989, "speed/FLOPS": 203313908350240.4, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.083808898925781, "optim/lr": 0.0027054, "optim/total_tokens": 1182269440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366652, "loss/out": 6.6152753829956055, "created_at": "2025-01-14T19:21:50.339125+00:00"} {"global_step": 4511, "acc_step": 0, "speed/wps": 12948.791499503586, "speed/FLOPS": 203378564003850.12, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0006, "optim/grad_norm": 31.601680755615234, "optim/lr": 0.002706, "optim/total_tokens": 1182531584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 6.45535135269165, "created_at": "2025-01-14T19:21:55.404161+00:00"} {"global_step": 4512, "acc_step": 0, "speed/wps": 12953.167162966696, "speed/FLOPS": 203447289811329.94, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 29.22470474243164, "optim/lr": 0.0027066, "optim/total_tokens": 1182793728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439871, "loss/out": 6.572409629821777, "created_at": "2025-01-14T19:22:00.475682+00:00"} {"global_step": 4513, "acc_step": 0, "speed/wps": 12948.828358384984, "speed/FLOPS": 203379142923231.88, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.6568477153778076, "optim/lr": 0.0027072, "optim/total_tokens": 1183055872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475696, "loss/out": 6.734038352966309, "created_at": "2025-01-14T19:22:05.545697+00:00"} {"global_step": 4514, "acc_step": 0, "speed/wps": 12952.903231424585, "speed/FLOPS": 203443144403781.16, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 235.8958740234375, "optim/lr": 0.0027078, "optim/total_tokens": 1183318016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434807, "loss/out": 6.5070953369140625, "created_at": "2025-01-14T19:22:10.610270+00:00"} {"global_step": 4515, "acc_step": 0, "speed/wps": 12956.041488221113, "speed/FLOPS": 203492435039187.38, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.651342391967773, "optim/lr": 0.0027084, "optim/total_tokens": 1183580160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 339173, "loss/out": 7.137685775756836, "created_at": "2025-01-14T19:22:15.672727+00:00"} {"global_step": 4516, "acc_step": 0, "speed/wps": 12954.228522171707, "speed/FLOPS": 203463959916105.03, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.6823570728302, "optim/lr": 0.002709, "optim/total_tokens": 1183842304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440399, "loss/out": 6.7321014404296875, "created_at": "2025-01-14T19:22:20.735411+00:00"} {"global_step": 4517, "acc_step": 0, "speed/wps": 12947.768088730412, "speed/FLOPS": 203362489931343.66, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.6330180168151855, "optim/lr": 0.0027096, "optim/total_tokens": 1184104448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 6.5849385261535645, "created_at": "2025-01-14T19:22:25.808697+00:00"} {"global_step": 4518, "acc_step": 0, "speed/wps": 12947.969328645866, "speed/FLOPS": 203365650680748.47, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6866769790649414, "optim/lr": 0.0027102, "optim/total_tokens": 1184366592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 6.691310882568359, "created_at": "2025-01-14T19:22:30.876802+00:00"} {"global_step": 4519, "acc_step": 0, "speed/wps": 12948.011591037935, "speed/FLOPS": 203366314469690.53, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.900622844696045, "optim/lr": 0.0027107999999999998, "optim/total_tokens": 1184628736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330862, "loss/out": 6.4912214279174805, "created_at": "2025-01-14T19:22:35.942905+00:00"} {"global_step": 4520, "acc_step": 0, "speed/wps": 12945.275994916543, "speed/FLOPS": 203323348173501.06, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0013, "optim/grad_norm": 2.4982051849365234, "optim/lr": 0.0027114, "optim/total_tokens": 1184890880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436080, "loss/out": 6.7822041511535645, "created_at": "2025-01-14T19:22:41.014144+00:00"} {"global_step": 4521, "acc_step": 0, "speed/wps": 12945.305787017898, "speed/FLOPS": 203323816099391.5, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.0732836723327637, "optim/lr": 0.002712, "optim/total_tokens": 1185153024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 6.530636787414551, "created_at": "2025-01-14T19:22:46.085747+00:00"} {"global_step": 4522, "acc_step": 0, "speed/wps": 12953.856368392657, "speed/FLOPS": 203458114729612.16, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5790092945098877, "optim/lr": 0.0027126, "optim/total_tokens": 1185415168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466369, "loss/out": 6.548321723937988, "created_at": "2025-01-14T19:22:51.149327+00:00"} {"global_step": 4523, "acc_step": 0, "speed/wps": 12955.11511219254, "speed/FLOPS": 203477885030684.56, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.14752197265625, "optim/lr": 0.0027132, "optim/total_tokens": 1185677312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426544, "loss/out": 6.44097900390625, "created_at": "2025-01-14T19:22:56.215623+00:00"} {"global_step": 4524, "acc_step": 0, "speed/wps": 12953.944230144976, "speed/FLOPS": 203459494719167.03, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.8252344131469727, "optim/lr": 0.0027137999999999997, "optim/total_tokens": 1185939456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354562, "loss/out": 6.498767375946045, "created_at": "2025-01-14T19:23:01.284723+00:00"} {"global_step": 4525, "acc_step": 0, "speed/wps": 12946.184348905828, "speed/FLOPS": 203337615121111.97, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.2496225833892822, "optim/lr": 0.0027144, "optim/total_tokens": 1186201600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 6.332507133483887, "created_at": "2025-01-14T19:23:06.351844+00:00"} {"global_step": 4526, "acc_step": 0, "speed/wps": 12954.134041352627, "speed/FLOPS": 203462475965011.12, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.466886520385742, "optim/lr": 0.002715, "optim/total_tokens": 1186463744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 6.364121913909912, "created_at": "2025-01-14T19:23:11.416026+00:00"} {"global_step": 4527, "acc_step": 0, "speed/wps": 12947.598451597605, "speed/FLOPS": 203359825547067.06, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.6129353046417236, "optim/lr": 0.0027156, "optim/total_tokens": 1186725888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415046, "loss/out": 6.474499225616455, "created_at": "2025-01-14T19:23:16.482864+00:00"} {"global_step": 4528, "acc_step": 0, "speed/wps": 12952.327354616758, "speed/FLOPS": 203434099467176.56, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.9202299118042, "optim/lr": 0.0027162, "optim/total_tokens": 1186988032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 6.436622619628906, "created_at": "2025-01-14T19:23:21.546233+00:00"} {"global_step": 4529, "acc_step": 0, "speed/wps": 12948.786219329342, "speed/FLOPS": 203378481071457.9, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.714536190032959, "optim/lr": 0.0027168, "optim/total_tokens": 1187250176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324713, "loss/out": 6.381120681762695, "created_at": "2025-01-14T19:23:26.611949+00:00"} {"global_step": 4530, "acc_step": 0, "speed/wps": 12963.310566114045, "speed/FLOPS": 203606605896255.66, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.402620792388916, "optim/lr": 0.0027174, "optim/total_tokens": 1187512320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365596, "loss/out": 6.311244010925293, "created_at": "2025-01-14T19:23:31.675229+00:00"} {"global_step": 4531, "acc_step": 0, "speed/wps": 12949.32791473387, "speed/FLOPS": 203386989142153.5, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.699127674102783, "optim/lr": 0.002718, "optim/total_tokens": 1187774464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468255, "loss/out": 6.544199466705322, "created_at": "2025-01-14T19:23:36.746140+00:00"} {"global_step": 4532, "acc_step": 0, "speed/wps": 12951.546457606593, "speed/FLOPS": 203421834406567.0, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.3057940006256104, "optim/lr": 0.0027186000000000003, "optim/total_tokens": 1188036608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427066, "loss/out": 6.3767266273498535, "created_at": "2025-01-14T19:23:41.814418+00:00"} {"global_step": 4533, "acc_step": 0, "speed/wps": 12949.114658652805, "speed/FLOPS": 203383639662355.8, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0013, "optim/grad_norm": 3.6134281158447266, "optim/lr": 0.0027192, "optim/total_tokens": 1188298752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441687, "loss/out": 6.377097129821777, "created_at": "2025-01-14T19:23:46.884480+00:00"} {"global_step": 4534, "acc_step": 0, "speed/wps": 12960.058578415044, "speed/FLOPS": 203555528960743.44, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.3181097507476807, "optim/lr": 0.0027198, "optim/total_tokens": 1188560896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299996, "loss/out": 6.503301620483398, "created_at": "2025-01-14T19:23:51.949677+00:00"} {"global_step": 4535, "acc_step": 0, "speed/wps": 12952.705941306698, "speed/FLOPS": 203440045691374.72, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.404389381408691, "optim/lr": 0.0027204000000000004, "optim/total_tokens": 1188823040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457095, "loss/out": 6.414287567138672, "created_at": "2025-01-14T19:23:57.013108+00:00"} {"global_step": 4536, "acc_step": 0, "speed/wps": 12950.888700572208, "speed/FLOPS": 203411503428489.1, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.27200174331665, "optim/lr": 0.0027210000000000003, "optim/total_tokens": 1189085184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 6.340220928192139, "created_at": "2025-01-14T19:24:02.081706+00:00"} {"global_step": 4537, "acc_step": 0, "speed/wps": 12947.544864092632, "speed/FLOPS": 203358983881664.5, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.381073236465454, "optim/lr": 0.0027216000000000002, "optim/total_tokens": 1189347328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 6.381396770477295, "created_at": "2025-01-14T19:24:07.148713+00:00"} {"global_step": 4538, "acc_step": 0, "speed/wps": 12950.823599736568, "speed/FLOPS": 203410480930407.44, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.4531049728393555, "optim/lr": 0.0027222, "optim/total_tokens": 1189609472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427066, "loss/out": 6.331484794616699, "created_at": "2025-01-14T19:24:12.215999+00:00"} {"global_step": 4539, "acc_step": 0, "speed/wps": 12948.593667469315, "speed/FLOPS": 203375456779902.2, "speed/curr_iter_time": 1.2771, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.4633893966674805, "optim/lr": 0.0027228, "optim/total_tokens": 1189871616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315310, "loss/out": 6.377586364746094, "created_at": "2025-01-14T19:24:17.285487+00:00"} {"global_step": 4540, "acc_step": 0, "speed/wps": 12951.147301293402, "speed/FLOPS": 203415565108170.06, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 18.60908317565918, "optim/lr": 0.0027234000000000004, "optim/total_tokens": 1190133760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463510, "loss/out": 6.300570487976074, "created_at": "2025-01-14T19:24:22.349566+00:00"} {"global_step": 4541, "acc_step": 0, "speed/wps": 12949.871855108397, "speed/FLOPS": 203395532473187.84, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.97945237159729, "optim/lr": 0.0027240000000000003, "optim/total_tokens": 1190395904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 6.505653381347656, "created_at": "2025-01-14T19:24:27.414062+00:00"} {"global_step": 4542, "acc_step": 0, "speed/wps": 12955.963397196652, "speed/FLOPS": 203491208512339.9, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.587249755859375, "optim/lr": 0.0027246, "optim/total_tokens": 1190658048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440919, "loss/out": 6.617987155914307, "created_at": "2025-01-14T19:24:32.478558+00:00"} {"global_step": 4543, "acc_step": 0, "speed/wps": 12955.048845499101, "speed/FLOPS": 203476844221204.12, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.209432601928711, "optim/lr": 0.0027252, "optim/total_tokens": 1190920192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 6.406391620635986, "created_at": "2025-01-14T19:24:37.542160+00:00"} {"global_step": 4544, "acc_step": 0, "speed/wps": 12951.821547556729, "speed/FLOPS": 203426155072247.06, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9241803884506226, "optim/lr": 0.0027258, "optim/total_tokens": 1191182336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310007, "loss/out": 6.594974517822266, "created_at": "2025-01-14T19:24:42.608369+00:00"} {"global_step": 4545, "acc_step": 0, "speed/wps": 12948.394588229565, "speed/FLOPS": 203372329966878.38, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0009, "optim/grad_norm": 5.076080322265625, "optim/lr": 0.0027264000000000004, "optim/total_tokens": 1191444480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450462, "loss/out": 6.529837608337402, "created_at": "2025-01-14T19:24:47.680989+00:00"} {"global_step": 4546, "acc_step": 0, "speed/wps": 12957.169410606195, "speed/FLOPS": 203510150610172.88, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.6536176204681396, "optim/lr": 0.0027270000000000003, "optim/total_tokens": 1191706624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461889, "loss/out": 6.7191972732543945, "created_at": "2025-01-14T19:24:52.742471+00:00"} {"global_step": 4547, "acc_step": 0, "speed/wps": 12950.608273249405, "speed/FLOPS": 203407098932036.94, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.993712902069092, "optim/lr": 0.0027276, "optim/total_tokens": 1191968768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464007, "loss/out": 6.491402626037598, "created_at": "2025-01-14T19:24:57.806579+00:00"} {"global_step": 4548, "acc_step": 0, "speed/wps": 12953.446806945696, "speed/FLOPS": 203451682004291.25, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.999483346939087, "optim/lr": 0.0027282, "optim/total_tokens": 1192230912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453375, "loss/out": 6.540437698364258, "created_at": "2025-01-14T19:25:02.871234+00:00"} {"global_step": 4549, "acc_step": 0, "speed/wps": 12951.898658940676, "speed/FLOPS": 203427366212494.1, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.574442386627197, "optim/lr": 0.0027288, "optim/total_tokens": 1192493056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348127, "loss/out": 6.617465972900391, "created_at": "2025-01-14T19:25:07.942256+00:00"} {"global_step": 4550, "acc_step": 0, "speed/wps": 12958.38847721345, "speed/FLOPS": 203529297726412.88, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.771604061126709, "optim/lr": 0.0027294000000000003, "optim/total_tokens": 1192755200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448890, "loss/out": 6.512476921081543, "created_at": "2025-01-14T19:25:13.009023+00:00"} {"global_step": 4551, "acc_step": 0, "speed/wps": 12952.27982870359, "speed/FLOPS": 203433353007402.8, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.642680048942566, "optim/lr": 0.0027300000000000002, "optim/total_tokens": 1193017344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442784, "loss/out": 6.365814685821533, "created_at": "2025-01-14T19:25:18.079199+00:00"} {"global_step": 4552, "acc_step": 0, "speed/wps": 12953.207279976796, "speed/FLOPS": 203447919904100.12, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.684320449829102, "optim/lr": 0.0027306, "optim/total_tokens": 1193279488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 6.467061996459961, "created_at": "2025-01-14T19:25:23.146948+00:00"} {"global_step": 4553, "acc_step": 0, "speed/wps": 12958.38783768928, "speed/FLOPS": 203529287681807.0, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.112834930419922, "optim/lr": 0.0027312, "optim/total_tokens": 1193541632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446505, "loss/out": 6.4334330558776855, "created_at": "2025-01-14T19:25:28.209094+00:00"} {"global_step": 4554, "acc_step": 0, "speed/wps": 12962.778517444382, "speed/FLOPS": 203598249340787.1, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.550304412841797, "optim/lr": 0.0027318, "optim/total_tokens": 1193803776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306094, "loss/out": 6.3298540115356445, "created_at": "2025-01-14T19:25:33.270103+00:00"} {"global_step": 4555, "acc_step": 0, "speed/wps": 12958.023576888556, "speed/FLOPS": 203523566465384.8, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.8842456340789795, "optim/lr": 0.0027324000000000003, "optim/total_tokens": 1194065920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419875, "loss/out": 6.414790630340576, "created_at": "2025-01-14T19:25:38.333970+00:00"} {"global_step": 4556, "acc_step": 0, "speed/wps": 12958.79893827905, "speed/FLOPS": 203535744581480.72, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.079632759094238, "optim/lr": 0.002733, "optim/total_tokens": 1194328064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 6.381699562072754, "created_at": "2025-01-14T19:25:43.398595+00:00"} {"global_step": 4557, "acc_step": 0, "speed/wps": 12952.388420026638, "speed/FLOPS": 203435058583351.4, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.797054290771484, "optim/lr": 0.0027336, "optim/total_tokens": 1194590208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435852, "loss/out": 6.369322776794434, "created_at": "2025-01-14T19:25:48.466115+00:00"} {"global_step": 4558, "acc_step": 0, "speed/wps": 12960.967702395612, "speed/FLOPS": 203569808002125.53, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.553614616394043, "optim/lr": 0.0027342, "optim/total_tokens": 1194852352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453131, "loss/out": 6.264705181121826, "created_at": "2025-01-14T19:25:53.530571+00:00"} {"global_step": 4559, "acc_step": 0, "speed/wps": 12959.28694214812, "speed/FLOPS": 203543409352832.03, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.222771167755127, "optim/lr": 0.0027348, "optim/total_tokens": 1195114496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293306, "loss/out": 6.440868377685547, "created_at": "2025-01-14T19:25:58.595638+00:00"} {"global_step": 4560, "acc_step": 0, "speed/wps": 12955.827524349872, "speed/FLOPS": 203489074442572.38, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.4989876747131348, "optim/lr": 0.0027354000000000002, "optim/total_tokens": 1195376640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447026, "loss/out": 6.410274982452393, "created_at": "2025-01-14T19:26:03.661919+00:00"} {"global_step": 4561, "acc_step": 0, "speed/wps": 12951.726172150231, "speed/FLOPS": 203424657070428.47, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.004279613494873, "optim/lr": 0.002736, "optim/total_tokens": 1195638784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440108, "loss/out": 6.271936416625977, "created_at": "2025-01-14T19:26:08.728557+00:00"} {"global_step": 4562, "acc_step": 0, "speed/wps": 12942.086306765099, "speed/FLOPS": 203273249737991.8, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.896096706390381, "optim/lr": 0.0027366, "optim/total_tokens": 1195900928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 6.247089862823486, "created_at": "2025-01-14T19:26:13.800931+00:00"} {"global_step": 4563, "acc_step": 0, "speed/wps": 12961.31173109584, "speed/FLOPS": 203575211445607.6, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.118985414505005, "optim/lr": 0.0027372, "optim/total_tokens": 1196163072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318033, "loss/out": 6.369100570678711, "created_at": "2025-01-14T19:26:18.860659+00:00"} {"global_step": 4564, "acc_step": 0, "speed/wps": 12954.624534349152, "speed/FLOPS": 203470179831529.56, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.619300365447998, "optim/lr": 0.0027378, "optim/total_tokens": 1196425216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 408061, "loss/out": 6.325819969177246, "created_at": "2025-01-14T19:26:23.923531+00:00"} {"global_step": 4565, "acc_step": 0, "speed/wps": 12956.838429733689, "speed/FLOPS": 203504952100754.78, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.076883792877197, "optim/lr": 0.0027383999999999998, "optim/total_tokens": 1196687360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451797, "loss/out": 6.354034423828125, "created_at": "2025-01-14T19:26:28.985745+00:00"} {"global_step": 4566, "acc_step": 0, "speed/wps": 12954.02108303028, "speed/FLOPS": 203460701799337.7, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.965318202972412, "optim/lr": 0.002739, "optim/total_tokens": 1196949504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443306, "loss/out": 6.351260662078857, "created_at": "2025-01-14T19:26:34.054187+00:00"} {"global_step": 4567, "acc_step": 0, "speed/wps": 12962.653953761757, "speed/FLOPS": 203596292896983.03, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.0906825065612793, "optim/lr": 0.0027396, "optim/total_tokens": 1197211648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 6.358788013458252, "created_at": "2025-01-14T19:26:39.114416+00:00"} {"global_step": 4568, "acc_step": 0, "speed/wps": 12953.32234623932, "speed/FLOPS": 203449727177870.66, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7618353366851807, "optim/lr": 0.0027402, "optim/total_tokens": 1197473792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418774, "loss/out": 6.234667778015137, "created_at": "2025-01-14T19:26:44.177460+00:00"} {"global_step": 4569, "acc_step": 0, "speed/wps": 12958.349790887301, "speed/FLOPS": 203528690104500.12, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.812347412109375, "optim/lr": 0.0027408, "optim/total_tokens": 1197735936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 6.2313337326049805, "created_at": "2025-01-14T19:26:49.243252+00:00"} {"global_step": 4570, "acc_step": 0, "speed/wps": 12946.685261457782, "speed/FLOPS": 203345482641066.38, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.213321208953857, "optim/lr": 0.0027413999999999997, "optim/total_tokens": 1197998080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314415, "loss/out": 6.345217227935791, "created_at": "2025-01-14T19:26:54.312348+00:00"} {"global_step": 4571, "acc_step": 0, "speed/wps": 12950.361057755063, "speed/FLOPS": 203403216072988.6, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.659693717956543, "optim/lr": 0.002742, "optim/total_tokens": 1198260224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444357, "loss/out": 6.295925140380859, "created_at": "2025-01-14T19:26:59.376800+00:00"} {"global_step": 4572, "acc_step": 0, "speed/wps": 12957.092789956225, "speed/FLOPS": 203508947177577.28, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.847346305847168, "optim/lr": 0.0027426, "optim/total_tokens": 1198522368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 6.156356334686279, "created_at": "2025-01-14T19:27:04.442040+00:00"} {"global_step": 4573, "acc_step": 0, "speed/wps": 12950.5175299451, "speed/FLOPS": 203405673683748.47, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0009, "optim/grad_norm": 4.183473587036133, "optim/lr": 0.0027432, "optim/total_tokens": 1198784512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461100, "loss/out": 6.321705341339111, "created_at": "2025-01-14T19:27:09.507037+00:00"} {"global_step": 4574, "acc_step": 0, "speed/wps": 12952.82186652368, "speed/FLOPS": 203441866456205.34, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.194828987121582, "optim/lr": 0.0027438, "optim/total_tokens": 1199046656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442208, "loss/out": 6.1907477378845215, "created_at": "2025-01-14T19:27:14.575164+00:00"} {"global_step": 4575, "acc_step": 0, "speed/wps": 12953.616287373816, "speed/FLOPS": 203454343927305.38, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.5660979747772217, "optim/lr": 0.0027443999999999997, "optim/total_tokens": 1199308800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356429, "loss/out": 6.298200607299805, "created_at": "2025-01-14T19:27:19.637957+00:00"} {"global_step": 4576, "acc_step": 0, "speed/wps": 12951.814236242757, "speed/FLOPS": 203426040238014.34, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7200257778167725, "optim/lr": 0.002745, "optim/total_tokens": 1199570944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454182, "loss/out": 6.310307502746582, "created_at": "2025-01-14T19:27:24.701974+00:00"} {"global_step": 4577, "acc_step": 0, "speed/wps": 12955.954132343611, "speed/FLOPS": 203491062995091.66, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.231808662414551, "optim/lr": 0.0027456, "optim/total_tokens": 1199833088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 6.322673797607422, "created_at": "2025-01-14T19:27:29.764130+00:00"} {"global_step": 4578, "acc_step": 0, "speed/wps": 12950.494408755405, "speed/FLOPS": 203405310533692.47, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.9415178298950195, "optim/lr": 0.0027462, "optim/total_tokens": 1200095232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411640, "loss/out": 6.226063251495361, "created_at": "2025-01-14T19:27:34.828986+00:00"} {"global_step": 4579, "acc_step": 0, "speed/wps": 12946.92034293694, "speed/FLOPS": 203349174918732.38, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.3737618923187256, "optim/lr": 0.0027467999999999998, "optim/total_tokens": 1200357376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475458, "loss/out": 6.437479019165039, "created_at": "2025-01-14T19:27:39.894518+00:00"} {"global_step": 4580, "acc_step": 0, "speed/wps": 12950.022526671568, "speed/FLOPS": 203397898977131.88, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0011, "optim/grad_norm": 4.535326957702637, "optim/lr": 0.0027474, "optim/total_tokens": 1200619520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463222, "loss/out": 6.202934265136719, "created_at": "2025-01-14T19:27:44.958714+00:00"} {"global_step": 4581, "acc_step": 0, "speed/wps": 12952.04109106582, "speed/FLOPS": 203429603304741.72, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.52711820602417, "optim/lr": 0.002748, "optim/total_tokens": 1200881664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334502, "loss/out": 6.414265155792236, "created_at": "2025-01-14T19:27:50.024430+00:00"} {"global_step": 4582, "acc_step": 0, "speed/wps": 12949.83099138662, "speed/FLOPS": 203394890652284.7, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.761948823928833, "optim/lr": 0.0027486, "optim/total_tokens": 1201143808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 408396, "loss/out": 6.253509521484375, "created_at": "2025-01-14T19:27:55.093437+00:00"} {"global_step": 4583, "acc_step": 0, "speed/wps": 12957.312934508765, "speed/FLOPS": 203512404850285.25, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.287570476531982, "optim/lr": 0.0027492, "optim/total_tokens": 1201405952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 6.299737453460693, "created_at": "2025-01-14T19:28:00.158257+00:00"} {"global_step": 4584, "acc_step": 0, "speed/wps": 12959.978954324739, "speed/FLOPS": 203554278354990.25, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.373536586761475, "optim/lr": 0.0027498, "optim/total_tokens": 1201668096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423627, "loss/out": 6.2724456787109375, "created_at": "2025-01-14T19:28:05.221191+00:00"} {"global_step": 4585, "acc_step": 0, "speed/wps": 12956.60270167273, "speed/FLOPS": 203501249667633.1, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.537508964538574, "optim/lr": 0.0027504, "optim/total_tokens": 1201930240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479622, "loss/out": 6.272679328918457, "created_at": "2025-01-14T19:28:10.283054+00:00"} {"global_step": 4586, "acc_step": 0, "speed/wps": 12935.70642891349, "speed/FLOPS": 203173044989461.4, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0009, "optim/grad_norm": 4.783749103546143, "optim/lr": 0.002751, "optim/total_tokens": 1202192384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320967, "loss/out": 6.485121726989746, "created_at": "2025-01-14T19:28:15.354781+00:00"} {"global_step": 4587, "acc_step": 0, "speed/wps": 12956.478566777747, "speed/FLOPS": 203499299958528.2, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.849879741668701, "optim/lr": 0.0027516000000000003, "optim/total_tokens": 1202454528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411271, "loss/out": 6.24071741104126, "created_at": "2025-01-14T19:28:20.418688+00:00"} {"global_step": 4588, "acc_step": 0, "speed/wps": 12949.595649299283, "speed/FLOPS": 203391194281409.16, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.639911413192749, "optim/lr": 0.0027522000000000002, "optim/total_tokens": 1202716672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 6.380317211151123, "created_at": "2025-01-14T19:28:25.484728+00:00"} {"global_step": 4589, "acc_step": 0, "speed/wps": 12951.539558994104, "speed/FLOPS": 203421726054378.16, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.385019779205322, "optim/lr": 0.0027528, "optim/total_tokens": 1202978816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425730, "loss/out": 6.396999359130859, "created_at": "2025-01-14T19:28:30.557796+00:00"} {"global_step": 4590, "acc_step": 0, "speed/wps": 12950.67855114552, "speed/FLOPS": 203408202742967.38, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.059563159942627, "optim/lr": 0.0027534, "optim/total_tokens": 1203240960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436151, "loss/out": 6.36171817779541, "created_at": "2025-01-14T19:28:35.623207+00:00"} {"global_step": 4591, "acc_step": 0, "speed/wps": 12951.81372371209, "speed/FLOPS": 203426032188015.94, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 48.288726806640625, "optim/lr": 0.0027540000000000004, "optim/total_tokens": 1203503104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296714, "loss/out": 6.337488174438477, "created_at": "2025-01-14T19:28:40.692332+00:00"} {"global_step": 4592, "acc_step": 0, "speed/wps": 12951.709218364631, "speed/FLOPS": 203424390787928.7, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.8248231410980225, "optim/lr": 0.0027546000000000003, "optim/total_tokens": 1203765248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446747, "loss/out": 6.51174259185791, "created_at": "2025-01-14T19:28:45.759374+00:00"} {"global_step": 4593, "acc_step": 0, "speed/wps": 12951.545198978749, "speed/FLOPS": 203421814638087.16, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.2356796264648438, "optim/lr": 0.0027552, "optim/total_tokens": 1204027392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453131, "loss/out": 6.52005672454834, "created_at": "2025-01-14T19:28:50.828389+00:00"} {"global_step": 4594, "acc_step": 0, "speed/wps": 12954.380396913415, "speed/FLOPS": 203466345317621.3, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.546003341674805, "optim/lr": 0.0027558, "optim/total_tokens": 1204289536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417752, "loss/out": 6.439690589904785, "created_at": "2025-01-14T19:28:55.896490+00:00"} {"global_step": 4595, "acc_step": 0, "speed/wps": 12960.929411269646, "speed/FLOPS": 203569206587374.72, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.680292129516602, "optim/lr": 0.0027564, "optim/total_tokens": 1204551680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 6.369723796844482, "created_at": "2025-01-14T19:29:00.968078+00:00"} {"global_step": 4596, "acc_step": 0, "speed/wps": 12957.965872881412, "speed/FLOPS": 203522660144659.0, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.825927734375, "optim/lr": 0.0027570000000000003, "optim/total_tokens": 1204813824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308786, "loss/out": 6.475408554077148, "created_at": "2025-01-14T19:29:06.033282+00:00"} {"global_step": 4597, "acc_step": 0, "speed/wps": 12957.98016819418, "speed/FLOPS": 203522884672189.84, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.745891571044922, "optim/lr": 0.0027576000000000002, "optim/total_tokens": 1205075968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413141, "loss/out": 6.373573303222656, "created_at": "2025-01-14T19:29:11.094636+00:00"} {"global_step": 4598, "acc_step": 0, "speed/wps": 12960.023503022225, "speed/FLOPS": 203554978053500.56, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.320850133895874, "optim/lr": 0.0027582, "optim/total_tokens": 1205338112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 6.380587577819824, "created_at": "2025-01-14T19:29:16.159634+00:00"} {"global_step": 4599, "acc_step": 0, "speed/wps": 12958.432427947404, "speed/FLOPS": 203529988033084.25, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0014, "optim/grad_norm": 6.555689811706543, "optim/lr": 0.0027588, "optim/total_tokens": 1205600256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436679, "loss/out": 6.298360824584961, "created_at": "2025-01-14T19:29:21.221799+00:00"} {"global_step": 4600, "acc_step": 0, "speed/wps": 12967.190092627412, "speed/FLOPS": 203667539191176.16, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.2420268058776855, "optim/lr": 0.0027594, "optim/total_tokens": 1205862400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447263, "loss/out": 6.5018744468688965, "created_at": "2025-01-14T19:29:26.282904+00:00"} {"global_step": 4601, "acc_step": 0, "speed/wps": 12959.38753398852, "speed/FLOPS": 203544989285913.22, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.7147388458251953, "optim/lr": 0.0027600000000000003, "optim/total_tokens": 1206124544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435329, "loss/out": 6.323522567749023, "created_at": "2025-01-14T19:29:31.346511+00:00"} {"global_step": 4602, "acc_step": 0, "speed/wps": 12959.673043560057, "speed/FLOPS": 203549473606065.47, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9579722881317139, "optim/lr": 0.0027606, "optim/total_tokens": 1206386688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359599, "loss/out": 6.4038567543029785, "created_at": "2025-01-14T19:29:36.413031+00:00"} {"global_step": 4603, "acc_step": 0, "speed/wps": 12959.853575635947, "speed/FLOPS": 203552309110394.38, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.072140693664551, "optim/lr": 0.0027612, "optim/total_tokens": 1206648832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 6.425839424133301, "created_at": "2025-01-14T19:29:41.476528+00:00"} {"global_step": 4604, "acc_step": 0, "speed/wps": 12960.657765621787, "speed/FLOPS": 203564940019194.84, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.696187973022461, "optim/lr": 0.0027618, "optim/total_tokens": 1206910976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459743, "loss/out": 6.4681782722473145, "created_at": "2025-01-14T19:29:46.537308+00:00"} {"global_step": 4605, "acc_step": 0, "speed/wps": 12956.988361632759, "speed/FLOPS": 203507306987257.7, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0009, "optim/grad_norm": 3.168200969696045, "optim/lr": 0.0027624, "optim/total_tokens": 1207173120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426829, "loss/out": 6.362627029418945, "created_at": "2025-01-14T19:29:51.604199+00:00"} {"global_step": 4606, "acc_step": 0, "speed/wps": 12951.522863142738, "speed/FLOPS": 203421463823089.94, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3022680282592773, "optim/lr": 0.0027630000000000003, "optim/total_tokens": 1207435264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 6.476478576660156, "created_at": "2025-01-14T19:29:56.672540+00:00"} {"global_step": 4607, "acc_step": 0, "speed/wps": 12957.146155910534, "speed/FLOPS": 203509785363222.2, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 27.26074981689453, "optim/lr": 0.0027636, "optim/total_tokens": 1207697408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432941, "loss/out": 6.451973915100098, "created_at": "2025-01-14T19:30:01.735301+00:00"} {"global_step": 4608, "acc_step": 0, "speed/wps": 12957.414650494253, "speed/FLOPS": 203514002439609.72, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.179758071899414, "optim/lr": 0.0027642, "optim/total_tokens": 1207959552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337829, "loss/out": 6.485669136047363, "created_at": "2025-01-14T19:30:06.796639+00:00"} {"global_step": 4609, "acc_step": 0, "speed/wps": 12959.205704660844, "speed/FLOPS": 203542133406462.38, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.8658642768859863, "optim/lr": 0.0027648, "optim/total_tokens": 1208221696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455517, "loss/out": 6.421278953552246, "created_at": "2025-01-14T19:30:11.858636+00:00"} {"global_step": 4610, "acc_step": 0, "speed/wps": 12951.653394191364, "speed/FLOPS": 203423513992577.1, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.202960968017578, "optim/lr": 0.0027654, "optim/total_tokens": 1208483840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 6.418558120727539, "created_at": "2025-01-14T19:30:16.923804+00:00"} {"global_step": 4611, "acc_step": 0, "speed/wps": 12960.70516684069, "speed/FLOPS": 203565684520473.8, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.358405590057373, "optim/lr": 0.0027660000000000002, "optim/total_tokens": 1208745984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428688, "loss/out": 6.428155899047852, "created_at": "2025-01-14T19:30:21.989391+00:00"} {"global_step": 4612, "acc_step": 0, "speed/wps": 12963.144375764625, "speed/FLOPS": 203603995648448.53, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.14454984664917, "optim/lr": 0.0027666, "optim/total_tokens": 1209008128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 6.377041339874268, "created_at": "2025-01-14T19:30:27.048601+00:00"} {"global_step": 4613, "acc_step": 0, "speed/wps": 12957.655051283065, "speed/FLOPS": 203517778264343.34, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.8515498638153076, "optim/lr": 0.0027672, "optim/total_tokens": 1209270272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326583, "loss/out": 6.398543357849121, "created_at": "2025-01-14T19:30:32.109799+00:00"} {"global_step": 4614, "acc_step": 0, "speed/wps": 12954.515242537807, "speed/FLOPS": 203468463253449.72, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5929460525512695, "optim/lr": 0.0027678, "optim/total_tokens": 1209532416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418774, "loss/out": 6.399769306182861, "created_at": "2025-01-14T19:30:37.173576+00:00"} {"global_step": 4615, "acc_step": 0, "speed/wps": 12955.18911684366, "speed/FLOPS": 203479047375423.44, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0429604053497314, "optim/lr": 0.0027684, "optim/total_tokens": 1209794560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468018, "loss/out": 6.201383113861084, "created_at": "2025-01-14T19:30:42.240514+00:00"} {"global_step": 4616, "acc_step": 0, "speed/wps": 12959.071766103316, "speed/FLOPS": 203540029717365.53, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.9035435914993286, "optim/lr": 0.002769, "optim/total_tokens": 1210056704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428118, "loss/out": 6.21811580657959, "created_at": "2025-01-14T19:30:47.320644+00:00"} {"global_step": 4617, "acc_step": 0, "speed/wps": 12960.232929445616, "speed/FLOPS": 203558267383262.0, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.434110164642334, "optim/lr": 0.0027696, "optim/total_tokens": 1210318848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440390, "loss/out": 6.338879108428955, "created_at": "2025-01-14T19:30:52.386236+00:00"} {"global_step": 4618, "acc_step": 0, "speed/wps": 12963.659603123004, "speed/FLOPS": 203612088002108.44, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9959580898284912, "optim/lr": 0.0027702, "optim/total_tokens": 1210580992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293306, "loss/out": 6.346175193786621, "created_at": "2025-01-14T19:30:57.447622+00:00"} {"global_step": 4619, "acc_step": 0, "speed/wps": 12962.429035997146, "speed/FLOPS": 203592760254420.4, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9250606298446655, "optim/lr": 0.0027708, "optim/total_tokens": 1210843136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336239, "loss/out": 6.18245267868042, "created_at": "2025-01-14T19:31:02.514445+00:00"} {"global_step": 4620, "acc_step": 0, "speed/wps": 12967.80687562597, "speed/FLOPS": 203677226615717.2, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.773866891860962, "optim/lr": 0.0027714, "optim/total_tokens": 1211105280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438240, "loss/out": 6.349725723266602, "created_at": "2025-01-14T19:31:07.577541+00:00"} {"global_step": 4621, "acc_step": 0, "speed/wps": 12956.348117367088, "speed/FLOPS": 203497251071277.5, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0009, "optim/grad_norm": 3.970717430114746, "optim/lr": 0.002772, "optim/total_tokens": 1211367424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 6.212424278259277, "created_at": "2025-01-14T19:31:12.643218+00:00"} {"global_step": 4622, "acc_step": 0, "speed/wps": 12956.160249830778, "speed/FLOPS": 203494300353462.72, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.8173747062683105, "optim/lr": 0.0027726, "optim/total_tokens": 1211629568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417230, "loss/out": 6.2364373207092285, "created_at": "2025-01-14T19:31:17.711426+00:00"} {"global_step": 4623, "acc_step": 0, "speed/wps": 12956.854929349181, "speed/FLOPS": 203505211249888.94, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.446046829223633, "optim/lr": 0.0027732, "optim/total_tokens": 1211891712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458410, "loss/out": 6.149248123168945, "created_at": "2025-01-14T19:31:22.777928+00:00"} {"global_step": 4624, "acc_step": 0, "speed/wps": 12961.97226401205, "speed/FLOPS": 203585586022724.03, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.8900632858276367, "optim/lr": 0.0027738, "optim/total_tokens": 1212153856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298593, "loss/out": 6.340828895568848, "created_at": "2025-01-14T19:31:27.842848+00:00"} {"global_step": 4625, "acc_step": 0, "speed/wps": 12963.053282373881, "speed/FLOPS": 203602564901571.47, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.196258068084717, "optim/lr": 0.0027743999999999998, "optim/total_tokens": 1212416000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 6.150662422180176, "created_at": "2025-01-14T19:31:32.902887+00:00"} {"global_step": 4626, "acc_step": 0, "speed/wps": 12958.720384457538, "speed/FLOPS": 203534510785769.62, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0009, "optim/grad_norm": 3.125317096710205, "optim/lr": 0.002775, "optim/total_tokens": 1212678144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 6.207787990570068, "created_at": "2025-01-14T19:31:37.965208+00:00"} {"global_step": 4627, "acc_step": 0, "speed/wps": 12953.40979300643, "speed/FLOPS": 203451100649512.6, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.812172532081604, "optim/lr": 0.0027756, "optim/total_tokens": 1212940288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437432, "loss/out": 6.285650253295898, "created_at": "2025-01-14T19:31:43.029296+00:00"} {"global_step": 4628, "acc_step": 0, "speed/wps": 12957.055862538275, "speed/FLOPS": 203508367181734.78, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.6933822631835938, "optim/lr": 0.0027762, "optim/total_tokens": 1213202432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 6.33518123626709, "created_at": "2025-01-14T19:31:48.091446+00:00"} {"global_step": 4629, "acc_step": 0, "speed/wps": 12957.238139730396, "speed/FLOPS": 203511230095512.16, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.632072925567627, "optim/lr": 0.0027768, "optim/total_tokens": 1213464576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430506, "loss/out": 6.180792808532715, "created_at": "2025-01-14T19:31:53.158123+00:00"} {"global_step": 4630, "acc_step": 0, "speed/wps": 12951.60594950516, "speed/FLOPS": 203422768808584.4, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0435171127319336, "optim/lr": 0.0027773999999999997, "optim/total_tokens": 1213726720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360518, "loss/out": 6.182846546173096, "created_at": "2025-01-14T19:31:58.226864+00:00"} {"global_step": 4631, "acc_step": 0, "speed/wps": 12960.411035189267, "speed/FLOPS": 203561064778708.97, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0688037872314453, "optim/lr": 0.002778, "optim/total_tokens": 1213988864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436151, "loss/out": 6.128960609436035, "created_at": "2025-01-14T19:32:03.288083+00:00"} {"global_step": 4632, "acc_step": 0, "speed/wps": 12957.23914179196, "speed/FLOPS": 203511245834266.0, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.7653932571411133, "optim/lr": 0.0027786, "optim/total_tokens": 1214251008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 6.2259697914123535, "created_at": "2025-01-14T19:32:08.350732+00:00"} {"global_step": 4633, "acc_step": 0, "speed/wps": 12960.211823946998, "speed/FLOPS": 203557935892404.22, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.318899631500244, "optim/lr": 0.0027792, "optim/total_tokens": 1214513152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495573, "loss/out": 6.185006618499756, "created_at": "2025-01-14T19:32:13.415430+00:00"} {"global_step": 4634, "acc_step": 0, "speed/wps": 12931.362874647613, "speed/FLOPS": 203104823501046.78, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0014, "optim/grad_norm": 1.6173399686813354, "optim/lr": 0.0027798, "optim/total_tokens": 1214775296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445933, "loss/out": 6.243347644805908, "created_at": "2025-01-14T19:32:18.488935+00:00"} {"global_step": 4635, "acc_step": 0, "speed/wps": 12965.093258537107, "speed/FLOPS": 203634605530435.9, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.748755693435669, "optim/lr": 0.0027804, "optim/total_tokens": 1215037440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456045, "loss/out": 6.259654521942139, "created_at": "2025-01-14T19:32:23.547856+00:00"} {"global_step": 4636, "acc_step": 0, "speed/wps": 12955.12771973192, "speed/FLOPS": 203478083049415.1, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7912765741348267, "optim/lr": 0.002781, "optim/total_tokens": 1215299584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326058, "loss/out": 6.0855793952941895, "created_at": "2025-01-14T19:32:28.610956+00:00"} {"global_step": 4637, "acc_step": 0, "speed/wps": 12971.112217697133, "speed/FLOPS": 203729141554960.75, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3237788677215576, "optim/lr": 0.0027816, "optim/total_tokens": 1215561728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353744, "loss/out": 6.209985256195068, "created_at": "2025-01-14T19:32:33.667081+00:00"} {"global_step": 4638, "acc_step": 0, "speed/wps": 12971.543575486648, "speed/FLOPS": 203735916621792.8, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.190687656402588, "optim/lr": 0.0027822, "optim/total_tokens": 1215823872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 6.195215702056885, "created_at": "2025-01-14T19:32:38.725916+00:00"} {"global_step": 4639, "acc_step": 0, "speed/wps": 12960.986552025512, "speed/FLOPS": 203570104061465.22, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.7683725357055664, "optim/lr": 0.0027828, "optim/total_tokens": 1216086016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 6.193681716918945, "created_at": "2025-01-14T19:32:43.790361+00:00"} {"global_step": 4640, "acc_step": 0, "speed/wps": 12964.330191988256, "speed/FLOPS": 203622620521722.9, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.382885456085205, "optim/lr": 0.0027834, "optim/total_tokens": 1216348160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447026, "loss/out": 6.181906223297119, "created_at": "2025-01-14T19:32:48.854390+00:00"} {"global_step": 4641, "acc_step": 0, "speed/wps": 12970.419685497065, "speed/FLOPS": 203718264385119.72, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5412931442260742, "optim/lr": 0.002784, "optim/total_tokens": 1216610304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 6.1842474937438965, "created_at": "2025-01-14T19:32:53.915815+00:00"} {"global_step": 4642, "acc_step": 0, "speed/wps": 12970.65917765202, "speed/FLOPS": 203722025938510.0, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.9439737796783447, "optim/lr": 0.0027846, "optim/total_tokens": 1216872448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448606, "loss/out": 6.284202575683594, "created_at": "2025-01-14T19:32:58.971975+00:00"} {"global_step": 4643, "acc_step": 0, "speed/wps": 12970.864811725392, "speed/FLOPS": 203725255704203.12, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0012, "optim/grad_norm": 2.473362445831299, "optim/lr": 0.0027852000000000003, "optim/total_tokens": 1217134592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326966, "loss/out": 6.237156391143799, "created_at": "2025-01-14T19:33:04.028447+00:00"} {"global_step": 4644, "acc_step": 0, "speed/wps": 12973.98020427738, "speed/FLOPS": 203774187225230.1, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8706454038619995, "optim/lr": 0.0027858, "optim/total_tokens": 1217396736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314655, "loss/out": 6.182592391967773, "created_at": "2025-01-14T19:33:09.083608+00:00"} {"global_step": 4645, "acc_step": 0, "speed/wps": 12973.64767036805, "speed/FLOPS": 203768964323234.56, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7994730472564697, "optim/lr": 0.0027864, "optim/total_tokens": 1217658880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337166, "loss/out": 6.254194259643555, "created_at": "2025-01-14T19:33:14.139635+00:00"} {"global_step": 4646, "acc_step": 0, "speed/wps": 12962.000903390406, "speed/FLOPS": 203586035843515.72, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.215566635131836, "optim/lr": 0.0027870000000000004, "optim/total_tokens": 1217921024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358407, "loss/out": 6.173415184020996, "created_at": "2025-01-14T19:33:19.199930+00:00"} {"global_step": 4647, "acc_step": 0, "speed/wps": 12968.724905015604, "speed/FLOPS": 203691645528785.94, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1274406909942627, "optim/lr": 0.0027876000000000003, "optim/total_tokens": 1218183168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453660, "loss/out": 6.206776142120361, "created_at": "2025-01-14T19:33:24.258046+00:00"} {"global_step": 4648, "acc_step": 0, "speed/wps": 12961.818005113602, "speed/FLOPS": 203583163174749.94, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0009, "optim/grad_norm": 3.131960391998291, "optim/lr": 0.0027882000000000002, "optim/total_tokens": 1218445312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 6.150959014892578, "created_at": "2025-01-14T19:33:29.322268+00:00"} {"global_step": 4649, "acc_step": 0, "speed/wps": 12968.52358345783, "speed/FLOPS": 203688483497076.3, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0305893421173096, "optim/lr": 0.0027888, "optim/total_tokens": 1218707456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424126, "loss/out": 6.291996955871582, "created_at": "2025-01-14T19:33:34.380569+00:00"} {"global_step": 4650, "acc_step": 0, "speed/wps": 12968.061470718254, "speed/FLOPS": 203681225381491.8, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6053929328918457, "optim/lr": 0.0027894, "optim/total_tokens": 1218969600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 6.197847366333008, "created_at": "2025-01-14T19:33:39.439047+00:00"} {"global_step": 4651, "acc_step": 0, "speed/wps": 12972.269315028709, "speed/FLOPS": 203747315358571.5, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4432036876678467, "optim/lr": 0.0027900000000000004, "optim/total_tokens": 1219231744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434003, "loss/out": 6.210319519042969, "created_at": "2025-01-14T19:33:44.498203+00:00"} {"global_step": 4652, "acc_step": 0, "speed/wps": 12967.063518224655, "speed/FLOPS": 203665551166249.4, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0009, "optim/grad_norm": 12.921635627746582, "optim/lr": 0.0027906000000000003, "optim/total_tokens": 1219493888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288203, "loss/out": 6.162687301635742, "created_at": "2025-01-14T19:33:49.560953+00:00"} {"global_step": 4653, "acc_step": 0, "speed/wps": 12966.367409149629, "speed/FLOPS": 203654617816672.0, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 12.012372016906738, "optim/lr": 0.0027912, "optim/total_tokens": 1219756032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458430, "loss/out": 6.482968330383301, "created_at": "2025-01-14T19:33:54.618777+00:00"} {"global_step": 4654, "acc_step": 0, "speed/wps": 12970.428096046617, "speed/FLOPS": 203718396484357.72, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.2022955417633057, "optim/lr": 0.0027918, "optim/total_tokens": 1220018176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 6.456263542175293, "created_at": "2025-01-14T19:33:59.675118+00:00"} {"global_step": 4655, "acc_step": 0, "speed/wps": 12974.460984286085, "speed/FLOPS": 203781738535927.75, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7526450157165527, "optim/lr": 0.0027924, "optim/total_tokens": 1220280320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464007, "loss/out": 6.348111152648926, "created_at": "2025-01-14T19:34:04.734777+00:00"} {"global_step": 4656, "acc_step": 0, "speed/wps": 12963.57073783961, "speed/FLOPS": 203610692250719.03, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0009, "optim/grad_norm": 7.265190124511719, "optim/lr": 0.0027930000000000003, "optim/total_tokens": 1220542464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 6.308598041534424, "created_at": "2025-01-14T19:34:09.798441+00:00"} {"global_step": 4657, "acc_step": 0, "speed/wps": 12968.571199972239, "speed/FLOPS": 203689231379867.03, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.253372669219971, "optim/lr": 0.0027936000000000002, "optim/total_tokens": 1220804608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456045, "loss/out": 6.373775482177734, "created_at": "2025-01-14T19:34:14.855656+00:00"} {"global_step": 4658, "acc_step": 0, "speed/wps": 12963.405185615033, "speed/FLOPS": 203608092025539.47, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 24.03877830505371, "optim/lr": 0.0027942, "optim/total_tokens": 1221066752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429739, "loss/out": 6.429539203643799, "created_at": "2025-01-14T19:34:19.914931+00:00"} {"global_step": 4659, "acc_step": 0, "speed/wps": 12960.783286391053, "speed/FLOPS": 203566911495359.78, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.001, "optim/grad_norm": 2.32098388671875, "optim/lr": 0.0027948, "optim/total_tokens": 1221328896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320435, "loss/out": 6.466580390930176, "created_at": "2025-01-14T19:34:24.978024+00:00"} {"global_step": 4660, "acc_step": 0, "speed/wps": 12951.215330112984, "speed/FLOPS": 203416633594263.0, "speed/curr_iter_time": 1.2793, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.033392429351807, "optim/lr": 0.0027954, "optim/total_tokens": 1221591040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366127, "loss/out": 6.180201053619385, "created_at": "2025-01-14T19:34:30.047011+00:00"} {"global_step": 4661, "acc_step": 0, "speed/wps": 12961.231270364515, "speed/FLOPS": 203573947699257.5, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.544197082519531, "optim/lr": 0.0027960000000000003, "optim/total_tokens": 1221853184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469590, "loss/out": 6.288180828094482, "created_at": "2025-01-14T19:34:35.106796+00:00"} {"global_step": 4662, "acc_step": 0, "speed/wps": 12967.291141735468, "speed/FLOPS": 203669126306274.2, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.7275702953338623, "optim/lr": 0.0027966, "optim/total_tokens": 1222115328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454466, "loss/out": 6.424627780914307, "created_at": "2025-01-14T19:34:40.165182+00:00"} {"global_step": 4663, "acc_step": 0, "speed/wps": 12958.746775781632, "speed/FLOPS": 203534925297780.0, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.505051612854004, "optim/lr": 0.0027972, "optim/total_tokens": 1222377472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464535, "loss/out": 6.285006999969482, "created_at": "2025-01-14T19:34:45.227225+00:00"} {"global_step": 4664, "acc_step": 0, "speed/wps": 12960.471845046019, "speed/FLOPS": 203562019881070.75, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0374832153320312, "optim/lr": 0.0027978, "optim/total_tokens": 1222639616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434515, "loss/out": 6.193945407867432, "created_at": "2025-01-14T19:34:50.291088+00:00"} {"global_step": 4665, "acc_step": 0, "speed/wps": 12961.06862590116, "speed/FLOPS": 203571393144463.6, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.928326606750488, "optim/lr": 0.0027984, "optim/total_tokens": 1222901760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291186, "loss/out": 6.339731693267822, "created_at": "2025-01-14T19:34:55.355054+00:00"} {"global_step": 4666, "acc_step": 0, "speed/wps": 12960.305498322194, "speed/FLOPS": 203559407177188.75, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.548886775970459, "optim/lr": 0.0027990000000000003, "optim/total_tokens": 1223163904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444119, "loss/out": 6.259408473968506, "created_at": "2025-01-14T19:35:00.419977+00:00"} {"global_step": 4667, "acc_step": 0, "speed/wps": 12965.693319900314, "speed/FLOPS": 203644030318715.8, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 26.957996368408203, "optim/lr": 0.0027996, "optim/total_tokens": 1223426048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 6.183908939361572, "created_at": "2025-01-14T19:35:05.484706+00:00"} {"global_step": 4668, "acc_step": 0, "speed/wps": 12959.803812934264, "speed/FLOPS": 203551527518783.34, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 22.293235778808594, "optim/lr": 0.0028002, "optim/total_tokens": 1223688192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 6.388164520263672, "created_at": "2025-01-14T19:35:10.548461+00:00"} {"global_step": 4669, "acc_step": 0, "speed/wps": 12962.211499228024, "speed/FLOPS": 203589343540534.5, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0008, "optim/grad_norm": 9.062047958374023, "optim/lr": 0.0028008, "optim/total_tokens": 1223950336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 6.44061279296875, "created_at": "2025-01-14T19:35:15.613740+00:00"} {"global_step": 4670, "acc_step": 0, "speed/wps": 12954.999348651989, "speed/FLOPS": 203476066805204.44, "speed/curr_iter_time": 1.2795, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.730560779571533, "optim/lr": 0.0028014, "optim/total_tokens": 1224212480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 275344, "loss/out": 6.36656379699707, "created_at": "2025-01-14T19:35:20.675936+00:00"} {"global_step": 4671, "acc_step": 0, "speed/wps": 12949.082926050587, "speed/FLOPS": 203383141258232.8, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0017, "optim/grad_norm": 317.97711181640625, "optim/lr": 0.0028020000000000002, "optim/total_tokens": 1224474624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443595, "loss/out": 6.3662309646606445, "created_at": "2025-01-14T19:35:25.748089+00:00"} {"global_step": 4672, "acc_step": 0, "speed/wps": 12960.070336292512, "speed/FLOPS": 203555713634366.1, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.753403186798096, "optim/lr": 0.0028026, "optim/total_tokens": 1224736768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 6.420000076293945, "created_at": "2025-01-14T19:35:30.812850+00:00"} {"global_step": 4673, "acc_step": 0, "speed/wps": 12956.21783279618, "speed/FLOPS": 203495204773061.62, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.207399368286133, "optim/lr": 0.0028032, "optim/total_tokens": 1224998912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443016, "loss/out": 6.525825500488281, "created_at": "2025-01-14T19:35:35.884332+00:00"} {"global_step": 4674, "acc_step": 0, "speed/wps": 12960.307629904995, "speed/FLOPS": 203559440656625.75, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0008, "optim/grad_norm": 15.458889961242676, "optim/lr": 0.0028038, "optim/total_tokens": 1225261056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 6.405763626098633, "created_at": "2025-01-14T19:35:40.945058+00:00"} {"global_step": 4675, "acc_step": 0, "speed/wps": 12963.674400588707, "speed/FLOPS": 203612320416641.16, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.095799922943115, "optim/lr": 0.0028044, "optim/total_tokens": 1225523200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334608, "loss/out": 6.229848384857178, "created_at": "2025-01-14T19:35:46.004268+00:00"} {"global_step": 4676, "acc_step": 0, "speed/wps": 12964.225439178477, "speed/FLOPS": 203620975234899.88, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.908838272094727, "optim/lr": 0.002805, "optim/total_tokens": 1225785344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354800, "loss/out": 6.4042840003967285, "created_at": "2025-01-14T19:35:51.066825+00:00"} {"global_step": 4677, "acc_step": 0, "speed/wps": 12957.801105631912, "speed/FLOPS": 203520072248591.7, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.639141082763672, "optim/lr": 0.0028056, "optim/total_tokens": 1226047488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449411, "loss/out": 6.266016006469727, "created_at": "2025-01-14T19:35:56.130436+00:00"} {"global_step": 4678, "acc_step": 0, "speed/wps": 12959.066940529327, "speed/FLOPS": 203539953925095.0, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0009, "optim/grad_norm": 11.660118103027344, "optim/lr": 0.0028062, "optim/total_tokens": 1226309632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459237, "loss/out": 6.347080230712891, "created_at": "2025-01-14T19:36:01.191168+00:00"} {"global_step": 4679, "acc_step": 0, "speed/wps": 12968.376957477596, "speed/FLOPS": 203686180534571.5, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.460981369018555, "optim/lr": 0.0028068, "optim/total_tokens": 1226571776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446747, "loss/out": 6.418293476104736, "created_at": "2025-01-14T19:36:06.249739+00:00"} {"global_step": 4680, "acc_step": 0, "speed/wps": 12959.585848736291, "speed/FLOPS": 203548104091539.97, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.823417901992798, "optim/lr": 0.0028074, "optim/total_tokens": 1226833920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 6.322850227355957, "created_at": "2025-01-14T19:36:11.310266+00:00"} {"global_step": 4681, "acc_step": 0, "speed/wps": 12963.744261514063, "speed/FLOPS": 203613417678472.84, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.554614305496216, "optim/lr": 0.002808, "optim/total_tokens": 1227096064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326583, "loss/out": 6.216053009033203, "created_at": "2025-01-14T19:36:16.370616+00:00"} {"global_step": 4682, "acc_step": 0, "speed/wps": 12963.098957862274, "speed/FLOPS": 203603282297881.38, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.628792762756348, "optim/lr": 0.0028086, "optim/total_tokens": 1227358208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362284, "loss/out": 6.2271623611450195, "created_at": "2025-01-14T19:36:21.429711+00:00"} {"global_step": 4683, "acc_step": 0, "speed/wps": 12949.469944788205, "speed/FLOPS": 203389219919326.56, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 64.29383087158203, "optim/lr": 0.0028092, "optim/total_tokens": 1227620352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435852, "loss/out": 6.257123947143555, "created_at": "2025-01-14T19:36:26.500250+00:00"} {"global_step": 4684, "acc_step": 0, "speed/wps": 12959.332872721412, "speed/FLOPS": 203544130755600.84, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.441788673400879, "optim/lr": 0.0028098, "optim/total_tokens": 1227882496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441450, "loss/out": 6.232267379760742, "created_at": "2025-01-14T19:36:31.567085+00:00"} {"global_step": 4685, "acc_step": 0, "speed/wps": 12957.619172886545, "speed/FLOPS": 203517214744823.9, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.6591131687164307, "optim/lr": 0.0028104, "optim/total_tokens": 1228144640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467734, "loss/out": 6.259788990020752, "created_at": "2025-01-14T19:36:36.633723+00:00"} {"global_step": 4686, "acc_step": 0, "speed/wps": 12964.982786756354, "speed/FLOPS": 203632870419315.06, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.273382186889648, "optim/lr": 0.002811, "optim/total_tokens": 1228406784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 6.253010272979736, "created_at": "2025-01-14T19:36:41.695542+00:00"} {"global_step": 4687, "acc_step": 0, "speed/wps": 12955.754161210407, "speed/FLOPS": 203487922173655.53, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 30.118982315063477, "optim/lr": 0.0028116, "optim/total_tokens": 1228668928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424678, "loss/out": 6.1018757820129395, "created_at": "2025-01-14T19:36:46.761929+00:00"} {"global_step": 4688, "acc_step": 0, "speed/wps": 12955.65362576696, "speed/FLOPS": 203486343126366.03, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.683795928955078, "optim/lr": 0.0028122, "optim/total_tokens": 1228931072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360703, "loss/out": 6.174004554748535, "created_at": "2025-01-14T19:36:51.823916+00:00"} {"global_step": 4689, "acc_step": 0, "speed/wps": 12957.774283896913, "speed/FLOPS": 203519650976386.6, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 15.791237831115723, "optim/lr": 0.0028128, "optim/total_tokens": 1229193216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 6.203287601470947, "created_at": "2025-01-14T19:36:56.886320+00:00"} {"global_step": 4690, "acc_step": 0, "speed/wps": 12960.81110171392, "speed/FLOPS": 203567348373227.66, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.772340774536133, "optim/lr": 0.0028133999999999998, "optim/total_tokens": 1229455360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 6.341527462005615, "created_at": "2025-01-14T19:37:01.947361+00:00"} {"global_step": 4691, "acc_step": 0, "speed/wps": 12966.520240045504, "speed/FLOPS": 203657018235903.38, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.28444766998291, "optim/lr": 0.002814, "optim/total_tokens": 1229717504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430268, "loss/out": 6.290215492248535, "created_at": "2025-01-14T19:37:07.009130+00:00"} {"global_step": 4692, "acc_step": 0, "speed/wps": 12957.142104777877, "speed/FLOPS": 203509721734616.97, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 22.669878005981445, "optim/lr": 0.0028146, "optim/total_tokens": 1229979648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439064, "loss/out": 6.212236404418945, "created_at": "2025-01-14T19:37:12.075779+00:00"} {"global_step": 4693, "acc_step": 0, "speed/wps": 12953.545619915283, "speed/FLOPS": 203453233997761.84, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 20.55328369140625, "optim/lr": 0.0028152, "optim/total_tokens": 1230241792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336771, "loss/out": 6.242335319519043, "created_at": "2025-01-14T19:37:17.139873+00:00"} {"global_step": 4694, "acc_step": 0, "speed/wps": 12958.590923194039, "speed/FLOPS": 203532477418727.94, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 30.55398941040039, "optim/lr": 0.0028158, "optim/total_tokens": 1230503936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445170, "loss/out": 6.247779369354248, "created_at": "2025-01-14T19:37:22.202996+00:00"} {"global_step": 4695, "acc_step": 0, "speed/wps": 12951.603954275759, "speed/FLOPS": 203422737470764.94, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 466.6520690917969, "optim/lr": 0.0028164, "optim/total_tokens": 1230766080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 6.305586338043213, "created_at": "2025-01-14T19:37:27.272209+00:00"} {"global_step": 4696, "acc_step": 0, "speed/wps": 12956.084497486823, "speed/FLOPS": 203493110558806.28, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.651347160339355, "optim/lr": 0.002817, "optim/total_tokens": 1231028224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 6.325669765472412, "created_at": "2025-01-14T19:37:32.341613+00:00"} {"global_step": 4697, "acc_step": 0, "speed/wps": 12962.383291209131, "speed/FLOPS": 203592041769664.72, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 29.332592010498047, "optim/lr": 0.0028176, "optim/total_tokens": 1231290368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444641, "loss/out": 6.296853542327881, "created_at": "2025-01-14T19:37:37.409678+00:00"} {"global_step": 4698, "acc_step": 0, "speed/wps": 12963.05546739671, "speed/FLOPS": 203602599220357.53, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.211689949035645, "optim/lr": 0.0028182000000000003, "optim/total_tokens": 1231552512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302985, "loss/out": 6.313779830932617, "created_at": "2025-01-14T19:37:42.469029+00:00"} {"global_step": 4699, "acc_step": 0, "speed/wps": 12960.31659306319, "speed/FLOPS": 203559581435341.84, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.508955001831055, "optim/lr": 0.0028188, "optim/total_tokens": 1231814656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323502, "loss/out": 6.331647872924805, "created_at": "2025-01-14T19:37:47.532312+00:00"} {"global_step": 4700, "acc_step": 0, "speed/wps": 12950.905475828602, "speed/FLOPS": 203411766906942.38, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0009, "optim/grad_norm": 16.093141555786133, "optim/lr": 0.0028194, "optim/total_tokens": 1232076800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451275, "loss/out": 6.167552947998047, "created_at": "2025-01-14T19:37:52.596320+00:00"} {"global_step": 4701, "acc_step": 0, "speed/wps": 12959.075338748687, "speed/FLOPS": 203540085830670.44, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.602730751037598, "optim/lr": 0.00282, "optim/total_tokens": 1232338944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 6.327528953552246, "created_at": "2025-01-14T19:37:57.659439+00:00"} {"global_step": 4702, "acc_step": 0, "speed/wps": 12957.229906130402, "speed/FLOPS": 203511100775510.1, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 29.40634536743164, "optim/lr": 0.0028206000000000004, "optim/total_tokens": 1232601088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 6.266214370727539, "created_at": "2025-01-14T19:38:02.720774+00:00"} {"global_step": 4703, "acc_step": 0, "speed/wps": 12955.185605968347, "speed/FLOPS": 203478992232302.22, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.107933521270752, "optim/lr": 0.0028212000000000003, "optim/total_tokens": 1232863232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 6.269189834594727, "created_at": "2025-01-14T19:38:07.782942+00:00"} {"global_step": 4704, "acc_step": 0, "speed/wps": 12951.554333571232, "speed/FLOPS": 203421958109414.34, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0006, "optim/grad_norm": 15.482186317443848, "optim/lr": 0.0028218, "optim/total_tokens": 1233125376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321343, "loss/out": 6.219048500061035, "created_at": "2025-01-14T19:38:12.848049+00:00"} {"global_step": 4705, "acc_step": 0, "speed/wps": 12955.94274451202, "speed/FLOPS": 203490884133548.0, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 19.519311904907227, "optim/lr": 0.0028224, "optim/total_tokens": 1233387520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362914, "loss/out": 6.384405136108398, "created_at": "2025-01-14T19:38:17.911145+00:00"} {"global_step": 4706, "acc_step": 0, "speed/wps": 12957.471817525973, "speed/FLOPS": 203514900326398.62, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.399740219116211, "optim/lr": 0.002823, "optim/total_tokens": 1233649664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462963, "loss/out": 6.354686737060547, "created_at": "2025-01-14T19:38:22.973054+00:00"} {"global_step": 4707, "acc_step": 0, "speed/wps": 12948.820697801179, "speed/FLOPS": 203379022603236.47, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 41.6763916015625, "optim/lr": 0.0028236000000000003, "optim/total_tokens": 1233911808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448361, "loss/out": 6.235079765319824, "created_at": "2025-01-14T19:38:28.038192+00:00"} {"global_step": 4708, "acc_step": 0, "speed/wps": 12946.479221431524, "speed/FLOPS": 203342246499325.34, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.533248901367188, "optim/lr": 0.0028242000000000002, "optim/total_tokens": 1234173952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 6.356218338012695, "created_at": "2025-01-14T19:38:33.106003+00:00"} {"global_step": 4709, "acc_step": 0, "speed/wps": 12946.656334805959, "speed/FLOPS": 203345028308250.12, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.10278034210205, "optim/lr": 0.0028248, "optim/total_tokens": 1234436096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324181, "loss/out": 6.291410446166992, "created_at": "2025-01-14T19:38:38.174406+00:00"} {"global_step": 4710, "acc_step": 0, "speed/wps": 12953.410026682635, "speed/FLOPS": 203451104319718.5, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.909368515014648, "optim/lr": 0.0028254, "optim/total_tokens": 1234698240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380268, "loss/out": 6.215006351470947, "created_at": "2025-01-14T19:38:43.240266+00:00"} {"global_step": 4711, "acc_step": 0, "speed/wps": 12952.325075462923, "speed/FLOPS": 203434063669933.7, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0009, "optim/grad_norm": 19.707597732543945, "optim/lr": 0.002826, "optim/total_tokens": 1234960384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439583, "loss/out": 6.206263542175293, "created_at": "2025-01-14T19:38:48.306061+00:00"} {"global_step": 4712, "acc_step": 0, "speed/wps": 12962.645376590235, "speed/FLOPS": 203596158180718.0, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 26.591400146484375, "optim/lr": 0.0028266000000000003, "optim/total_tokens": 1235222528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456330, "loss/out": 6.28455924987793, "created_at": "2025-01-14T19:38:53.369830+00:00"} {"global_step": 4713, "acc_step": 0, "speed/wps": 12953.094983971065, "speed/FLOPS": 203446156141019.22, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0006, "optim/grad_norm": 18.454486846923828, "optim/lr": 0.0028272, "optim/total_tokens": 1235484672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 6.281869888305664, "created_at": "2025-01-14T19:38:58.435620+00:00"} {"global_step": 4714, "acc_step": 0, "speed/wps": 12956.79698286585, "speed/FLOPS": 203504301120741.4, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 11.035529136657715, "optim/lr": 0.0028278, "optim/total_tokens": 1235746816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462672, "loss/out": 6.147764682769775, "created_at": "2025-01-14T19:39:03.498584+00:00"} {"global_step": 4715, "acc_step": 0, "speed/wps": 12952.213567172308, "speed/FLOPS": 203432312279001.16, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.76968765258789, "optim/lr": 0.0028284, "optim/total_tokens": 1236008960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439054, "loss/out": 6.249780654907227, "created_at": "2025-01-14T19:39:08.562426+00:00"} {"global_step": 4716, "acc_step": 0, "speed/wps": 12958.28990532995, "speed/FLOPS": 203527749519530.53, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.398876190185547, "optim/lr": 0.002829, "optim/total_tokens": 1236271104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292430, "loss/out": 6.2742767333984375, "created_at": "2025-01-14T19:39:13.625195+00:00"} {"global_step": 4717, "acc_step": 0, "speed/wps": 12948.91822974291, "speed/FLOPS": 203380554476404.3, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.869024753570557, "optim/lr": 0.0028296000000000003, "optim/total_tokens": 1236533248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444119, "loss/out": 6.260409832000732, "created_at": "2025-01-14T19:39:18.689903+00:00"} {"global_step": 4718, "acc_step": 0, "speed/wps": 12959.725965349415, "speed/FLOPS": 203550304815489.5, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 19.15431022644043, "optim/lr": 0.0028302, "optim/total_tokens": 1236795392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 6.157844543457031, "created_at": "2025-01-14T19:39:23.755513+00:00"} {"global_step": 4719, "acc_step": 0, "speed/wps": 12956.341726527518, "speed/FLOPS": 203497150694360.12, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.097473621368408, "optim/lr": 0.0028308, "optim/total_tokens": 1237057536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440628, "loss/out": 6.19801139831543, "created_at": "2025-01-14T19:39:28.823205+00:00"} {"global_step": 4720, "acc_step": 0, "speed/wps": 12957.06248550458, "speed/FLOPS": 203508471204521.53, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 15.584611892700195, "optim/lr": 0.0028314, "optim/total_tokens": 1237319680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440105, "loss/out": 6.1850762367248535, "created_at": "2025-01-14T19:39:33.890733+00:00"} {"global_step": 4721, "acc_step": 0, "speed/wps": 12955.434210218942, "speed/FLOPS": 203482896903675.84, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 17.634275436401367, "optim/lr": 0.002832, "optim/total_tokens": 1237581824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307440, "loss/out": 6.193192481994629, "created_at": "2025-01-14T19:39:38.959062+00:00"} {"global_step": 4722, "acc_step": 0, "speed/wps": 12952.274498003404, "speed/FLOPS": 203433269281431.25, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0013, "optim/grad_norm": 7.758630752563477, "optim/lr": 0.0028326000000000002, "optim/total_tokens": 1237843968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445933, "loss/out": 6.346486568450928, "created_at": "2025-01-14T19:39:44.027096+00:00"} {"global_step": 4723, "acc_step": 0, "speed/wps": 12957.626893990413, "speed/FLOPS": 203517336015370.06, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 26.866132736206055, "optim/lr": 0.0028332, "optim/total_tokens": 1238106112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447839, "loss/out": 6.224322319030762, "created_at": "2025-01-14T19:39:49.095814+00:00"} {"global_step": 4724, "acc_step": 0, "speed/wps": 12957.7199965297, "speed/FLOPS": 203518798318685.84, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0008, "optim/grad_norm": 20.31806182861328, "optim/lr": 0.0028338, "optim/total_tokens": 1238368256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 6.377177715301514, "created_at": "2025-01-14T19:39:54.162861+00:00"} {"global_step": 4725, "acc_step": 0, "speed/wps": 12953.091138039621, "speed/FLOPS": 203446095735381.03, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.52559757232666, "optim/lr": 0.0028344, "optim/total_tokens": 1238630400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445976, "loss/out": 6.339034080505371, "created_at": "2025-01-14T19:39:59.231364+00:00"} {"global_step": 4726, "acc_step": 0, "speed/wps": 12952.464181606041, "speed/FLOPS": 203436248523063.25, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 28.4341983795166, "optim/lr": 0.002835, "optim/total_tokens": 1238892544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 286904, "loss/out": 6.295472621917725, "created_at": "2025-01-14T19:40:04.294669+00:00"} {"global_step": 4727, "acc_step": 0, "speed/wps": 12950.646224553298, "speed/FLOPS": 203407695009415.16, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.4640889167785645, "optim/lr": 0.0028356, "optim/total_tokens": 1239154688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459779, "loss/out": 6.4887375831604, "created_at": "2025-01-14T19:40:09.364168+00:00"} {"global_step": 4728, "acc_step": 0, "speed/wps": 12947.436256297333, "speed/FLOPS": 203357278046996.6, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0009, "optim/grad_norm": 12.078975677490234, "optim/lr": 0.0028362, "optim/total_tokens": 1239416832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 6.2880353927612305, "created_at": "2025-01-14T19:40:14.429611+00:00"} {"global_step": 4729, "acc_step": 0, "speed/wps": 12947.74243537312, "speed/FLOPS": 203362087010115.97, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 18.793964385986328, "optim/lr": 0.0028368, "optim/total_tokens": 1239678976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437717, "loss/out": 6.437620162963867, "created_at": "2025-01-14T19:40:19.495234+00:00"} {"global_step": 4730, "acc_step": 0, "speed/wps": 12955.670072274112, "speed/FLOPS": 203486601441360.72, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.5593671798706055, "optim/lr": 0.0028374, "optim/total_tokens": 1239941120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 6.471057891845703, "created_at": "2025-01-14T19:40:24.560068+00:00"} {"global_step": 4731, "acc_step": 0, "speed/wps": 12951.424784228288, "speed/FLOPS": 203419923358963.97, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.5471415519714355, "optim/lr": 0.002838, "optim/total_tokens": 1240203264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291186, "loss/out": 6.327219486236572, "created_at": "2025-01-14T19:40:29.626308+00:00"} {"global_step": 4732, "acc_step": 0, "speed/wps": 12954.656659008622, "speed/FLOPS": 203470684393450.44, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.7308807373046875, "optim/lr": 0.0028386, "optim/total_tokens": 1240465408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462150, "loss/out": 6.4133782386779785, "created_at": "2025-01-14T19:40:34.694159+00:00"} {"global_step": 4733, "acc_step": 0, "speed/wps": 12954.184862270604, "speed/FLOPS": 203463274177363.22, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 19.332679748535156, "optim/lr": 0.0028392, "optim/total_tokens": 1240727552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 6.149477958679199, "created_at": "2025-01-14T19:40:39.759919+00:00"} {"global_step": 4734, "acc_step": 0, "speed/wps": 12954.330297052527, "speed/FLOPS": 203465558430461.66, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.120083332061768, "optim/lr": 0.0028398, "optim/total_tokens": 1240989696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452093, "loss/out": 6.404899597167969, "created_at": "2025-01-14T19:40:44.830194+00:00"} {"global_step": 4735, "acc_step": 0, "speed/wps": 12954.77691206225, "speed/FLOPS": 203472573132902.7, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.230642318725586, "optim/lr": 0.0028404, "optim/total_tokens": 1241251840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436679, "loss/out": 6.2400288581848145, "created_at": "2025-01-14T19:40:49.897481+00:00"} {"global_step": 4736, "acc_step": 0, "speed/wps": 12963.171379734815, "speed/FLOPS": 203604419782907.7, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.945528745651245, "optim/lr": 0.0028409999999999998, "optim/total_tokens": 1241513984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358828, "loss/out": 6.233868598937988, "created_at": "2025-01-14T19:40:54.964503+00:00"} {"global_step": 4737, "acc_step": 0, "speed/wps": 12954.90403853321, "speed/FLOPS": 203474569828819.2, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.277964115142822, "optim/lr": 0.0028416, "optim/total_tokens": 1241776128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454995, "loss/out": 6.3281989097595215, "created_at": "2025-01-14T19:41:00.029283+00:00"} {"global_step": 4738, "acc_step": 0, "speed/wps": 12953.215258105427, "speed/FLOPS": 203448045211573.38, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.001, "optim/grad_norm": 8.277653694152832, "optim/lr": 0.0028422, "optim/total_tokens": 1242038272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454466, "loss/out": 6.301835536956787, "created_at": "2025-01-14T19:41:05.101361+00:00"} {"global_step": 4739, "acc_step": 0, "speed/wps": 12955.206955077174, "speed/FLOPS": 203479327549393.1, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.21325159072876, "optim/lr": 0.0028428, "optim/total_tokens": 1242300416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481279, "loss/out": 6.309206485748291, "created_at": "2025-01-14T19:41:10.164540+00:00"} {"global_step": 4740, "acc_step": 0, "speed/wps": 12954.5125524185, "speed/FLOPS": 203468421001429.38, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 40.09342956542969, "optim/lr": 0.0028434, "optim/total_tokens": 1242562560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465633, "loss/out": 6.327032089233398, "created_at": "2025-01-14T19:41:15.228231+00:00"} {"global_step": 4741, "acc_step": 0, "speed/wps": 12956.988542951183, "speed/FLOPS": 203507309835112.7, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 97.02919006347656, "optim/lr": 0.0028439999999999997, "optim/total_tokens": 1242824704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321253, "loss/out": 6.268702030181885, "created_at": "2025-01-14T19:41:20.294093+00:00"} {"global_step": 4742, "acc_step": 0, "speed/wps": 12956.687422987005, "speed/FLOPS": 203502580332293.1, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.356507301330566, "optim/lr": 0.0028446, "optim/total_tokens": 1243086848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423001, "loss/out": 6.379317760467529, "created_at": "2025-01-14T19:41:25.360663+00:00"} {"global_step": 4743, "acc_step": 0, "speed/wps": 12959.251691936744, "speed/FLOPS": 203542855699823.53, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.787081480026245, "optim/lr": 0.0028452, "optim/total_tokens": 1243348992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 6.349731922149658, "created_at": "2025-01-14T19:41:30.422573+00:00"} {"global_step": 4744, "acc_step": 0, "speed/wps": 12953.23647196388, "speed/FLOPS": 203448378404371.28, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0009, "optim/grad_norm": 11.847254753112793, "optim/lr": 0.0028458, "optim/total_tokens": 1243611136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438536, "loss/out": 6.343499183654785, "created_at": "2025-01-14T19:41:35.485812+00:00"} {"global_step": 4745, "acc_step": 0, "speed/wps": 12957.444066638853, "speed/FLOPS": 203514464460582.7, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.288032531738281, "optim/lr": 0.0028464, "optim/total_tokens": 1243873280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479906, "loss/out": 6.30210018157959, "created_at": "2025-01-14T19:41:40.547329+00:00"} {"global_step": 4746, "acc_step": 0, "speed/wps": 12960.064254464476, "speed/FLOPS": 203555618110899.47, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.913690090179443, "optim/lr": 0.002847, "optim/total_tokens": 1244135424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442784, "loss/out": 6.284196853637695, "created_at": "2025-01-14T19:41:45.608388+00:00"} {"global_step": 4747, "acc_step": 0, "speed/wps": 12954.692707025186, "speed/FLOPS": 203471250577085.75, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.947366237640381, "optim/lr": 0.0028476, "optim/total_tokens": 1244397568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356429, "loss/out": 6.316740989685059, "created_at": "2025-01-14T19:41:50.670764+00:00"} {"global_step": 4748, "acc_step": 0, "speed/wps": 12959.483356400291, "speed/FLOPS": 203546494308563.88, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.687421798706055, "optim/lr": 0.0028482, "optim/total_tokens": 1244659712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460287, "loss/out": 6.270909309387207, "created_at": "2025-01-14T19:41:55.731213+00:00"} {"global_step": 4749, "acc_step": 0, "speed/wps": 12959.038987114309, "speed/FLOPS": 203539514878300.12, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0008, "optim/grad_norm": 8.204987525939941, "optim/lr": 0.0028488, "optim/total_tokens": 1244921856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457665, "loss/out": 6.290434837341309, "created_at": "2025-01-14T19:42:00.797059+00:00"} {"global_step": 4750, "acc_step": 0, "speed/wps": 12961.077363320868, "speed/FLOPS": 203571530377646.6, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.299851417541504, "optim/lr": 0.0028494, "optim/total_tokens": 1245184000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 6.318153381347656, "created_at": "2025-01-14T19:42:05.862113+00:00"} {"global_step": 4751, "acc_step": 0, "speed/wps": 12957.517067163952, "speed/FLOPS": 203515611034142.6, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.110769033432007, "optim/lr": 0.00285, "optim/total_tokens": 1245446144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442778, "loss/out": 6.2161760330200195, "created_at": "2025-01-14T19:42:10.924600+00:00"} {"global_step": 4752, "acc_step": 0, "speed/wps": 12950.440797276233, "speed/FLOPS": 203404468491742.16, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.836597442626953, "optim/lr": 0.0028506, "optim/total_tokens": 1245708288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338641, "loss/out": 6.25145149230957, "created_at": "2025-01-14T19:42:15.990697+00:00"} {"global_step": 4753, "acc_step": 0, "speed/wps": 12956.653057676407, "speed/FLOPS": 203502040577867.25, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.109865188598633, "optim/lr": 0.0028512, "optim/total_tokens": 1245970432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302985, "loss/out": 6.254734992980957, "created_at": "2025-01-14T19:42:21.052734+00:00"} {"global_step": 4754, "acc_step": 0, "speed/wps": 12958.942429075812, "speed/FLOPS": 203537998301620.9, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 37.146602630615234, "optim/lr": 0.0028518000000000003, "optim/total_tokens": 1246232576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 408061, "loss/out": 6.325697898864746, "created_at": "2025-01-14T19:42:26.131239+00:00"} {"global_step": 4755, "acc_step": 0, "speed/wps": 12960.480323859205, "speed/FLOPS": 203562153052482.97, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.7039031982421875, "optim/lr": 0.0028524, "optim/total_tokens": 1246494720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 6.222254753112793, "created_at": "2025-01-14T19:42:31.192351+00:00"} {"global_step": 4756, "acc_step": 0, "speed/wps": 12962.081583790232, "speed/FLOPS": 203587303040061.47, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.594414710998535, "optim/lr": 0.002853, "optim/total_tokens": 1246756864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444119, "loss/out": 6.2758355140686035, "created_at": "2025-01-14T19:42:36.253221+00:00"} {"global_step": 4757, "acc_step": 0, "speed/wps": 12960.647004492921, "speed/FLOPS": 203564771000878.66, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0012, "optim/grad_norm": 6.2377753257751465, "optim/lr": 0.0028536000000000004, "optim/total_tokens": 1247019008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457652, "loss/out": 6.346705913543701, "created_at": "2025-01-14T19:42:41.317552+00:00"} {"global_step": 4758, "acc_step": 0, "speed/wps": 12961.205179271623, "speed/FLOPS": 203573537902790.7, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.34163498878479, "optim/lr": 0.0028542000000000003, "optim/total_tokens": 1247281152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295714, "loss/out": 6.266602516174316, "created_at": "2025-01-14T19:42:46.381619+00:00"} {"global_step": 4759, "acc_step": 0, "speed/wps": 12957.370093485235, "speed/FLOPS": 203513302610555.34, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 20.013790130615234, "optim/lr": 0.0028548000000000002, "optim/total_tokens": 1247543296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438014, "loss/out": 6.342155456542969, "created_at": "2025-01-14T19:42:51.446970+00:00"} {"global_step": 4760, "acc_step": 0, "speed/wps": 12959.908857101513, "speed/FLOPS": 203553177381775.78, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.202247619628906, "optim/lr": 0.0028554, "optim/total_tokens": 1247805440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 6.3491997718811035, "created_at": "2025-01-14T19:42:56.511689+00:00"} {"global_step": 4761, "acc_step": 0, "speed/wps": 12954.825760882115, "speed/FLOPS": 203473340370743.97, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.894323348999023, "optim/lr": 0.002856, "optim/total_tokens": 1248067584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 6.328439235687256, "created_at": "2025-01-14T19:43:01.578191+00:00"} {"global_step": 4762, "acc_step": 0, "speed/wps": 12960.91639138347, "speed/FLOPS": 203569002092171.12, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.121513366699219, "optim/lr": 0.0028566000000000004, "optim/total_tokens": 1248329728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 6.3387908935546875, "created_at": "2025-01-14T19:43:06.641105+00:00"} {"global_step": 4763, "acc_step": 0, "speed/wps": 12958.148485883754, "speed/FLOPS": 203525528332797.34, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.152775526046753, "optim/lr": 0.0028572000000000003, "optim/total_tokens": 1248591872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310132, "loss/out": 6.268121719360352, "created_at": "2025-01-14T19:43:11.704638+00:00"} {"global_step": 4764, "acc_step": 0, "speed/wps": 12963.15786818349, "speed/FLOPS": 203604207565427.34, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.321633815765381, "optim/lr": 0.0028578, "optim/total_tokens": 1248854016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447032, "loss/out": 6.233985424041748, "created_at": "2025-01-14T19:43:16.788715+00:00"} {"global_step": 4765, "acc_step": 0, "speed/wps": 12958.128105416221, "speed/FLOPS": 203525208229549.03, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.127445936203003, "optim/lr": 0.0028584, "optim/total_tokens": 1249116160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445170, "loss/out": 6.2359466552734375, "created_at": "2025-01-14T19:43:21.864820+00:00"} {"global_step": 4766, "acc_step": 0, "speed/wps": 12961.193696263756, "speed/FLOPS": 203573357546372.84, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.380430221557617, "optim/lr": 0.002859, "optim/total_tokens": 1249378304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 6.311190128326416, "created_at": "2025-01-14T19:43:26.926547+00:00"} {"global_step": 4767, "acc_step": 0, "speed/wps": 12962.92784658235, "speed/FLOPS": 203600594760098.62, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.626021862030029, "optim/lr": 0.0028596000000000003, "optim/total_tokens": 1249640448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 6.250197887420654, "created_at": "2025-01-14T19:43:31.991265+00:00"} {"global_step": 4768, "acc_step": 0, "speed/wps": 12967.050689116983, "speed/FLOPS": 203665349667484.34, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.839329719543457, "optim/lr": 0.0028602000000000002, "optim/total_tokens": 1249902592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303511, "loss/out": 6.249514102935791, "created_at": "2025-01-14T19:43:37.049346+00:00"} {"global_step": 4769, "acc_step": 0, "speed/wps": 12962.974602938793, "speed/FLOPS": 203601329132926.53, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.335155010223389, "optim/lr": 0.0028608, "optim/total_tokens": 1250164736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448361, "loss/out": 6.271004676818848, "created_at": "2025-01-14T19:43:42.112217+00:00"} {"global_step": 4770, "acc_step": 0, "speed/wps": 12961.51616643778, "speed/FLOPS": 203578422383576.34, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.050809860229492, "optim/lr": 0.0028614, "optim/total_tokens": 1250426880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 6.255260467529297, "created_at": "2025-01-14T19:43:47.175826+00:00"} {"global_step": 4771, "acc_step": 0, "speed/wps": 12966.700344099268, "speed/FLOPS": 203659847017556.66, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.4415693283081055, "optim/lr": 0.002862, "optim/total_tokens": 1250689024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437432, "loss/out": 6.233559608459473, "created_at": "2025-01-14T19:43:52.240763+00:00"} {"global_step": 4772, "acc_step": 0, "speed/wps": 12963.648354587911, "speed/FLOPS": 203611911328407.72, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.7721707820892334, "optim/lr": 0.0028626000000000003, "optim/total_tokens": 1250951168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454466, "loss/out": 6.219612121582031, "created_at": "2025-01-14T19:43:57.303501+00:00"} {"global_step": 4773, "acc_step": 0, "speed/wps": 12963.705648285439, "speed/FLOPS": 203612811204657.16, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.2932193279266357, "optim/lr": 0.0028632, "optim/total_tokens": 1251213312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310540, "loss/out": 6.248205184936523, "created_at": "2025-01-14T19:44:02.363033+00:00"} {"global_step": 4774, "acc_step": 0, "speed/wps": 12971.103858861667, "speed/FLOPS": 203729010267963.5, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.6965320110321045, "optim/lr": 0.0028638, "optim/total_tokens": 1251475456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363102, "loss/out": 6.307587623596191, "created_at": "2025-01-14T19:44:07.419010+00:00"} {"global_step": 4775, "acc_step": 0, "speed/wps": 12960.856964314396, "speed/FLOPS": 203568068708389.97, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.547433376312256, "optim/lr": 0.0028644, "optim/total_tokens": 1251737600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447032, "loss/out": 6.117364883422852, "created_at": "2025-01-14T19:44:12.479374+00:00"} {"global_step": 4776, "acc_step": 0, "speed/wps": 12963.895717006906, "speed/FLOPS": 203615796495109.62, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.155564785003662, "optim/lr": 0.002865, "optim/total_tokens": 1251999744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 6.290630340576172, "created_at": "2025-01-14T19:44:17.542632+00:00"} {"global_step": 4777, "acc_step": 0, "speed/wps": 12963.900319287846, "speed/FLOPS": 203615868780255.97, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 20.892868041992188, "optim/lr": 0.0028656000000000003, "optim/total_tokens": 1252261888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426252, "loss/out": 6.253384590148926, "created_at": "2025-01-14T19:44:22.601979+00:00"} {"global_step": 4778, "acc_step": 0, "speed/wps": 12964.588508119776, "speed/FLOPS": 203626677731531.97, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.495584964752197, "optim/lr": 0.0028662, "optim/total_tokens": 1252524032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 6.370987892150879, "created_at": "2025-01-14T19:44:27.668471+00:00"} {"global_step": 4779, "acc_step": 0, "speed/wps": 12959.267534221963, "speed/FLOPS": 203543104524682.47, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.586190223693848, "optim/lr": 0.0028668, "optim/total_tokens": 1252786176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460556, "loss/out": 6.2256760597229, "created_at": "2025-01-14T19:44:32.729577+00:00"} {"global_step": 4780, "acc_step": 0, "speed/wps": 12966.774215948144, "speed/FLOPS": 203661007276453.8, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 19.469547271728516, "optim/lr": 0.0028674, "optim/total_tokens": 1253048320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456330, "loss/out": 6.190362930297852, "created_at": "2025-01-14T19:44:37.790214+00:00"} {"global_step": 4781, "acc_step": 0, "speed/wps": 12960.094093189617, "speed/FLOPS": 203556086769080.53, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.85147762298584, "optim/lr": 0.002868, "optim/total_tokens": 1253310464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371206, "loss/out": 6.338247299194336, "created_at": "2025-01-14T19:44:42.850675+00:00"} {"global_step": 4782, "acc_step": 0, "speed/wps": 12963.950108677202, "speed/FLOPS": 203616650791033.6, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0009, "optim/grad_norm": 15.382805824279785, "optim/lr": 0.0028686000000000002, "optim/total_tokens": 1253572608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435044, "loss/out": 6.171053409576416, "created_at": "2025-01-14T19:44:47.913600+00:00"} {"global_step": 4783, "acc_step": 0, "speed/wps": 12959.269484081076, "speed/FLOPS": 203543135149899.25, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.60333251953125, "optim/lr": 0.0028692, "optim/total_tokens": 1253834752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457665, "loss/out": 6.256633281707764, "created_at": "2025-01-14T19:44:52.976348+00:00"} {"global_step": 4784, "acc_step": 0, "speed/wps": 12950.147388707495, "speed/FLOPS": 203399860106984.53, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 23.795866012573242, "optim/lr": 0.0028698, "optim/total_tokens": 1254096896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 6.249499320983887, "created_at": "2025-01-14T19:44:58.040746+00:00"} {"global_step": 4785, "acc_step": 0, "speed/wps": 12963.969941425781, "speed/FLOPS": 203616962291603.1, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 22.549365997314453, "optim/lr": 0.0028704, "optim/total_tokens": 1254359040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 6.243775844573975, "created_at": "2025-01-14T19:45:03.102699+00:00"} {"global_step": 4786, "acc_step": 0, "speed/wps": 12961.633930195638, "speed/FLOPS": 203580272025217.88, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 50.9874267578125, "optim/lr": 0.002871, "optim/total_tokens": 1254621184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336239, "loss/out": 6.384689807891846, "created_at": "2025-01-14T19:45:08.166739+00:00"} {"global_step": 4787, "acc_step": 0, "speed/wps": 12963.561801226288, "speed/FLOPS": 203610551888926.62, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 316.5701599121094, "optim/lr": 0.0028716, "optim/total_tokens": 1254883328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390782, "loss/out": 6.386848449707031, "created_at": "2025-01-14T19:45:13.227437+00:00"} {"global_step": 4788, "acc_step": 0, "speed/wps": 12952.950606934457, "speed/FLOPS": 203443888501264.47, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 362.4132995605469, "optim/lr": 0.0028722, "optim/total_tokens": 1255145472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 6.410477638244629, "created_at": "2025-01-14T19:45:18.291898+00:00"} {"global_step": 4789, "acc_step": 0, "speed/wps": 12957.714348779882, "speed/FLOPS": 203518709613014.12, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 185.9089813232422, "optim/lr": 0.0028728, "optim/total_tokens": 1255407616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432656, "loss/out": 6.4968976974487305, "created_at": "2025-01-14T19:45:23.354273+00:00"} {"global_step": 4790, "acc_step": 0, "speed/wps": 12957.315220190094, "speed/FLOPS": 203512440750051.34, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 195.9290771484375, "optim/lr": 0.0028734, "optim/total_tokens": 1255669760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 6.6582722663879395, "created_at": "2025-01-14T19:45:28.421293+00:00"} {"global_step": 4791, "acc_step": 0, "speed/wps": 12956.302830568222, "speed/FLOPS": 203496539779870.72, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 183.28965759277344, "optim/lr": 0.002874, "optim/total_tokens": 1255931904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441687, "loss/out": 6.651465892791748, "created_at": "2025-01-14T19:45:33.486886+00:00"} {"global_step": 4792, "acc_step": 0, "speed/wps": 12960.403668821033, "speed/FLOPS": 203560949079773.4, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0009, "optim/grad_norm": 177.196044921875, "optim/lr": 0.0028746, "optim/total_tokens": 1256194048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299469, "loss/out": 6.657670021057129, "created_at": "2025-01-14T19:45:38.552738+00:00"} {"global_step": 4793, "acc_step": 0, "speed/wps": 12951.795155826136, "speed/FLOPS": 203425740553852.06, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 102.70782470703125, "optim/lr": 0.0028752, "optim/total_tokens": 1256456192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357199, "loss/out": 6.757598876953125, "created_at": "2025-01-14T19:45:43.618775+00:00"} {"global_step": 4794, "acc_step": 0, "speed/wps": 12954.707486412226, "speed/FLOPS": 203471482707668.22, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 294.52117919921875, "optim/lr": 0.0028758, "optim/total_tokens": 1256718336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452859, "loss/out": 6.667380332946777, "created_at": "2025-01-14T19:45:48.683343+00:00"} {"global_step": 4795, "acc_step": 0, "speed/wps": 12954.55824405583, "speed/FLOPS": 203469138651380.53, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 25.562976837158203, "optim/lr": 0.0028764, "optim/total_tokens": 1256980480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453375, "loss/out": 6.802481651306152, "created_at": "2025-01-14T19:45:53.748135+00:00"} {"global_step": 4796, "acc_step": 0, "speed/wps": 12965.494562577585, "speed/FLOPS": 203640908561841.2, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 336.3460693359375, "optim/lr": 0.002877, "optim/total_tokens": 1257242624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448605, "loss/out": 6.566284656524658, "created_at": "2025-01-14T19:45:58.806556+00:00"} {"global_step": 4797, "acc_step": 0, "speed/wps": 12953.488392142997, "speed/FLOPS": 203452335156958.9, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 150.64569091796875, "optim/lr": 0.0028776, "optim/total_tokens": 1257504768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 6.609507083892822, "created_at": "2025-01-14T19:46:03.876772+00:00"} {"global_step": 4798, "acc_step": 0, "speed/wps": 12962.526859007487, "speed/FLOPS": 203594296699220.72, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 141.15121459960938, "optim/lr": 0.0028782, "optim/total_tokens": 1257766912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310904, "loss/out": 6.593267440795898, "created_at": "2025-01-14T19:46:08.937377+00:00"} {"global_step": 4799, "acc_step": 0, "speed/wps": 12963.134154977524, "speed/FLOPS": 203603835116942.28, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 26.16084861755371, "optim/lr": 0.0028788, "optim/total_tokens": 1258029056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371118, "loss/out": 6.695302486419678, "created_at": "2025-01-14T19:46:14.004072+00:00"} {"global_step": 4800, "acc_step": 0, "speed/wps": 12956.284610020695, "speed/FLOPS": 203496253601134.16, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.591814041137695, "optim/lr": 0.0028794, "optim/total_tokens": 1258291200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461889, "loss/out": 6.595165252685547, "created_at": "2025-01-14T19:46:19.065920+00:00"} {"global_step": 4801, "acc_step": 0, "speed/wps": 12962.774689639473, "speed/FLOPS": 203598189219851.03, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0012, "optim/grad_norm": 73.8409652709961, "optim/lr": 0.0028799999999999997, "optim/total_tokens": 1258553344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445407, "loss/out": 6.49422550201416, "created_at": "2025-01-14T19:46:24.128429+00:00"} {"global_step": 4802, "acc_step": 0, "speed/wps": 12961.257698177547, "speed/FLOPS": 203574362784376.78, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 31.172067642211914, "optim/lr": 0.0028806, "optim/total_tokens": 1258815488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 6.545374870300293, "created_at": "2025-01-14T19:46:29.191696+00:00"} {"global_step": 4803, "acc_step": 0, "speed/wps": 12961.266944356554, "speed/FLOPS": 203574508008323.66, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 36.9236946105957, "optim/lr": 0.0028812, "optim/total_tokens": 1259077632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442256, "loss/out": 6.507633686065674, "created_at": "2025-01-14T19:46:34.251465+00:00"} {"global_step": 4804, "acc_step": 0, "speed/wps": 12955.769637147529, "speed/FLOPS": 203488165244514.22, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 22.16718101501465, "optim/lr": 0.0028818, "optim/total_tokens": 1259339776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308786, "loss/out": 6.537022113800049, "created_at": "2025-01-14T19:46:39.318613+00:00"} {"global_step": 4805, "acc_step": 0, "speed/wps": 12959.409359960579, "speed/FLOPS": 203545332092796.4, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 46.77931213378906, "optim/lr": 0.0028824, "optim/total_tokens": 1259601920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 342920, "loss/out": 6.517637252807617, "created_at": "2025-01-14T19:46:44.385952+00:00"} {"global_step": 4806, "acc_step": 0, "speed/wps": 12958.826592352756, "speed/FLOPS": 203536178926708.97, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0008, "optim/grad_norm": 28.54228973388672, "optim/lr": 0.002883, "optim/total_tokens": 1259864064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452325, "loss/out": 6.581520080566406, "created_at": "2025-01-14T19:46:49.450284+00:00"} {"global_step": 4807, "acc_step": 0, "speed/wps": 12954.076345749907, "speed/FLOPS": 203461569776288.34, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 25.93094825744629, "optim/lr": 0.0028836, "optim/total_tokens": 1260126208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 6.491359233856201, "created_at": "2025-01-14T19:46:54.513027+00:00"} {"global_step": 4808, "acc_step": 0, "speed/wps": 12952.835800005643, "speed/FLOPS": 203442085300686.2, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 915.292236328125, "optim/lr": 0.0028842, "optim/total_tokens": 1260388352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 412169, "loss/out": 6.524962425231934, "created_at": "2025-01-14T19:46:59.580277+00:00"} {"global_step": 4809, "acc_step": 0, "speed/wps": 12950.064166148435, "speed/FLOPS": 203398552982334.7, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 119.4906005859375, "optim/lr": 0.0028848000000000003, "optim/total_tokens": 1260650496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 6.62183952331543, "created_at": "2025-01-14T19:47:04.646006+00:00"} {"global_step": 4810, "acc_step": 0, "speed/wps": 12951.570028318678, "speed/FLOPS": 203422204616989.84, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 28.565570831298828, "optim/lr": 0.0028854, "optim/total_tokens": 1260912640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318851, "loss/out": 6.784541130065918, "created_at": "2025-01-14T19:47:09.711124+00:00"} {"global_step": 4811, "acc_step": 0, "speed/wps": 12947.503476660655, "speed/FLOPS": 203358333835173.03, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0008, "optim/grad_norm": 28.72486114501953, "optim/lr": 0.002886, "optim/total_tokens": 1261174784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430484, "loss/out": 6.790886878967285, "created_at": "2025-01-14T19:47:14.777679+00:00"} {"global_step": 4812, "acc_step": 0, "speed/wps": 12954.290661964937, "speed/FLOPS": 203464935906946.72, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 90.43029022216797, "optim/lr": 0.0028866, "optim/total_tokens": 1261436928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 6.6630449295043945, "created_at": "2025-01-14T19:47:19.841690+00:00"} {"global_step": 4813, "acc_step": 0, "speed/wps": 12954.006903761729, "speed/FLOPS": 203460479094440.72, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.221290588378906, "optim/lr": 0.0028872000000000004, "optim/total_tokens": 1261699072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425730, "loss/out": 6.7338457107543945, "created_at": "2025-01-14T19:47:24.908527+00:00"} {"global_step": 4814, "acc_step": 0, "speed/wps": 12957.702566544529, "speed/FLOPS": 203518524556817.44, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 43.64469528198242, "optim/lr": 0.0028878000000000003, "optim/total_tokens": 1261961216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 6.932772636413574, "created_at": "2025-01-14T19:47:29.969889+00:00"} {"global_step": 4815, "acc_step": 0, "speed/wps": 12949.319328976115, "speed/FLOPS": 203386854291029.88, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 23.947542190551758, "optim/lr": 0.0028884, "optim/total_tokens": 1262223360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327115, "loss/out": 7.149308204650879, "created_at": "2025-01-14T19:47:35.036101+00:00"} {"global_step": 4816, "acc_step": 0, "speed/wps": 12952.567006397101, "speed/FLOPS": 203437863527702.9, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0008, "optim/grad_norm": 17.803308486938477, "optim/lr": 0.002889, "optim/total_tokens": 1262485504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 381609, "loss/out": 6.847656726837158, "created_at": "2025-01-14T19:47:40.099355+00:00"} {"global_step": 4817, "acc_step": 0, "speed/wps": 12959.541988033032, "speed/FLOPS": 203547415198924.38, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 37.74095153808594, "optim/lr": 0.0028896, "optim/total_tokens": 1262747648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 6.8541765213012695, "created_at": "2025-01-14T19:47:45.160546+00:00"} {"global_step": 4818, "acc_step": 0, "speed/wps": 12956.518809268087, "speed/FLOPS": 203499932022137.97, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.266440391540527, "optim/lr": 0.0028902000000000003, "optim/total_tokens": 1263009792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 6.846749782562256, "created_at": "2025-01-14T19:47:50.225282+00:00"} {"global_step": 4819, "acc_step": 0, "speed/wps": 12955.442990132857, "speed/FLOPS": 203483034804288.75, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 17.746389389038086, "optim/lr": 0.0028908000000000002, "optim/total_tokens": 1263271936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480663, "loss/out": 6.844038963317871, "created_at": "2025-01-14T19:47:55.292131+00:00"} {"global_step": 4820, "acc_step": 0, "speed/wps": 12955.656938917833, "speed/FLOPS": 203486395163953.25, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 42.068748474121094, "optim/lr": 0.0028914, "optim/total_tokens": 1263534080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 6.993190765380859, "created_at": "2025-01-14T19:48:00.357648+00:00"} {"global_step": 4821, "acc_step": 0, "speed/wps": 12958.518552491956, "speed/FLOPS": 203531340737403.97, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 23.630094528198242, "optim/lr": 0.002892, "optim/total_tokens": 1263796224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 7.101726055145264, "created_at": "2025-01-14T19:48:05.422534+00:00"} {"global_step": 4822, "acc_step": 0, "speed/wps": 12957.456106085729, "speed/FLOPS": 203514653556639.8, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.4163522720336914, "optim/lr": 0.0028926, "optim/total_tokens": 1264058368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303689, "loss/out": 6.906587600708008, "created_at": "2025-01-14T19:48:10.487703+00:00"} {"global_step": 4823, "acc_step": 0, "speed/wps": 12954.827835808408, "speed/FLOPS": 203473372960312.66, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0013, "optim/grad_norm": 19.0672550201416, "optim/lr": 0.0028932000000000003, "optim/total_tokens": 1264320512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444878, "loss/out": 7.267672538757324, "created_at": "2025-01-14T19:48:15.552828+00:00"} {"global_step": 4824, "acc_step": 0, "speed/wps": 12959.885029278441, "speed/FLOPS": 203552803133071.6, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 28.575660705566406, "optim/lr": 0.0028938, "optim/total_tokens": 1264582656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 7.412652969360352, "created_at": "2025-01-14T19:48:20.618446+00:00"} {"global_step": 4825, "acc_step": 0, "speed/wps": 12964.508644440393, "speed/FLOPS": 203625423362703.1, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 123.592529296875, "optim/lr": 0.0028944, "optim/total_tokens": 1264844800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446503, "loss/out": 7.240777969360352, "created_at": "2025-01-14T19:48:25.680500+00:00"} {"global_step": 4826, "acc_step": 0, "speed/wps": 12965.783246338426, "speed/FLOPS": 203645442736998.22, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.926977157592773, "optim/lr": 0.002895, "optim/total_tokens": 1265106944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 7.6056413650512695, "created_at": "2025-01-14T19:48:30.738645+00:00"} {"global_step": 4827, "acc_step": 0, "speed/wps": 12968.249666827474, "speed/FLOPS": 203684181259995.66, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.418415546417236, "optim/lr": 0.0028956, "optim/total_tokens": 1265369088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432127, "loss/out": 7.210488319396973, "created_at": "2025-01-14T19:48:35.796484+00:00"} {"global_step": 4828, "acc_step": 0, "speed/wps": 12960.185760827717, "speed/FLOPS": 203557526535301.4, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 63.17164993286133, "optim/lr": 0.0028962000000000002, "optim/total_tokens": 1265631232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 346784, "loss/out": 7.116784572601318, "created_at": "2025-01-14T19:48:40.861321+00:00"} {"global_step": 4829, "acc_step": 0, "speed/wps": 12962.225089988326, "speed/FLOPS": 203589557002100.72, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.275423526763916, "optim/lr": 0.0028968, "optim/total_tokens": 1265893376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439820, "loss/out": 7.343908309936523, "created_at": "2025-01-14T19:48:45.924036+00:00"} {"global_step": 4830, "acc_step": 0, "speed/wps": 12964.10221384498, "speed/FLOPS": 203619039811707.97, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.300437927246094, "optim/lr": 0.0028974, "optim/total_tokens": 1266155520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476195, "loss/out": 7.175455093383789, "created_at": "2025-01-14T19:48:50.982981+00:00"} {"global_step": 4831, "acc_step": 0, "speed/wps": 12958.964993362544, "speed/FLOPS": 203538352704750.6, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 22.973682403564453, "optim/lr": 0.002898, "optim/total_tokens": 1266417664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438536, "loss/out": 7.14577579498291, "created_at": "2025-01-14T19:48:56.052541+00:00"} {"global_step": 4832, "acc_step": 0, "speed/wps": 12950.575865972161, "speed/FLOPS": 203406589931215.56, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.599592208862305, "optim/lr": 0.0028986, "optim/total_tokens": 1266679808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 7.16514778137207, "created_at": "2025-01-14T19:49:01.117492+00:00"} {"global_step": 4833, "acc_step": 0, "speed/wps": 12955.18922178833, "speed/FLOPS": 203479049023723.72, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.03545618057251, "optim/lr": 0.0028992, "optim/total_tokens": 1266941952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455808, "loss/out": 7.14535665512085, "created_at": "2025-01-14T19:49:06.181665+00:00"} {"global_step": 4834, "acc_step": 0, "speed/wps": 12963.724348255486, "speed/FLOPS": 203613104913383.22, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.352168560028076, "optim/lr": 0.0028998, "optim/total_tokens": 1267204096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295474, "loss/out": 7.008489608764648, "created_at": "2025-01-14T19:49:11.241162+00:00"} {"global_step": 4835, "acc_step": 0, "speed/wps": 12957.449098580932, "speed/FLOPS": 203514543494147.66, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.901830673217773, "optim/lr": 0.0029004, "optim/total_tokens": 1267466240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429438, "loss/out": 7.0755157470703125, "created_at": "2025-01-14T19:49:16.306797+00:00"} {"global_step": 4836, "acc_step": 0, "speed/wps": 12959.326175835007, "speed/FLOPS": 203544025571797.34, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 19.19717025756836, "optim/lr": 0.002901, "optim/total_tokens": 1267728384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 7.026147365570068, "created_at": "2025-01-14T19:49:21.371450+00:00"} {"global_step": 4837, "acc_step": 0, "speed/wps": 12961.457996810932, "speed/FLOPS": 203577508749653.25, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 18.72122573852539, "optim/lr": 0.0029016, "optim/total_tokens": 1267990528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 6.930672645568848, "created_at": "2025-01-14T19:49:26.435656+00:00"} {"global_step": 4838, "acc_step": 0, "speed/wps": 12962.561945564124, "speed/FLOPS": 203594847781806.7, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.591512203216553, "optim/lr": 0.0029022, "optim/total_tokens": 1268252672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 6.988406658172607, "created_at": "2025-01-14T19:49:31.495007+00:00"} {"global_step": 4839, "acc_step": 0, "speed/wps": 12956.205750326151, "speed/FLOPS": 203495015001266.78, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.690274715423584, "optim/lr": 0.0029028, "optim/total_tokens": 1268514816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325000, "loss/out": 7.0216569900512695, "created_at": "2025-01-14T19:49:36.565436+00:00"} {"global_step": 4840, "acc_step": 0, "speed/wps": 12962.460019809803, "speed/FLOPS": 203593246897774.44, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.06607723236084, "optim/lr": 0.0029034, "optim/total_tokens": 1268776960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421973, "loss/out": 6.838526725769043, "created_at": "2025-01-14T19:49:41.624726+00:00"} {"global_step": 4841, "acc_step": 0, "speed/wps": 12958.455944981732, "speed/FLOPS": 203530357400424.16, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.161005020141602, "optim/lr": 0.002904, "optim/total_tokens": 1269039104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473602, "loss/out": 6.78240966796875, "created_at": "2025-01-14T19:49:46.709756+00:00"} {"global_step": 4842, "acc_step": 0, "speed/wps": 12957.966156830733, "speed/FLOPS": 203522664604473.28, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.619585990905762, "optim/lr": 0.0029046, "optim/total_tokens": 1269301248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 6.76191520690918, "created_at": "2025-01-14T19:49:51.773315+00:00"} {"global_step": 4843, "acc_step": 0, "speed/wps": 12957.000760484563, "speed/FLOPS": 203507501728263.1, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.394482612609863, "optim/lr": 0.0029052, "optim/total_tokens": 1269563392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435329, "loss/out": 6.5640740394592285, "created_at": "2025-01-14T19:49:56.841672+00:00"} {"global_step": 4844, "acc_step": 0, "speed/wps": 12962.916365297784, "speed/FLOPS": 203600414430747.56, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 23.487083435058594, "optim/lr": 0.0029058, "optim/total_tokens": 1269825536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333311, "loss/out": 6.740088939666748, "created_at": "2025-01-14T19:50:01.902019+00:00"} {"global_step": 4845, "acc_step": 0, "speed/wps": 12954.02370617452, "speed/FLOPS": 203460742999422.53, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0014, "optim/grad_norm": 12.227155685424805, "optim/lr": 0.0029064, "optim/total_tokens": 1270087680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413141, "loss/out": 6.716719627380371, "created_at": "2025-01-14T19:50:06.968101+00:00"} {"global_step": 4846, "acc_step": 0, "speed/wps": 12951.481197598618, "speed/FLOPS": 203420809408465.1, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0009, "optim/grad_norm": 12.443216323852539, "optim/lr": 0.002907, "optim/total_tokens": 1270349824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 6.623530387878418, "created_at": "2025-01-14T19:50:12.035399+00:00"} {"global_step": 4847, "acc_step": 0, "speed/wps": 12956.799933975226, "speed/FLOPS": 203504347471969.4, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 9.92879581451416, "optim/lr": 0.0029075999999999998, "optim/total_tokens": 1270611968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457380, "loss/out": 6.545525550842285, "created_at": "2025-01-14T19:50:17.102878+00:00"} {"global_step": 4848, "acc_step": 0, "speed/wps": 12954.808364917015, "speed/FLOPS": 203473067143207.53, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.555213928222656, "optim/lr": 0.0029082, "optim/total_tokens": 1270874112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 6.677136421203613, "created_at": "2025-01-14T19:50:22.168249+00:00"} {"global_step": 4849, "acc_step": 0, "speed/wps": 12958.14024874058, "speed/FLOPS": 203525398957144.75, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 32.06349182128906, "optim/lr": 0.0029088, "optim/total_tokens": 1271136256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444067, "loss/out": 6.552714824676514, "created_at": "2025-01-14T19:50:27.232542+00:00"} {"global_step": 4850, "acc_step": 0, "speed/wps": 12957.728953723072, "speed/FLOPS": 203518939003716.22, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.163042068481445, "optim/lr": 0.0029094, "optim/total_tokens": 1271398400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363341, "loss/out": 6.6057448387146, "created_at": "2025-01-14T19:50:32.297302+00:00"} {"global_step": 4851, "acc_step": 0, "speed/wps": 12955.474651350214, "speed/FLOPS": 203483532087214.4, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 18.277206420898438, "optim/lr": 0.00291, "optim/total_tokens": 1271660544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437438, "loss/out": 6.706584930419922, "created_at": "2025-01-14T19:50:37.362554+00:00"} {"global_step": 4852, "acc_step": 0, "speed/wps": 12957.77241084155, "speed/FLOPS": 203519621557478.25, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.7938175201416, "optim/lr": 0.0029105999999999997, "optim/total_tokens": 1271922688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 6.656982898712158, "created_at": "2025-01-14T19:50:42.425460+00:00"} {"global_step": 4853, "acc_step": 0, "speed/wps": 12957.468940074841, "speed/FLOPS": 203514855132074.56, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 24.084312438964844, "optim/lr": 0.0029112, "optim/total_tokens": 1272184832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443022, "loss/out": 6.618350982666016, "created_at": "2025-01-14T19:50:47.488002+00:00"} {"global_step": 4854, "acc_step": 0, "speed/wps": 12960.879719083403, "speed/FLOPS": 203568426103305.53, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 320.79949951171875, "optim/lr": 0.0029118, "optim/total_tokens": 1272446976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460578, "loss/out": 6.5663533210754395, "created_at": "2025-01-14T19:50:52.552800+00:00"} {"global_step": 4855, "acc_step": 0, "speed/wps": 12961.186894799459, "speed/FLOPS": 203573250720029.88, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 344.87078857421875, "optim/lr": 0.0029124, "optim/total_tokens": 1272709120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440637, "loss/out": 6.596217155456543, "created_at": "2025-01-14T19:50:57.616915+00:00"} {"global_step": 4856, "acc_step": 0, "speed/wps": 12967.149167326495, "speed/FLOPS": 203666896403088.0, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 48.9785041809082, "optim/lr": 0.002913, "optim/total_tokens": 1272971264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295894, "loss/out": 6.536359786987305, "created_at": "2025-01-14T19:51:02.678045+00:00"} {"global_step": 4857, "acc_step": 0, "speed/wps": 12960.826892906834, "speed/FLOPS": 203567596395611.75, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 22.776662826538086, "optim/lr": 0.0029136, "optim/total_tokens": 1273233408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443015, "loss/out": 6.627694129943848, "created_at": "2025-01-14T19:51:07.739344+00:00"} {"global_step": 4858, "acc_step": 0, "speed/wps": 12967.82842168733, "speed/FLOPS": 203677565026218.22, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.890368461608887, "optim/lr": 0.0029142, "optim/total_tokens": 1273495552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444119, "loss/out": 6.672884941101074, "created_at": "2025-01-14T19:51:12.801512+00:00"} {"global_step": 4859, "acc_step": 0, "speed/wps": 12962.64881987432, "speed/FLOPS": 203596212262226.16, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.600332260131836, "optim/lr": 0.0029148, "optim/total_tokens": 1273757696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453945, "loss/out": 6.604974269866943, "created_at": "2025-01-14T19:51:17.860934+00:00"} {"global_step": 4860, "acc_step": 0, "speed/wps": 12964.485107820581, "speed/FLOPS": 203625053687746.22, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 41.817115783691406, "optim/lr": 0.0029154, "optim/total_tokens": 1274019840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 6.547892093658447, "created_at": "2025-01-14T19:51:22.920788+00:00"} {"global_step": 4861, "acc_step": 0, "speed/wps": 12963.142503544474, "speed/FLOPS": 203603966242658.34, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.001, "optim/grad_norm": 8.789510726928711, "optim/lr": 0.002916, "optim/total_tokens": 1274281984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425201, "loss/out": 6.667473316192627, "created_at": "2025-01-14T19:51:27.984443+00:00"} {"global_step": 4862, "acc_step": 0, "speed/wps": 12952.810715468077, "speed/FLOPS": 203441691313553.94, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 53.7747917175293, "optim/lr": 0.0029166, "optim/total_tokens": 1274544128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467205, "loss/out": 6.614138603210449, "created_at": "2025-01-14T19:51:33.050859+00:00"} {"global_step": 4863, "acc_step": 0, "speed/wps": 12957.243302734976, "speed/FLOPS": 203511311187593.78, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 76.40422821044922, "optim/lr": 0.0029172, "optim/total_tokens": 1274806272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294835, "loss/out": 6.864575386047363, "created_at": "2025-01-14T19:51:38.116694+00:00"} {"global_step": 4864, "acc_step": 0, "speed/wps": 12960.558254506623, "speed/FLOPS": 203563377060388.6, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 138.0736846923828, "optim/lr": 0.0029178, "optim/total_tokens": 1275068416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446742, "loss/out": 6.963149070739746, "created_at": "2025-01-14T19:51:43.180619+00:00"} {"global_step": 4865, "acc_step": 0, "speed/wps": 12959.83683447431, "speed/FLOPS": 203552046167446.12, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 56.7780647277832, "optim/lr": 0.0029184000000000002, "optim/total_tokens": 1275330560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 6.940707206726074, "created_at": "2025-01-14T19:51:48.245697+00:00"} {"global_step": 4866, "acc_step": 0, "speed/wps": 12962.276253773747, "speed/FLOPS": 203590360599657.16, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 39.56693649291992, "optim/lr": 0.002919, "optim/total_tokens": 1275592704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455517, "loss/out": 6.924266338348389, "created_at": "2025-01-14T19:51:53.309466+00:00"} {"global_step": 4867, "acc_step": 0, "speed/wps": 12959.642801026805, "speed/FLOPS": 203548998605523.12, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.001, "optim/grad_norm": 16.172584533691406, "optim/lr": 0.0029196, "optim/total_tokens": 1275854848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470119, "loss/out": 6.922449111938477, "created_at": "2025-01-14T19:51:58.369984+00:00"} {"global_step": 4868, "acc_step": 0, "speed/wps": 12964.478761498814, "speed/FLOPS": 203624954010042.12, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.29111671447754, "optim/lr": 0.0029202000000000004, "optim/total_tokens": 1276116992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322976, "loss/out": 6.775907516479492, "created_at": "2025-01-14T19:52:03.430112+00:00"} {"global_step": 4869, "acc_step": 0, "speed/wps": 12963.029216127448, "speed/FLOPS": 203602186908100.75, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.980160713195801, "optim/lr": 0.0029208000000000003, "optim/total_tokens": 1276379136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 381895, "loss/out": 6.854400157928467, "created_at": "2025-01-14T19:52:08.490653+00:00"} {"global_step": 4870, "acc_step": 0, "speed/wps": 12958.345610143459, "speed/FLOPS": 203528624440173.12, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.975319862365723, "optim/lr": 0.0029214, "optim/total_tokens": 1276641280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451797, "loss/out": 6.702285289764404, "created_at": "2025-01-14T19:52:13.554698+00:00"} {"global_step": 4871, "acc_step": 0, "speed/wps": 12960.137692681006, "speed/FLOPS": 203556771559004.78, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.611040115356445, "optim/lr": 0.002922, "optim/total_tokens": 1276903424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 6.593849182128906, "created_at": "2025-01-14T19:52:18.617407+00:00"} {"global_step": 4872, "acc_step": 0, "speed/wps": 12955.827345449283, "speed/FLOPS": 203489071632692.78, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 52.952545166015625, "optim/lr": 0.0029226, "optim/total_tokens": 1277165568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443835, "loss/out": 6.727014541625977, "created_at": "2025-01-14T19:52:23.685505+00:00"} {"global_step": 4873, "acc_step": 0, "speed/wps": 12967.474868380015, "speed/FLOPS": 203672011985693.7, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 2231.906494140625, "optim/lr": 0.0029232000000000004, "optim/total_tokens": 1277427712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464535, "loss/out": 6.944096565246582, "created_at": "2025-01-14T19:52:28.746682+00:00"} {"global_step": 4874, "acc_step": 0, "speed/wps": 12962.821958293303, "speed/FLOPS": 203598931639013.38, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 18.31629753112793, "optim/lr": 0.0029238000000000003, "optim/total_tokens": 1277689856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464535, "loss/out": 7.061746597290039, "created_at": "2025-01-14T19:52:33.808068+00:00"} {"global_step": 4875, "acc_step": 0, "speed/wps": 12959.957849653301, "speed/FLOPS": 203553946877124.44, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 27.793745040893555, "optim/lr": 0.0029244, "optim/total_tokens": 1277952000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290078, "loss/out": 6.689783096313477, "created_at": "2025-01-14T19:52:38.871249+00:00"} {"global_step": 4876, "acc_step": 0, "speed/wps": 12961.6964418803, "speed/FLOPS": 203581253857125.53, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 15.55234146118164, "optim/lr": 0.002925, "optim/total_tokens": 1278214144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452847, "loss/out": 6.901112079620361, "created_at": "2025-01-14T19:52:43.934505+00:00"} {"global_step": 4877, "acc_step": 0, "speed/wps": 12963.704184292616, "speed/FLOPS": 203612788210638.16, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 177.6596221923828, "optim/lr": 0.0029256, "optim/total_tokens": 1278476288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 6.717927932739258, "created_at": "2025-01-14T19:52:48.994039+00:00"} {"global_step": 4878, "acc_step": 0, "speed/wps": 12971.272701775939, "speed/FLOPS": 203731662177946.62, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.457033157348633, "optim/lr": 0.0029262000000000003, "optim/total_tokens": 1278738432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443022, "loss/out": 6.731572151184082, "created_at": "2025-01-14T19:52:54.050086+00:00"} {"global_step": 4879, "acc_step": 0, "speed/wps": 12964.302086885418, "speed/FLOPS": 203622179092462.38, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 17.77887725830078, "optim/lr": 0.0029268000000000002, "optim/total_tokens": 1279000576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452615, "loss/out": 6.69075345993042, "created_at": "2025-01-14T19:52:59.111382+00:00"} {"global_step": 4880, "acc_step": 0, "speed/wps": 12962.202024870869, "speed/FLOPS": 203589194732736.25, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0012, "optim/grad_norm": 7.5128350257873535, "optim/lr": 0.0029274, "optim/total_tokens": 1279262720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457380, "loss/out": 6.719828128814697, "created_at": "2025-01-14T19:53:04.174153+00:00"} {"global_step": 4881, "acc_step": 0, "speed/wps": 12967.80456473389, "speed/FLOPS": 203677190319981.53, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 23.24917984008789, "optim/lr": 0.002928, "optim/total_tokens": 1279524864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313596, "loss/out": 6.709386825561523, "created_at": "2025-01-14T19:53:09.231631+00:00"} {"global_step": 4882, "acc_step": 0, "speed/wps": 12958.999370938282, "speed/FLOPS": 203538892651817.28, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 36.57487487792969, "optim/lr": 0.0029286, "optim/total_tokens": 1279787008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413463, "loss/out": 6.9625468254089355, "created_at": "2025-01-14T19:53:14.293909+00:00"} {"global_step": 4883, "acc_step": 0, "speed/wps": 12963.866016013486, "speed/FLOPS": 203615330000194.5, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.012836456298828, "optim/lr": 0.0029292000000000003, "optim/total_tokens": 1280049152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452847, "loss/out": 6.76790189743042, "created_at": "2025-01-14T19:53:19.355574+00:00"} {"global_step": 4884, "acc_step": 0, "speed/wps": 12969.214577026503, "speed/FLOPS": 203699336500596.97, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 22.393003463745117, "optim/lr": 0.0029298, "optim/total_tokens": 1280311296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 6.682334899902344, "created_at": "2025-01-14T19:53:24.415632+00:00"} {"global_step": 4885, "acc_step": 0, "speed/wps": 12970.63279277323, "speed/FLOPS": 203721611527731.97, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 41.303680419921875, "optim/lr": 0.0029304, "optim/total_tokens": 1280573440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461605, "loss/out": 6.782825946807861, "created_at": "2025-01-14T19:53:29.471963+00:00"} {"global_step": 4886, "acc_step": 0, "speed/wps": 12970.313495477392, "speed/FLOPS": 203716596524940.38, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.376866340637207, "optim/lr": 0.002931, "optim/total_tokens": 1280835584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426755, "loss/out": 6.877921104431152, "created_at": "2025-01-14T19:53:34.534343+00:00"} {"global_step": 4887, "acc_step": 0, "speed/wps": 12965.966729573616, "speed/FLOPS": 203648324593339.28, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.369994163513184, "optim/lr": 0.0029316, "optim/total_tokens": 1281097728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 6.8196611404418945, "created_at": "2025-01-14T19:53:39.597480+00:00"} {"global_step": 4888, "acc_step": 0, "speed/wps": 12961.223067498846, "speed/FLOPS": 203573818861980.25, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 74.27740478515625, "optim/lr": 0.0029322000000000003, "optim/total_tokens": 1281359872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324713, "loss/out": 6.750553607940674, "created_at": "2025-01-14T19:53:44.660524+00:00"} {"global_step": 4889, "acc_step": 0, "speed/wps": 12963.436209916568, "speed/FLOPS": 203608579304826.94, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0009, "optim/grad_norm": 56.84479904174805, "optim/lr": 0.0029328, "optim/total_tokens": 1281622016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367614, "loss/out": 7.029514789581299, "created_at": "2025-01-14T19:53:49.722695+00:00"} {"global_step": 4890, "acc_step": 0, "speed/wps": 12962.964313168279, "speed/FLOPS": 203601167517940.97, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 38.96941375732422, "optim/lr": 0.0029334, "optim/total_tokens": 1281884160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444878, "loss/out": 6.977316856384277, "created_at": "2025-01-14T19:53:54.781814+00:00"} {"global_step": 4891, "acc_step": 0, "speed/wps": 12956.983606797483, "speed/FLOPS": 203507232306035.84, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.016069412231445, "optim/lr": 0.002934, "optim/total_tokens": 1282146304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 6.832180023193359, "created_at": "2025-01-14T19:53:59.844070+00:00"} {"global_step": 4892, "acc_step": 0, "speed/wps": 12964.0380715829, "speed/FLOPS": 203618032369341.28, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 12.92048168182373, "optim/lr": 0.0029346, "optim/total_tokens": 1282408448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440624, "loss/out": 6.705101013183594, "created_at": "2025-01-14T19:54:04.902916+00:00"} {"global_step": 4893, "acc_step": 0, "speed/wps": 12956.07224351442, "speed/FLOPS": 203492918093330.97, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 20.038545608520508, "optim/lr": 0.0029352000000000002, "optim/total_tokens": 1282670592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 6.783290863037109, "created_at": "2025-01-14T19:54:09.964782+00:00"} {"global_step": 4894, "acc_step": 0, "speed/wps": 12956.3735448125, "speed/FLOPS": 203497650444248.7, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.159075736999512, "optim/lr": 0.0029358, "optim/total_tokens": 1282932736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 6.705348968505859, "created_at": "2025-01-14T19:54:15.029213+00:00"} {"global_step": 4895, "acc_step": 0, "speed/wps": 12957.589717771043, "speed/FLOPS": 203516752111759.38, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.201106071472168, "optim/lr": 0.0029364, "optim/total_tokens": 1283194880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 340231, "loss/out": 6.822058200836182, "created_at": "2025-01-14T19:54:20.095495+00:00"} {"global_step": 4896, "acc_step": 0, "speed/wps": 12954.30491578289, "speed/FLOPS": 203465159782744.47, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 21.804603576660156, "optim/lr": 0.002937, "optim/total_tokens": 1283457024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441971, "loss/out": 6.7387237548828125, "created_at": "2025-01-14T19:54:25.163009+00:00"} {"global_step": 4897, "acc_step": 0, "speed/wps": 12962.273061397686, "speed/FLOPS": 203590310459004.34, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 125.02386474609375, "optim/lr": 0.0029376, "optim/total_tokens": 1283719168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 6.692100524902344, "created_at": "2025-01-14T19:54:30.223677+00:00"} {"global_step": 4898, "acc_step": 0, "speed/wps": 12964.366876651447, "speed/FLOPS": 203623196704768.34, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.462499618530273, "optim/lr": 0.0029382, "optim/total_tokens": 1283981312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 6.815678119659424, "created_at": "2025-01-14T19:54:35.286877+00:00"} {"global_step": 4899, "acc_step": 0, "speed/wps": 12962.1417264516, "speed/FLOPS": 203588247663202.94, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 38.40065383911133, "optim/lr": 0.0029388, "optim/total_tokens": 1284243456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424963, "loss/out": 6.6606364250183105, "created_at": "2025-01-14T19:54:40.352453+00:00"} {"global_step": 4900, "acc_step": 0, "speed/wps": 12951.5845312145, "speed/FLOPS": 203422432404897.78, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 734.2031860351562, "optim/lr": 0.0029394, "optim/total_tokens": 1284505600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460034, "loss/out": 7.046679496765137, "created_at": "2025-01-14T19:54:45.419066+00:00"} {"global_step": 4901, "acc_step": 0, "speed/wps": 12964.893504421676, "speed/FLOPS": 203631468117562.16, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 468.2171325683594, "optim/lr": 0.00294, "optim/total_tokens": 1284767744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316820, "loss/out": 6.899439811706543, "created_at": "2025-01-14T19:54:50.479501+00:00"} {"global_step": 4902, "acc_step": 0, "speed/wps": 12953.928572566436, "speed/FLOPS": 203459248795380.28, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0012, "optim/grad_norm": 175.50270080566406, "optim/lr": 0.0029406, "optim/total_tokens": 1285029888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356961, "loss/out": 7.103164196014404, "created_at": "2025-01-14T19:54:55.556308+00:00"} {"global_step": 4903, "acc_step": 0, "speed/wps": 12964.514878529979, "speed/FLOPS": 203625521277646.56, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 101.32178497314453, "optim/lr": 0.0029412, "optim/total_tokens": 1285292032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462150, "loss/out": 7.273777961730957, "created_at": "2025-01-14T19:55:00.620925+00:00"} {"global_step": 4904, "acc_step": 0, "speed/wps": 12965.092761676224, "speed/FLOPS": 203634597726553.0, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.710332870483398, "optim/lr": 0.0029418, "optim/total_tokens": 1285554176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472789, "loss/out": 7.221589088439941, "created_at": "2025-01-14T19:55:05.683129+00:00"} {"global_step": 4905, "acc_step": 0, "speed/wps": 12962.79047606592, "speed/FLOPS": 203598437167371.22, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.001, "optim/grad_norm": 10.177633285522461, "optim/lr": 0.0029424, "optim/total_tokens": 1285816320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462672, "loss/out": 6.981752395629883, "created_at": "2025-01-14T19:55:10.747952+00:00"} {"global_step": 4906, "acc_step": 0, "speed/wps": 12966.608261725578, "speed/FLOPS": 203658400737343.4, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 43.18362045288086, "optim/lr": 0.002943, "optim/total_tokens": 1286078464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445170, "loss/out": 6.935373783111572, "created_at": "2025-01-14T19:55:15.810152+00:00"} {"global_step": 4907, "acc_step": 0, "speed/wps": 12964.860765246192, "speed/FLOPS": 203630953903822.84, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 24.8421688079834, "optim/lr": 0.0029435999999999998, "optim/total_tokens": 1286340608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298121, "loss/out": 6.9098734855651855, "created_at": "2025-01-14T19:55:20.869841+00:00"} {"global_step": 4908, "acc_step": 0, "speed/wps": 12961.016304282144, "speed/FLOPS": 203570571361539.5, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.478001832962036, "optim/lr": 0.0029442, "optim/total_tokens": 1286602752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433975, "loss/out": 6.956600189208984, "created_at": "2025-01-14T19:55:25.931028+00:00"} {"global_step": 4909, "acc_step": 0, "speed/wps": 12960.850690772744, "speed/FLOPS": 203567970173797.56, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 52.812286376953125, "optim/lr": 0.0029448, "optim/total_tokens": 1286864896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470641, "loss/out": 7.020140647888184, "created_at": "2025-01-14T19:55:30.991151+00:00"} {"global_step": 4910, "acc_step": 0, "speed/wps": 12971.355073349645, "speed/FLOPS": 203732955936702.1, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 425.5030822753906, "optim/lr": 0.0029454, "optim/total_tokens": 1287127040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 7.184394836425781, "created_at": "2025-01-14T19:55:36.053385+00:00"} {"global_step": 4911, "acc_step": 0, "speed/wps": 12962.205324658004, "speed/FLOPS": 203589246560427.6, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 72.09276580810547, "optim/lr": 0.002946, "optim/total_tokens": 1287389184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431320, "loss/out": 7.569645404815674, "created_at": "2025-01-14T19:55:41.124822+00:00"} {"global_step": 4912, "acc_step": 0, "speed/wps": 12966.950640152298, "speed/FLOPS": 203663778261012.28, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0009, "optim/grad_norm": 8.169479370117188, "optim/lr": 0.0029465999999999997, "optim/total_tokens": 1287651328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475696, "loss/out": 7.58292818069458, "created_at": "2025-01-14T19:55:46.185390+00:00"} {"global_step": 4913, "acc_step": 0, "speed/wps": 12963.69633653248, "speed/FLOPS": 203612664950781.25, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.7294504642486572, "optim/lr": 0.0029472, "optim/total_tokens": 1287913472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454979, "loss/out": 7.271599292755127, "created_at": "2025-01-14T19:55:51.244551+00:00"} {"global_step": 4914, "acc_step": 0, "speed/wps": 12969.216752175524, "speed/FLOPS": 203699370664301.3, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 25.203628540039062, "optim/lr": 0.0029478, "optim/total_tokens": 1288175616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322837, "loss/out": 7.132390975952148, "created_at": "2025-01-14T19:55:56.306712+00:00"} {"global_step": 4915, "acc_step": 0, "speed/wps": 12968.896673696578, "speed/FLOPS": 203694343391955.53, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0012, "optim/grad_norm": 23.32767677307129, "optim/lr": 0.0029484, "optim/total_tokens": 1288437760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367376, "loss/out": 7.2815446853637695, "created_at": "2025-01-14T19:56:01.364796+00:00"} {"global_step": 4916, "acc_step": 0, "speed/wps": 12966.419114549735, "speed/FLOPS": 203655429921031.34, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 22.018199920654297, "optim/lr": 0.0029490000000000002, "optim/total_tokens": 1288699904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441450, "loss/out": 7.169034004211426, "created_at": "2025-01-14T19:56:06.425678+00:00"} {"global_step": 4917, "acc_step": 0, "speed/wps": 12965.056321333948, "speed/FLOPS": 203634025380903.28, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 238.30018615722656, "optim/lr": 0.0029496, "optim/total_tokens": 1288962048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 7.060055732727051, "created_at": "2025-01-14T19:56:11.486122+00:00"} {"global_step": 4918, "acc_step": 0, "speed/wps": 12964.262684534164, "speed/FLOPS": 203621560224391.4, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.473937034606934, "optim/lr": 0.0029502, "optim/total_tokens": 1289224192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445691, "loss/out": 7.330756187438965, "created_at": "2025-01-14T19:56:16.550140+00:00"} {"global_step": 4919, "acc_step": 0, "speed/wps": 12968.59412762909, "speed/FLOPS": 203689591490223.78, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.4818172454833984, "optim/lr": 0.0029508, "optim/total_tokens": 1289486336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 7.13401985168457, "created_at": "2025-01-14T19:56:21.607372+00:00"} {"global_step": 4920, "acc_step": 0, "speed/wps": 12965.396971878621, "speed/FLOPS": 203639375765810.28, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.0135817527771, "optim/lr": 0.0029514000000000003, "optim/total_tokens": 1289748480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 7.048984527587891, "created_at": "2025-01-14T19:56:26.665772+00:00"} {"global_step": 4921, "acc_step": 0, "speed/wps": 12968.557509859904, "speed/FLOPS": 203689016357840.62, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.9301557540893555, "optim/lr": 0.002952, "optim/total_tokens": 1290010624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424156, "loss/out": 7.146426200866699, "created_at": "2025-01-14T19:56:31.752026+00:00"} {"global_step": 4922, "acc_step": 0, "speed/wps": 12966.47864042141, "speed/FLOPS": 203656364856643.1, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1710407733917236, "optim/lr": 0.0029526, "optim/total_tokens": 1290272768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468018, "loss/out": 7.108859539031982, "created_at": "2025-01-14T19:56:36.816479+00:00"} {"global_step": 4923, "acc_step": 0, "speed/wps": 12957.992225340287, "speed/FLOPS": 203523074046237.72, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.604566812515259, "optim/lr": 0.0029532, "optim/total_tokens": 1290534912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433718, "loss/out": 6.961859703063965, "created_at": "2025-01-14T19:56:41.882598+00:00"} {"global_step": 4924, "acc_step": 0, "speed/wps": 12958.281114398165, "speed/FLOPS": 203527611445866.84, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.5205262899398804, "optim/lr": 0.0029538000000000003, "optim/total_tokens": 1290797056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 7.048465728759766, "created_at": "2025-01-14T19:56:46.949691+00:00"} {"global_step": 4925, "acc_step": 0, "speed/wps": 12963.060775930655, "speed/FLOPS": 203602682598177.8, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.93709659576416, "optim/lr": 0.0029544000000000003, "optim/total_tokens": 1291059200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 7.311388969421387, "created_at": "2025-01-14T19:56:52.013704+00:00"} {"global_step": 4926, "acc_step": 0, "speed/wps": 12964.3104011298, "speed/FLOPS": 203622309679095.38, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.2893126010894775, "optim/lr": 0.002955, "optim/total_tokens": 1291321344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457095, "loss/out": 7.310997009277344, "created_at": "2025-01-14T19:56:57.072401+00:00"} {"global_step": 4927, "acc_step": 0, "speed/wps": 12966.875812918077, "speed/FLOPS": 203662602996475.47, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5920835733413696, "optim/lr": 0.0029556, "optim/total_tokens": 1291583488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317214, "loss/out": 7.15587043762207, "created_at": "2025-01-14T19:57:02.134259+00:00"} {"global_step": 4928, "acc_step": 0, "speed/wps": 12962.583022900271, "speed/FLOPS": 203595178830334.2, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.543492317199707, "optim/lr": 0.0029562000000000004, "optim/total_tokens": 1291845632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 403274, "loss/out": 7.219419479370117, "created_at": "2025-01-14T19:57:07.198911+00:00"} {"global_step": 4929, "acc_step": 0, "speed/wps": 12965.502594071828, "speed/FLOPS": 203641034707494.7, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 18.751937866210938, "optim/lr": 0.0029568000000000003, "optim/total_tokens": 1292107776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471217, "loss/out": 7.104196548461914, "created_at": "2025-01-14T19:57:12.262691+00:00"} {"global_step": 4930, "acc_step": 0, "speed/wps": 12963.115287160312, "speed/FLOPS": 203603538771945.84, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.589641809463501, "optim/lr": 0.0029574000000000002, "optim/total_tokens": 1292369920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444878, "loss/out": 7.1529717445373535, "created_at": "2025-01-14T19:57:17.323238+00:00"} {"global_step": 4931, "acc_step": 0, "speed/wps": 12972.206122607931, "speed/FLOPS": 203746322834767.72, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3376439809799194, "optim/lr": 0.002958, "optim/total_tokens": 1292632064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479416, "loss/out": 7.185859680175781, "created_at": "2025-01-14T19:57:22.382508+00:00"} {"global_step": 4932, "acc_step": 0, "speed/wps": 12972.625822688626, "speed/FLOPS": 203752914801263.34, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3641072511672974, "optim/lr": 0.0029586, "optim/total_tokens": 1292894208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 7.148975849151611, "created_at": "2025-01-14T19:57:27.442753+00:00"} {"global_step": 4933, "acc_step": 0, "speed/wps": 12965.002871091705, "speed/FLOPS": 203633185871400.53, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.6403238773345947, "optim/lr": 0.0029592000000000004, "optim/total_tokens": 1293156352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317059, "loss/out": 6.921029090881348, "created_at": "2025-01-14T19:57:32.500925+00:00"} {"global_step": 4934, "acc_step": 0, "speed/wps": 12967.616708929796, "speed/FLOPS": 203674239786437.3, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3438513278961182, "optim/lr": 0.0029598000000000003, "optim/total_tokens": 1293418496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424364, "loss/out": 6.9033613204956055, "created_at": "2025-01-14T19:57:37.562827+00:00"} {"global_step": 4935, "acc_step": 0, "speed/wps": 12964.916323528903, "speed/FLOPS": 203631826522997.9, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5505914688110352, "optim/lr": 0.0029604, "optim/total_tokens": 1293680640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463248, "loss/out": 6.931520938873291, "created_at": "2025-01-14T19:57:42.623078+00:00"} {"global_step": 4936, "acc_step": 0, "speed/wps": 12971.200820580807, "speed/FLOPS": 203730533185001.97, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.6496872901916504, "optim/lr": 0.002961, "optim/total_tokens": 1293942784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 6.917513847351074, "created_at": "2025-01-14T19:57:47.680262+00:00"} {"global_step": 4937, "acc_step": 0, "speed/wps": 12970.162265319916, "speed/FLOPS": 203714221247505.2, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 23.052936553955078, "optim/lr": 0.0029616, "optim/total_tokens": 1294204928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452847, "loss/out": 6.861548900604248, "created_at": "2025-01-14T19:57:52.742470+00:00"} {"global_step": 4938, "acc_step": 0, "speed/wps": 12967.442902351251, "speed/FLOPS": 203671509915286.0, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.752771854400635, "optim/lr": 0.0029622000000000003, "optim/total_tokens": 1294467072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431605, "loss/out": 6.931941986083984, "created_at": "2025-01-14T19:57:57.805220+00:00"} {"global_step": 4939, "acc_step": 0, "speed/wps": 12967.905229103506, "speed/FLOPS": 203678771392233.7, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.508495330810547, "optim/lr": 0.0029628000000000002, "optim/total_tokens": 1294729216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456346, "loss/out": 6.90745735168457, "created_at": "2025-01-14T19:58:02.863172+00:00"} {"global_step": 4940, "acc_step": 0, "speed/wps": 12969.53303856083, "speed/FLOPS": 203704338376605.6, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.769190311431885, "optim/lr": 0.0029634, "optim/total_tokens": 1294991360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 6.849124908447266, "created_at": "2025-01-14T19:58:07.919861+00:00"} {"global_step": 4941, "acc_step": 0, "speed/wps": 12968.349152690149, "speed/FLOPS": 203685743822176.88, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8117096424102783, "optim/lr": 0.002964, "optim/total_tokens": 1295253504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461368, "loss/out": 6.780022621154785, "created_at": "2025-01-14T19:58:12.982583+00:00"} {"global_step": 4942, "acc_step": 0, "speed/wps": 12965.361714697385, "speed/FLOPS": 203638822003330.5, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.914414882659912, "optim/lr": 0.0029646, "optim/total_tokens": 1295515648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301284, "loss/out": 6.782467365264893, "created_at": "2025-01-14T19:58:18.043966+00:00"} {"global_step": 4943, "acc_step": 0, "speed/wps": 12971.944166996433, "speed/FLOPS": 203742208461924.38, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0011, "optim/grad_norm": 2.5550734996795654, "optim/lr": 0.0029652, "optim/total_tokens": 1295777792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 398431, "loss/out": 6.73494815826416, "created_at": "2025-01-14T19:58:23.099608+00:00"} {"global_step": 4944, "acc_step": 0, "speed/wps": 12974.585456821042, "speed/FLOPS": 203783693548132.44, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1725382804870605, "optim/lr": 0.0029658, "optim/total_tokens": 1296039936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443307, "loss/out": 6.758237361907959, "created_at": "2025-01-14T19:58:28.154368+00:00"} {"global_step": 4945, "acc_step": 0, "speed/wps": 12974.84600237121, "speed/FLOPS": 203787785774024.94, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.29754638671875, "optim/lr": 0.0029664, "optim/total_tokens": 1296302080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 6.6878252029418945, "created_at": "2025-01-14T19:58:33.221778+00:00"} {"global_step": 4946, "acc_step": 0, "speed/wps": 12969.143803812098, "speed/FLOPS": 203698224910012.16, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.541944742202759, "optim/lr": 0.002967, "optim/total_tokens": 1296564224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462701, "loss/out": 6.783395290374756, "created_at": "2025-01-14T19:58:38.286861+00:00"} {"global_step": 4947, "acc_step": 0, "speed/wps": 12974.690171382148, "speed/FLOPS": 203785338234207.5, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.9624180793762207, "optim/lr": 0.0029676, "optim/total_tokens": 1296826368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455808, "loss/out": 6.710733413696289, "created_at": "2025-01-14T19:58:43.342945+00:00"} {"global_step": 4948, "acc_step": 0, "speed/wps": 12961.668204946049, "speed/FLOPS": 203580810357271.5, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.2731120586395264, "optim/lr": 0.0029682, "optim/total_tokens": 1297088512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323215, "loss/out": 6.632946014404297, "created_at": "2025-01-14T19:58:48.403888+00:00"} {"global_step": 4949, "acc_step": 0, "speed/wps": 12972.480696655668, "speed/FLOPS": 203750635397491.7, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9488362073898315, "optim/lr": 0.0029688, "optim/total_tokens": 1297350656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351876, "loss/out": 6.655394554138184, "created_at": "2025-01-14T19:58:53.462074+00:00"} {"global_step": 4950, "acc_step": 0, "speed/wps": 12971.594939592476, "speed/FLOPS": 203736723365656.5, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.54261314868927, "optim/lr": 0.0029694, "optim/total_tokens": 1297612800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457380, "loss/out": 6.556674480438232, "created_at": "2025-01-14T19:58:58.523168+00:00"} {"global_step": 4951, "acc_step": 0, "speed/wps": 12970.455579828658, "speed/FLOPS": 203718828154921.38, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.539243459701538, "optim/lr": 0.00297, "optim/total_tokens": 1297874944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446742, "loss/out": 6.6993913650512695, "created_at": "2025-01-14T19:59:03.583405+00:00"} {"global_step": 4952, "acc_step": 0, "speed/wps": 12957.298060898614, "speed/FLOPS": 203512171239799.34, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.395719051361084, "optim/lr": 0.0029706, "optim/total_tokens": 1298137088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 6.693848133087158, "created_at": "2025-01-14T19:59:08.647300+00:00"} {"global_step": 4953, "acc_step": 0, "speed/wps": 12960.698712023705, "speed/FLOPS": 203565583138703.22, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.799500465393066, "optim/lr": 0.0029712, "optim/total_tokens": 1298399232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449411, "loss/out": 6.666055679321289, "created_at": "2025-01-14T19:59:13.710240+00:00"} {"global_step": 4954, "acc_step": 0, "speed/wps": 12953.141760026803, "speed/FLOPS": 203446890823251.62, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.6445345878601074, "optim/lr": 0.0029718, "optim/total_tokens": 1298661376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 6.7325544357299805, "created_at": "2025-01-14T19:59:18.775057+00:00"} {"global_step": 4955, "acc_step": 0, "speed/wps": 12958.987808310883, "speed/FLOPS": 203538711044865.28, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.814663887023926, "optim/lr": 0.0029724, "optim/total_tokens": 1298923520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432365, "loss/out": 6.689459800720215, "created_at": "2025-01-14T19:59:23.868559+00:00"} {"global_step": 4956, "acc_step": 0, "speed/wps": 12965.386666317772, "speed/FLOPS": 203639213902815.8, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.123242139816284, "optim/lr": 0.002973, "optim/total_tokens": 1299185664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 6.601953983306885, "created_at": "2025-01-14T19:59:28.931413+00:00"} {"global_step": 4957, "acc_step": 0, "speed/wps": 12973.73862090016, "speed/FLOPS": 203770392826320.5, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.353939056396484, "optim/lr": 0.0029736, "optim/total_tokens": 1299447808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445695, "loss/out": 6.712177276611328, "created_at": "2025-01-14T19:59:33.991722+00:00"} {"global_step": 4958, "acc_step": 0, "speed/wps": 12967.07579538667, "speed/FLOPS": 203665743995949.44, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.4761316776275635, "optim/lr": 0.0029741999999999998, "optim/total_tokens": 1299709952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481190, "loss/out": 6.706661701202393, "created_at": "2025-01-14T19:59:39.053872+00:00"} {"global_step": 4959, "acc_step": 0, "speed/wps": 12965.325865275538, "speed/FLOPS": 203638258938898.12, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.857462763786316, "optim/lr": 0.0029748, "optim/total_tokens": 1299972096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445404, "loss/out": 6.681491374969482, "created_at": "2025-01-14T19:59:44.112426+00:00"} {"global_step": 4960, "acc_step": 0, "speed/wps": 12967.625603340859, "speed/FLOPS": 203674379485385.28, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.7222617864608765, "optim/lr": 0.0029754, "optim/total_tokens": 1300234240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319225, "loss/out": 6.544875144958496, "created_at": "2025-01-14T19:59:49.174140+00:00"} {"global_step": 4961, "acc_step": 0, "speed/wps": 12972.091466703749, "speed/FLOPS": 203744522006238.78, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 22.040603637695312, "optim/lr": 0.002976, "optim/total_tokens": 1300496384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444881, "loss/out": 6.591460227966309, "created_at": "2025-01-14T19:59:54.235523+00:00"} {"global_step": 4962, "acc_step": 0, "speed/wps": 12969.385285418031, "speed/FLOPS": 203702017710463.8, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.179574489593506, "optim/lr": 0.0029766, "optim/total_tokens": 1300758528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 6.53841495513916, "created_at": "2025-01-14T19:59:59.299516+00:00"} {"global_step": 4963, "acc_step": 0, "speed/wps": 12971.388227904463, "speed/FLOPS": 203733476674544.62, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.548738956451416, "optim/lr": 0.0029771999999999997, "optim/total_tokens": 1301020672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446742, "loss/out": 6.627058029174805, "created_at": "2025-01-14T20:00:04.358885+00:00"} {"global_step": 4964, "acc_step": 0, "speed/wps": 12972.80359717704, "speed/FLOPS": 203755706993891.66, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 11.906570434570312, "optim/lr": 0.0029778, "optim/total_tokens": 1301282816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435338, "loss/out": 6.725441932678223, "created_at": "2025-01-14T20:00:09.440391+00:00"} {"global_step": 4965, "acc_step": 0, "speed/wps": 12970.473197147212, "speed/FLOPS": 203719104859118.0, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0014, "optim/grad_norm": 2.1379635334014893, "optim/lr": 0.0029784, "optim/total_tokens": 1301544960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447839, "loss/out": 6.467568397521973, "created_at": "2025-01-14T20:00:14.498029+00:00"} {"global_step": 4966, "acc_step": 0, "speed/wps": 12971.803659326173, "speed/FLOPS": 203740001595885.2, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.4653029441833496, "optim/lr": 0.002979, "optim/total_tokens": 1301807104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448606, "loss/out": 6.505438327789307, "created_at": "2025-01-14T20:00:19.559098+00:00"} {"global_step": 4967, "acc_step": 0, "speed/wps": 12975.245206872896, "speed/FLOPS": 203794055829288.38, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.345993995666504, "optim/lr": 0.0029796, "optim/total_tokens": 1302069248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350820, "loss/out": 6.502627372741699, "created_at": "2025-01-14T20:00:24.613458+00:00"} {"global_step": 4968, "acc_step": 0, "speed/wps": 12963.746539911905, "speed/FLOPS": 203613453463841.8, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 21.270858764648438, "optim/lr": 0.0029802, "optim/total_tokens": 1302331392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467734, "loss/out": 6.435576438903809, "created_at": "2025-01-14T20:00:29.675594+00:00"} {"global_step": 4969, "acc_step": 0, "speed/wps": 12974.527399482975, "speed/FLOPS": 203782781677858.97, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.2148218154907227, "optim/lr": 0.0029808, "optim/total_tokens": 1302593536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 6.469451904296875, "created_at": "2025-01-14T20:00:34.735198+00:00"} {"global_step": 4970, "acc_step": 0, "speed/wps": 12963.049755294867, "speed/FLOPS": 203602509503948.84, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.4656026363372803, "optim/lr": 0.0029814, "optim/total_tokens": 1302855680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434230, "loss/out": 6.56291389465332, "created_at": "2025-01-14T20:00:39.794583+00:00"} {"global_step": 4971, "acc_step": 0, "speed/wps": 12972.84664605456, "speed/FLOPS": 203756383135668.47, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.132875442504883, "optim/lr": 0.002982, "optim/total_tokens": 1303117824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 6.55023193359375, "created_at": "2025-01-14T20:00:44.853773+00:00"} {"global_step": 4972, "acc_step": 0, "speed/wps": 12966.634374575991, "speed/FLOPS": 203658810875541.97, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.367431640625, "optim/lr": 0.0029826, "optim/total_tokens": 1303379968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440628, "loss/out": 6.565637588500977, "created_at": "2025-01-14T20:00:49.916031+00:00"} {"global_step": 4973, "acc_step": 0, "speed/wps": 12965.299346816313, "speed/FLOPS": 203637842430055.84, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.265397071838379, "optim/lr": 0.0029832, "optim/total_tokens": 1303642112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300577, "loss/out": 6.5407185554504395, "created_at": "2025-01-14T20:00:54.980090+00:00"} {"global_step": 4974, "acc_step": 0, "speed/wps": 12966.611241199123, "speed/FLOPS": 203658447534069.66, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.26035213470459, "optim/lr": 0.0029838, "optim/total_tokens": 1303904256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443306, "loss/out": 6.577012062072754, "created_at": "2025-01-14T20:01:00.040116+00:00"} {"global_step": 4975, "acc_step": 0, "speed/wps": 12962.987151752746, "speed/FLOPS": 203601526229293.53, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.635505676269531, "optim/lr": 0.0029844, "optim/total_tokens": 1304166400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 6.520323753356934, "created_at": "2025-01-14T20:01:05.102112+00:00"} {"global_step": 4976, "acc_step": 0, "speed/wps": 12964.540830082276, "speed/FLOPS": 203625928882436.56, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 57.573333740234375, "optim/lr": 0.0029850000000000002, "optim/total_tokens": 1304428544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445695, "loss/out": 6.562061786651611, "created_at": "2025-01-14T20:01:10.164457+00:00"} {"global_step": 4977, "acc_step": 0, "speed/wps": 12967.589662539309, "speed/FLOPS": 203673814985709.1, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 32.274349212646484, "optim/lr": 0.0029856, "optim/total_tokens": 1304690688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444119, "loss/out": 6.562216758728027, "created_at": "2025-01-14T20:01:15.221967+00:00"} {"global_step": 4978, "acc_step": 0, "speed/wps": 12970.113403568952, "speed/FLOPS": 203713453806563.22, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.903684616088867, "optim/lr": 0.0029862, "optim/total_tokens": 1304952832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463485, "loss/out": 6.615097999572754, "created_at": "2025-01-14T20:01:20.279600+00:00"} {"global_step": 4979, "acc_step": 0, "speed/wps": 12971.742334190543, "speed/FLOPS": 203739038400360.56, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 19.470640182495117, "optim/lr": 0.0029868000000000004, "optim/total_tokens": 1305214976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 6.445932388305664, "created_at": "2025-01-14T20:01:25.338579+00:00"} {"global_step": 4980, "acc_step": 0, "speed/wps": 12970.777789001644, "speed/FLOPS": 203723888892744.9, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.8367185592651367, "optim/lr": 0.0029874000000000003, "optim/total_tokens": 1305477120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439302, "loss/out": 6.54878044128418, "created_at": "2025-01-14T20:01:30.397552+00:00"} {"global_step": 4981, "acc_step": 0, "speed/wps": 12954.999372502352, "speed/FLOPS": 203476067179807.16, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 29.902605056762695, "optim/lr": 0.002988, "optim/total_tokens": 1305739264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354562, "loss/out": 6.377048492431641, "created_at": "2025-01-14T20:01:35.462944+00:00"} {"global_step": 4982, "acc_step": 0, "speed/wps": 12963.807400179176, "speed/FLOPS": 203614409357970.06, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.757725238800049, "optim/lr": 0.0029886, "optim/total_tokens": 1306001408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441450, "loss/out": 6.587324142456055, "created_at": "2025-01-14T20:01:40.523181+00:00"} {"global_step": 4983, "acc_step": 0, "speed/wps": 12969.188494417269, "speed/FLOPS": 203698926837377.7, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 152.1013946533203, "optim/lr": 0.0029892, "optim/total_tokens": 1306263552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 6.59938907623291, "created_at": "2025-01-14T20:01:45.581030+00:00"} {"global_step": 4984, "acc_step": 0, "speed/wps": 12969.926568559871, "speed/FLOPS": 203710519306007.4, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 20.8971004486084, "optim/lr": 0.0029898000000000004, "optim/total_tokens": 1306525696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441158, "loss/out": 6.492408752441406, "created_at": "2025-01-14T20:01:50.640332+00:00"} {"global_step": 4985, "acc_step": 0, "speed/wps": 12964.979846250537, "speed/FLOPS": 203632824234630.53, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 28.44501495361328, "optim/lr": 0.0029904000000000003, "optim/total_tokens": 1306787840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 6.554598331451416, "created_at": "2025-01-14T20:01:55.700606+00:00"} {"global_step": 4986, "acc_step": 0, "speed/wps": 12971.20171959892, "speed/FLOPS": 203730547305316.8, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.1437177658081055, "optim/lr": 0.002991, "optim/total_tokens": 1307049984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 6.630583763122559, "created_at": "2025-01-14T20:02:00.759486+00:00"} {"global_step": 4987, "acc_step": 0, "speed/wps": 12968.841987156331, "speed/FLOPS": 203693484464693.78, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.927814960479736, "optim/lr": 0.0029916, "optim/total_tokens": 1307312128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326870, "loss/out": 6.426669120788574, "created_at": "2025-01-14T20:02:05.816449+00:00"} {"global_step": 4988, "acc_step": 0, "speed/wps": 12972.304215205151, "speed/FLOPS": 203747863513800.4, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 88.68047332763672, "optim/lr": 0.0029922, "optim/total_tokens": 1307574272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 346546, "loss/out": 6.555879592895508, "created_at": "2025-01-14T20:02:10.876773+00:00"} {"global_step": 4989, "acc_step": 0, "speed/wps": 12967.905869567428, "speed/FLOPS": 203678781451599.7, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.852819919586182, "optim/lr": 0.0029928000000000003, "optim/total_tokens": 1307836416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448361, "loss/out": 6.511415004730225, "created_at": "2025-01-14T20:02:15.934558+00:00"} {"global_step": 4990, "acc_step": 0, "speed/wps": 12970.54940234343, "speed/FLOPS": 203720301766441.5, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.228132724761963, "optim/lr": 0.0029934000000000002, "optim/total_tokens": 1308098560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442256, "loss/out": 6.435147762298584, "created_at": "2025-01-14T20:02:20.996412+00:00"} {"global_step": 4991, "acc_step": 0, "speed/wps": 12972.181603875315, "speed/FLOPS": 203745937734379.7, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.056772232055664, "optim/lr": 0.002994, "optim/total_tokens": 1308360704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 6.509485721588135, "created_at": "2025-01-14T20:02:26.052491+00:00"} {"global_step": 4992, "acc_step": 0, "speed/wps": 12970.206731052458, "speed/FLOPS": 203714919642937.1, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 51.08488845825195, "optim/lr": 0.0029946, "optim/total_tokens": 1308622848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495573, "loss/out": 6.476512432098389, "created_at": "2025-01-14T20:02:31.108839+00:00"} {"global_step": 4993, "acc_step": 0, "speed/wps": 12971.577340746848, "speed/FLOPS": 203736446951602.56, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.741966247558594, "optim/lr": 0.0029952, "optim/total_tokens": 1308884992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460034, "loss/out": 6.626613140106201, "created_at": "2025-01-14T20:02:36.169174+00:00"} {"global_step": 4994, "acc_step": 0, "speed/wps": 12965.669640418762, "speed/FLOPS": 203643658399919.03, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 38.494468688964844, "optim/lr": 0.0029958000000000003, "optim/total_tokens": 1309147136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451559, "loss/out": 6.599359035491943, "created_at": "2025-01-14T20:02:41.232211+00:00"} {"global_step": 4995, "acc_step": 0, "speed/wps": 12968.816147448479, "speed/FLOPS": 203693078616574.47, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 35.15415954589844, "optim/lr": 0.0029964, "optim/total_tokens": 1309409280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451547, "loss/out": 6.517385482788086, "created_at": "2025-01-14T20:02:46.293939+00:00"} {"global_step": 4996, "acc_step": 0, "speed/wps": 12968.67637675389, "speed/FLOPS": 203690883325750.9, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 22.48912811279297, "optim/lr": 0.002997, "optim/total_tokens": 1309671424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312537, "loss/out": 6.519833564758301, "created_at": "2025-01-14T20:02:51.356486+00:00"} {"global_step": 4997, "acc_step": 0, "speed/wps": 12963.773799547354, "speed/FLOPS": 203613881613875.22, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 18.63835906982422, "optim/lr": 0.0029976, "optim/total_tokens": 1309933568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 6.5315399169921875, "created_at": "2025-01-14T20:02:56.415470+00:00"} {"global_step": 4998, "acc_step": 0, "speed/wps": 12978.208878364274, "speed/FLOPS": 203840604362571.88, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 17.037282943725586, "optim/lr": 0.0029982, "optim/total_tokens": 1310195712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445166, "loss/out": 6.480851173400879, "created_at": "2025-01-14T20:03:01.473329+00:00"} {"global_step": 4999, "acc_step": 0, "speed/wps": 12969.472174720051, "speed/FLOPS": 203703382426350.47, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 17.360937118530273, "optim/lr": 0.0029988000000000003, "optim/total_tokens": 1310457856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 6.535923957824707, "created_at": "2025-01-14T20:03:06.538478+00:00"} {"global_step": 5000, "acc_step": 0, "speed/wps": 12967.735838176182, "speed/FLOPS": 203676110874952.56, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.821642875671387, "optim/lr": 0.0029994, "optim/total_tokens": 1310720000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459237, "loss/out": 6.3874006271362305, "created_at": "2025-01-14T20:03:11.599707+00:00"} {"global_step": 5001, "acc_step": 0, "speed/wps": 3474.245532007098, "speed/FLOPS": 54567800193823.81, "speed/curr_iter_time": 1.2616, "speed/data_load_time": 0.2692, "optim/grad_norm": 5.334745407104492, "optim/lr": 0.003, "optim/total_tokens": 1310982144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292963, "loss/out": 6.489900588989258, "created_at": "2025-01-14T20:03:30.468075+00:00"} {"global_step": 5002, "acc_step": 0, "speed/wps": 13036.162055726001, "speed/FLOPS": 204750838649048.75, "speed/curr_iter_time": 1.2693, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.884615898132324, "optim/lr": 0.002999999997552993, "optim/total_tokens": 1311244288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452093, "loss/out": 6.541272163391113, "created_at": "2025-01-14T20:03:35.508711+00:00"} {"global_step": 5003, "acc_step": 0, "speed/wps": 13010.90761100433, "speed/FLOPS": 204354182891451.22, "speed/curr_iter_time": 1.2707, "speed/data_load_time": 0.001, "optim/grad_norm": 21.216278076171875, "optim/lr": 0.0029999999902119725, "optim/total_tokens": 1311506432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440921, "loss/out": 6.455909252166748, "created_at": "2025-01-14T20:03:40.550369+00:00"} {"global_step": 5004, "acc_step": 0, "speed/wps": 13007.459122131306, "speed/FLOPS": 204300019634980.1, "speed/curr_iter_time": 1.2697, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.790366172790527, "optim/lr": 0.002999999977976938, "optim/total_tokens": 1311768576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434807, "loss/out": 6.434762954711914, "created_at": "2025-01-14T20:03:45.597592+00:00"} {"global_step": 5005, "acc_step": 0, "speed/wps": 12995.444306816964, "speed/FLOPS": 204111310450381.97, "speed/curr_iter_time": 1.2709, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.685356140136719, "optim/lr": 0.0029999999608478897, "optim/total_tokens": 1312030720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318405, "loss/out": 6.383855819702148, "created_at": "2025-01-14T20:03:50.649495+00:00"} {"global_step": 5006, "acc_step": 0, "speed/wps": 12992.425022300118, "speed/FLOPS": 204063888438113.2, "speed/curr_iter_time": 1.2716, "speed/data_load_time": 0.0006, "optim/grad_norm": 149.04942321777344, "optim/lr": 0.002999999938824828, "optim/total_tokens": 1312292864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336239, "loss/out": 6.4225029945373535, "created_at": "2025-01-14T20:03:55.769305+00:00"} {"global_step": 5007, "acc_step": 0, "speed/wps": 12995.077398192285, "speed/FLOPS": 204105547646246.1, "speed/curr_iter_time": 1.2712, "speed/data_load_time": 0.0006, "optim/grad_norm": 157.59356689453125, "optim/lr": 0.0029999999119077517, "optim/total_tokens": 1312555008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 6.429601669311523, "created_at": "2025-01-14T20:04:00.816538+00:00"} {"global_step": 5008, "acc_step": 0, "speed/wps": 12987.994119864497, "speed/FLOPS": 203994295026663.78, "speed/curr_iter_time": 1.2715, "speed/data_load_time": 0.0007, "optim/grad_norm": 26.802112579345703, "optim/lr": 0.0029999998800966632, "optim/total_tokens": 1312817152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446505, "loss/out": 6.573607444763184, "created_at": "2025-01-14T20:04:05.869811+00:00"} {"global_step": 5009, "acc_step": 0, "speed/wps": 12985.008939452284, "speed/FLOPS": 203947408666222.22, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.852649688720703, "optim/lr": 0.0029999998433915603, "optim/total_tokens": 1313079296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444881, "loss/out": 6.578507900238037, "created_at": "2025-01-14T20:04:10.923133+00:00"} {"global_step": 5010, "acc_step": 0, "speed/wps": 12989.19080474775, "speed/FLOPS": 204013090607172.62, "speed/curr_iter_time": 1.2716, "speed/data_load_time": 0.0007, "optim/grad_norm": 70.0043716430664, "optim/lr": 0.0029999998017924447, "optim/total_tokens": 1313341440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 6.53277587890625, "created_at": "2025-01-14T20:04:15.973234+00:00"} {"global_step": 5011, "acc_step": 0, "speed/wps": 12982.594600160603, "speed/FLOPS": 203909488149996.5, "speed/curr_iter_time": 1.2719, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.523502349853516, "optim/lr": 0.002999999755299316, "optim/total_tokens": 1313603584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 339173, "loss/out": 6.526205062866211, "created_at": "2025-01-14T20:04:21.028719+00:00"} {"global_step": 5012, "acc_step": 0, "speed/wps": 12981.525163324139, "speed/FLOPS": 203892691174919.62, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0008, "optim/grad_norm": 9.846352577209473, "optim/lr": 0.002999999703912174, "optim/total_tokens": 1313865728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420140, "loss/out": 6.477242469787598, "created_at": "2025-01-14T20:04:26.081478+00:00"} {"global_step": 5013, "acc_step": 0, "speed/wps": 12978.24886579806, "speed/FLOPS": 203841232420167.97, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.967940330505371, "optim/lr": 0.0029999996476310193, "optim/total_tokens": 1314127872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441157, "loss/out": 6.442215919494629, "created_at": "2025-01-14T20:04:31.135361+00:00"} {"global_step": 5014, "acc_step": 0, "speed/wps": 12977.453924232235, "speed/FLOPS": 203828746770512.5, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.60677433013916, "optim/lr": 0.002999999586455852, "optim/total_tokens": 1314390016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 6.470346450805664, "created_at": "2025-01-14T20:04:36.193841+00:00"} {"global_step": 5015, "acc_step": 0, "speed/wps": 12980.055654400096, "speed/FLOPS": 203869610517950.66, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.610786437988281, "optim/lr": 0.002999999520386672, "optim/total_tokens": 1314652160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 6.541499137878418, "created_at": "2025-01-14T20:04:41.247398+00:00"} {"global_step": 5016, "acc_step": 0, "speed/wps": 12969.52082367618, "speed/FLOPS": 203704146525057.12, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 9.040163040161133, "optim/lr": 0.0029999994494234795, "optim/total_tokens": 1314914304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475100, "loss/out": 6.482975482940674, "created_at": "2025-01-14T20:04:46.303940+00:00"} {"global_step": 5017, "acc_step": 0, "speed/wps": 12982.336270167378, "speed/FLOPS": 203905430722470.3, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.762166976928711, "optim/lr": 0.0029999993735662753, "optim/total_tokens": 1315176448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462963, "loss/out": 6.503973960876465, "created_at": "2025-01-14T20:04:51.361260+00:00"} {"global_step": 5018, "acc_step": 0, "speed/wps": 12974.43158890688, "speed/FLOPS": 203781276841103.88, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.303279876708984, "optim/lr": 0.0029999992928150592, "optim/total_tokens": 1315438592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458430, "loss/out": 6.476573467254639, "created_at": "2025-01-14T20:04:56.418354+00:00"} {"global_step": 5019, "acc_step": 0, "speed/wps": 12977.731166548898, "speed/FLOPS": 203833101242066.56, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.575662136077881, "optim/lr": 0.0029999992071698314, "optim/total_tokens": 1315700736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 6.35786771774292, "created_at": "2025-01-14T20:05:01.477577+00:00"} {"global_step": 5020, "acc_step": 0, "speed/wps": 12972.57282658207, "speed/FLOPS": 203752082424585.62, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.091256618499756, "optim/lr": 0.0029999991166305925, "optim/total_tokens": 1315962880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330091, "loss/out": 6.406370162963867, "created_at": "2025-01-14T20:05:06.534267+00:00"} {"global_step": 5021, "acc_step": 0, "speed/wps": 12977.898939827839, "speed/FLOPS": 203835736351956.97, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.71825885772705, "optim/lr": 0.002999999021197343, "optim/total_tokens": 1316225024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308039, "loss/out": 6.397779941558838, "created_at": "2025-01-14T20:05:11.590126+00:00"} {"global_step": 5022, "acc_step": 0, "speed/wps": 12977.967711922074, "speed/FLOPS": 203836816512199.72, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.546818256378174, "optim/lr": 0.0029999989208700827, "optim/total_tokens": 1316487168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436366, "loss/out": 6.393786430358887, "created_at": "2025-01-14T20:05:16.645090+00:00"} {"global_step": 5023, "acc_step": 0, "speed/wps": 12980.550099776538, "speed/FLOPS": 203877376462026.84, "speed/curr_iter_time": 1.2714, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.879927635192871, "optim/lr": 0.002999998815648812, "optim/total_tokens": 1316749312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445404, "loss/out": 6.29728364944458, "created_at": "2025-01-14T20:05:21.697777+00:00"} {"global_step": 5024, "acc_step": 0, "speed/wps": 12966.462704044321, "speed/FLOPS": 203656114553942.06, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.933480978012085, "optim/lr": 0.0029999987055335316, "optim/total_tokens": 1317011456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 6.314146518707275, "created_at": "2025-01-14T20:05:26.781108+00:00"} {"global_step": 5025, "acc_step": 0, "speed/wps": 12979.243518207857, "speed/FLOPS": 203856854803059.47, "speed/curr_iter_time": 1.272, "speed/data_load_time": 0.0012, "optim/grad_norm": 23.65226173400879, "optim/lr": 0.0029999985905242413, "optim/total_tokens": 1317273600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 6.5535078048706055, "created_at": "2025-01-14T20:05:31.834066+00:00"} {"global_step": 5026, "acc_step": 0, "speed/wps": 12978.719326669776, "speed/FLOPS": 203848621654639.25, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.959864616394043, "optim/lr": 0.002999998470620942, "optim/total_tokens": 1317535744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448361, "loss/out": 6.568107604980469, "created_at": "2025-01-14T20:05:36.887104+00:00"} {"global_step": 5027, "acc_step": 0, "speed/wps": 12977.27022583069, "speed/FLOPS": 203825861534688.22, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.293072700500488, "optim/lr": 0.002999998345823634, "optim/total_tokens": 1317797888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447263, "loss/out": 6.520144939422607, "created_at": "2025-01-14T20:05:41.940665+00:00"} {"global_step": 5028, "acc_step": 0, "speed/wps": 12970.807624481115, "speed/FLOPS": 203724357499948.25, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.670802593231201, "optim/lr": 0.002999998216132317, "optim/total_tokens": 1318060032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447554, "loss/out": 6.496731281280518, "created_at": "2025-01-14T20:05:47.003654+00:00"} {"global_step": 5029, "acc_step": 0, "speed/wps": 12975.300825482074, "speed/FLOPS": 203794929395973.5, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.870133399963379, "optim/lr": 0.0029999980815469925, "optim/total_tokens": 1318322176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444359, "loss/out": 6.461740493774414, "created_at": "2025-01-14T20:05:52.060559+00:00"} {"global_step": 5030, "acc_step": 0, "speed/wps": 12975.679363850533, "speed/FLOPS": 203800874861216.38, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 46.11066436767578, "optim/lr": 0.0029999979420676603, "optim/total_tokens": 1318584320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318166, "loss/out": 6.4772796630859375, "created_at": "2025-01-14T20:05:57.117577+00:00"} {"global_step": 5031, "acc_step": 0, "speed/wps": 12974.309169397706, "speed/FLOPS": 203779354074490.25, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 15.296009063720703, "optim/lr": 0.002999997797694321, "optim/total_tokens": 1318846464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372786, "loss/out": 6.568942070007324, "created_at": "2025-01-14T20:06:02.180377+00:00"} {"global_step": 5032, "acc_step": 0, "speed/wps": 12968.395711556195, "speed/FLOPS": 203686475093146.72, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 41.31781005859375, "optim/lr": 0.002999997648426975, "optim/total_tokens": 1319108608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446220, "loss/out": 6.5669708251953125, "created_at": "2025-01-14T20:06:07.237422+00:00"} {"global_step": 5033, "acc_step": 0, "speed/wps": 12970.562412901243, "speed/FLOPS": 203720506115130.34, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0009, "optim/grad_norm": 6.848090171813965, "optim/lr": 0.002999997494265623, "optim/total_tokens": 1319370752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 6.709990978240967, "created_at": "2025-01-14T20:06:12.296721+00:00"} {"global_step": 5034, "acc_step": 0, "speed/wps": 12964.266823716378, "speed/FLOPS": 203621625235935.94, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.001, "optim/grad_norm": 30.18943214416504, "optim/lr": 0.0029999973352102657, "optim/total_tokens": 1319632896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422528, "loss/out": 6.503556251525879, "created_at": "2025-01-14T20:06:17.361727+00:00"} {"global_step": 5035, "acc_step": 0, "speed/wps": 12972.429940139717, "speed/FLOPS": 203749838196661.9, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.942409515380859, "optim/lr": 0.002999997171260903, "optim/total_tokens": 1319895040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 6.49235200881958, "created_at": "2025-01-14T20:06:22.417185+00:00"} {"global_step": 5036, "acc_step": 0, "speed/wps": 12973.6442929954, "speed/FLOPS": 203768911276956.1, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 15.68104076385498, "optim/lr": 0.0029999970024175353, "optim/total_tokens": 1320157184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463248, "loss/out": 6.565594673156738, "created_at": "2025-01-14T20:06:27.477198+00:00"} {"global_step": 5037, "acc_step": 0, "speed/wps": 12972.512881237231, "speed/FLOPS": 203751140900571.4, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.655823707580566, "optim/lr": 0.002999996828680164, "optim/total_tokens": 1320419328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 6.524444103240967, "created_at": "2025-01-14T20:06:32.535511+00:00"} {"global_step": 5038, "acc_step": 0, "speed/wps": 12967.932123863122, "speed/FLOPS": 203679193811390.75, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 40.178096771240234, "optim/lr": 0.002999996650048789, "optim/total_tokens": 1320681472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413993, "loss/out": 6.658091068267822, "created_at": "2025-01-14T20:06:37.595518+00:00"} {"global_step": 5039, "acc_step": 0, "speed/wps": 12972.48326988577, "speed/FLOPS": 203750675813606.44, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.437038421630859, "optim/lr": 0.002999996466523411, "optim/total_tokens": 1320943616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433993, "loss/out": 6.665622711181641, "created_at": "2025-01-14T20:06:42.656802+00:00"} {"global_step": 5040, "acc_step": 0, "speed/wps": 12973.059471944687, "speed/FLOPS": 203759725858726.3, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.811871528625488, "optim/lr": 0.002999996278104031, "optim/total_tokens": 1321205760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 6.665070056915283, "created_at": "2025-01-14T20:06:47.713848+00:00"} {"global_step": 5041, "acc_step": 0, "speed/wps": 12965.187481248127, "speed/FLOPS": 203636085427588.06, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 30.93785858154297, "optim/lr": 0.0029999960847906498, "optim/total_tokens": 1321467904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443543, "loss/out": 6.584600448608398, "created_at": "2025-01-14T20:06:52.776982+00:00"} {"global_step": 5042, "acc_step": 0, "speed/wps": 12969.879970014872, "speed/FLOPS": 203709787411825.1, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.54729175567627, "optim/lr": 0.002999995886583267, "optim/total_tokens": 1321730048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436666, "loss/out": 6.74982213973999, "created_at": "2025-01-14T20:06:57.863714+00:00"} {"global_step": 5043, "acc_step": 0, "speed/wps": 12982.176760033424, "speed/FLOPS": 203902925396626.66, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.184412956237793, "optim/lr": 0.0029999956834818836, "optim/total_tokens": 1321992192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448881, "loss/out": 6.5204339027404785, "created_at": "2025-01-14T20:07:02.915528+00:00"} {"global_step": 5044, "acc_step": 0, "speed/wps": 12969.888018895736, "speed/FLOPS": 203709913830559.38, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.495530128479004, "optim/lr": 0.0029999954754865, "optim/total_tokens": 1322254336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432127, "loss/out": 6.535529136657715, "created_at": "2025-01-14T20:07:07.975838+00:00"} {"global_step": 5045, "acc_step": 0, "speed/wps": 12978.25123546696, "speed/FLOPS": 203841269639074.3, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.2257628440856934, "optim/lr": 0.0029999952625971186, "optim/total_tokens": 1322516480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303689, "loss/out": 6.512328147888184, "created_at": "2025-01-14T20:07:13.030806+00:00"} {"global_step": 5046, "acc_step": 0, "speed/wps": 12970.908500451578, "speed/FLOPS": 203725941895682.47, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 24.513591766357422, "optim/lr": 0.002999995044813738, "optim/total_tokens": 1322778624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455769, "loss/out": 6.506735324859619, "created_at": "2025-01-14T20:07:18.091908+00:00"} {"global_step": 5047, "acc_step": 0, "speed/wps": 12961.465101761782, "speed/FLOPS": 203577620342669.56, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.358916282653809, "optim/lr": 0.00299999482213636, "optim/total_tokens": 1323040768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 6.568526268005371, "created_at": "2025-01-14T20:07:23.156374+00:00"} {"global_step": 5048, "acc_step": 0, "speed/wps": 12974.156135701369, "speed/FLOPS": 203776950469998.94, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 1744.6363525390625, "optim/lr": 0.0029999945945649855, "optim/total_tokens": 1323302912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438531, "loss/out": 6.469847679138184, "created_at": "2025-01-14T20:07:28.216431+00:00"} {"global_step": 5049, "acc_step": 0, "speed/wps": 12970.798598989033, "speed/FLOPS": 203724215742192.78, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.167194366455078, "optim/lr": 0.0029999943620996144, "optim/total_tokens": 1323565056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495573, "loss/out": 6.563886642456055, "created_at": "2025-01-14T20:07:33.276584+00:00"} {"global_step": 5050, "acc_step": 0, "speed/wps": 12976.582207184567, "speed/FLOPS": 203815055256414.12, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.246537208557129, "optim/lr": 0.002999994124740248, "optim/total_tokens": 1323827200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435582, "loss/out": 6.642950057983398, "created_at": "2025-01-14T20:07:38.335625+00:00"} {"global_step": 5051, "acc_step": 0, "speed/wps": 12973.359966142883, "speed/FLOPS": 203764445533031.3, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.104392051696777, "optim/lr": 0.002999993882486887, "optim/total_tokens": 1324089344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310132, "loss/out": 6.4857587814331055, "created_at": "2025-01-14T20:07:43.392464+00:00"} {"global_step": 5052, "acc_step": 0, "speed/wps": 12974.439667360044, "speed/FLOPS": 203781403724311.75, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.0546135902404785, "optim/lr": 0.0029999936353395317, "optim/total_tokens": 1324351488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372023, "loss/out": 6.532922267913818, "created_at": "2025-01-14T20:07:48.449294+00:00"} {"global_step": 5053, "acc_step": 0, "speed/wps": 12974.544434589376, "speed/FLOPS": 203783049237613.66, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.001, "optim/grad_norm": 3.1650426387786865, "optim/lr": 0.0029999933832981834, "optim/total_tokens": 1324613632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448605, "loss/out": 6.881747245788574, "created_at": "2025-01-14T20:07:53.505202+00:00"} {"global_step": 5054, "acc_step": 0, "speed/wps": 12974.910419093214, "speed/FLOPS": 203788797527161.84, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.01150131225586, "optim/lr": 0.002999993126362843, "optim/total_tokens": 1324875776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 6.542688369750977, "created_at": "2025-01-14T20:07:58.560827+00:00"} {"global_step": 5055, "acc_step": 0, "speed/wps": 12972.50791650759, "speed/FLOPS": 203751062922670.12, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.167085647583008, "optim/lr": 0.0029999928645335116, "optim/total_tokens": 1325137920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452325, "loss/out": 6.641349792480469, "created_at": "2025-01-14T20:08:03.620474+00:00"} {"global_step": 5056, "acc_step": 0, "speed/wps": 12965.679277517746, "speed/FLOPS": 203643809763801.3, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.4085497856140137, "optim/lr": 0.0029999925978101897, "optim/total_tokens": 1325400064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 6.647457122802734, "created_at": "2025-01-14T20:08:08.678569+00:00"} {"global_step": 5057, "acc_step": 0, "speed/wps": 12981.616545248484, "speed/FLOPS": 203894126453613.4, "speed/curr_iter_time": 1.2722, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7254559993743896, "optim/lr": 0.0029999923261928778, "optim/total_tokens": 1325662208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460578, "loss/out": 6.404329299926758, "created_at": "2025-01-14T20:08:13.731412+00:00"} {"global_step": 5058, "acc_step": 0, "speed/wps": 12977.167314781273, "speed/FLOPS": 203824245175242.5, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 19.967552185058594, "optim/lr": 0.002999992049681577, "optim/total_tokens": 1325924352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416700, "loss/out": 6.628148078918457, "created_at": "2025-01-14T20:08:18.789159+00:00"} {"global_step": 5059, "acc_step": 0, "speed/wps": 12976.048377055717, "speed/FLOPS": 203806670720681.06, "speed/curr_iter_time": 1.2719, "speed/data_load_time": 0.0008, "optim/grad_norm": 14.343753814697266, "optim/lr": 0.0029999917682762885, "optim/total_tokens": 1326186496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 6.797016143798828, "created_at": "2025-01-14T20:08:23.847886+00:00"} {"global_step": 5060, "acc_step": 0, "speed/wps": 12968.47767853403, "speed/FLOPS": 203687762497168.03, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.293938398361206, "optim/lr": 0.002999991481977013, "optim/total_tokens": 1326448640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443306, "loss/out": 6.843384742736816, "created_at": "2025-01-14T20:08:28.905107+00:00"} {"global_step": 5061, "acc_step": 0, "speed/wps": 12980.991414941629, "speed/FLOPS": 203884307923125.2, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9317567348480225, "optim/lr": 0.0029999911907837517, "optim/total_tokens": 1326710784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433179, "loss/out": 6.764765739440918, "created_at": "2025-01-14T20:08:33.965252+00:00"} {"global_step": 5062, "acc_step": 0, "speed/wps": 12976.540117023815, "speed/FLOPS": 203814394172601.56, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.0909790992736816, "optim/lr": 0.002999990894696505, "optim/total_tokens": 1326972928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297240, "loss/out": 6.656180381774902, "created_at": "2025-01-14T20:08:39.025267+00:00"} {"global_step": 5063, "acc_step": 0, "speed/wps": 12982.468223043445, "speed/FLOPS": 203907503223711.2, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.9977192878723145, "optim/lr": 0.002999990593715274, "optim/total_tokens": 1327235072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401696, "loss/out": 6.738913059234619, "created_at": "2025-01-14T20:08:44.077016+00:00"} {"global_step": 5064, "acc_step": 0, "speed/wps": 12972.646973421715, "speed/FLOPS": 203753247002590.7, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.44242262840271, "optim/lr": 0.0029999902878400605, "optim/total_tokens": 1327497216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443022, "loss/out": 6.608514785766602, "created_at": "2025-01-14T20:08:49.132514+00:00"} {"global_step": 5065, "acc_step": 0, "speed/wps": 12975.610946189809, "speed/FLOPS": 203799800267839.0, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.2368667125701904, "optim/lr": 0.0029999899770708645, "optim/total_tokens": 1327759360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 6.5317840576171875, "created_at": "2025-01-14T20:08:54.187189+00:00"} {"global_step": 5066, "acc_step": 0, "speed/wps": 12971.816983260594, "speed/FLOPS": 203740210866584.28, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.3920512199401855, "optim/lr": 0.0029999896614076875, "optim/total_tokens": 1328021504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 6.544197082519531, "created_at": "2025-01-14T20:08:59.243537+00:00"} {"global_step": 5067, "acc_step": 0, "speed/wps": 12981.010449774092, "speed/FLOPS": 203884606891325.62, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.3137319087982178, "optim/lr": 0.00299998934085053, "optim/total_tokens": 1328283648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 6.702977657318115, "created_at": "2025-01-14T20:09:04.300616+00:00"} {"global_step": 5068, "acc_step": 0, "speed/wps": 12971.659888354292, "speed/FLOPS": 203737743475209.78, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.386611938476562, "optim/lr": 0.0029999890153993938, "optim/total_tokens": 1328545792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440108, "loss/out": 6.592836380004883, "created_at": "2025-01-14T20:09:09.366638+00:00"} {"global_step": 5069, "acc_step": 0, "speed/wps": 12978.399887244925, "speed/FLOPS": 203843604419516.2, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.397317409515381, "optim/lr": 0.0029999886850542797, "optim/total_tokens": 1328807936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439820, "loss/out": 6.877325057983398, "created_at": "2025-01-14T20:09:14.421254+00:00"} {"global_step": 5070, "acc_step": 0, "speed/wps": 12976.141476509845, "speed/FLOPS": 203808132975540.22, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.5856869220733643, "optim/lr": 0.002999988349815189, "optim/total_tokens": 1329070080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 6.678030014038086, "created_at": "2025-01-14T20:09:19.480898+00:00"} {"global_step": 5071, "acc_step": 0, "speed/wps": 12977.654359969043, "speed/FLOPS": 203831894889186.66, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7157845497131348, "optim/lr": 0.002999988009682122, "optim/total_tokens": 1329332224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444878, "loss/out": 6.510671615600586, "created_at": "2025-01-14T20:09:24.535142+00:00"} {"global_step": 5072, "acc_step": 0, "speed/wps": 12978.41920171798, "speed/FLOPS": 203843907779856.4, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.260274410247803, "optim/lr": 0.0029999876646550805, "optim/total_tokens": 1329594368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 6.597621440887451, "created_at": "2025-01-14T20:09:29.593631+00:00"} {"global_step": 5073, "acc_step": 0, "speed/wps": 12974.76406013421, "speed/FLOPS": 203786498758592.4, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7239741086959839, "optim/lr": 0.002999987314734065, "optim/total_tokens": 1329856512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301284, "loss/out": 6.624154090881348, "created_at": "2025-01-14T20:09:34.651711+00:00"} {"global_step": 5074, "acc_step": 0, "speed/wps": 12973.94697427873, "speed/FLOPS": 203773665302438.47, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.383636474609375, "optim/lr": 0.002999986959919078, "optim/total_tokens": 1330118656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 384529, "loss/out": 6.667316436767578, "created_at": "2025-01-14T20:09:39.706645+00:00"} {"global_step": 5075, "acc_step": 0, "speed/wps": 12979.845536576902, "speed/FLOPS": 203866310328805.25, "speed/curr_iter_time": 1.2721, "speed/data_load_time": 0.0009, "optim/grad_norm": 2.866797924041748, "optim/lr": 0.002999986600210119, "optim/total_tokens": 1330380800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441734, "loss/out": 6.806634902954102, "created_at": "2025-01-14T20:09:44.764621+00:00"} {"global_step": 5076, "acc_step": 0, "speed/wps": 12971.071595196756, "speed/FLOPS": 203728503522771.16, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.579887866973877, "optim/lr": 0.00299998623560719, "optim/total_tokens": 1330642944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 6.563382148742676, "created_at": "2025-01-14T20:09:49.823217+00:00"} {"global_step": 5077, "acc_step": 0, "speed/wps": 12972.177775297812, "speed/FLOPS": 203745877601309.0, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9359136819839478, "optim/lr": 0.002999985866110292, "optim/total_tokens": 1330905088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457095, "loss/out": 6.489528179168701, "created_at": "2025-01-14T20:09:54.879833+00:00"} {"global_step": 5078, "acc_step": 0, "speed/wps": 12981.924393352416, "speed/FLOPS": 203898961631113.28, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.7840495109558105, "optim/lr": 0.0029999854917194266, "optim/total_tokens": 1331167232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421271, "loss/out": 6.56754207611084, "created_at": "2025-01-14T20:09:59.932585+00:00"} {"global_step": 5079, "acc_step": 0, "speed/wps": 12973.790782170201, "speed/FLOPS": 203771212090743.6, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.111580848693848, "optim/lr": 0.0029999851124345943, "optim/total_tokens": 1331429376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433464, "loss/out": 6.4433770179748535, "created_at": "2025-01-14T20:10:04.988885+00:00"} {"global_step": 5080, "acc_step": 0, "speed/wps": 12973.499323563186, "speed/FLOPS": 203766634332814.3, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.8593711853027344, "optim/lr": 0.0029999847282557975, "optim/total_tokens": 1331691520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480229, "loss/out": 6.589554309844971, "created_at": "2025-01-14T20:10:10.046685+00:00"} {"global_step": 5081, "acc_step": 0, "speed/wps": 12976.336135925587, "speed/FLOPS": 203811190369154.5, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.8736162185668945, "optim/lr": 0.002999984339183036, "optim/total_tokens": 1331953664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 6.440838813781738, "created_at": "2025-01-14T20:10:15.101171+00:00"} {"global_step": 5082, "acc_step": 0, "speed/wps": 12982.619781028689, "speed/FLOPS": 203909883650131.94, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1819469928741455, "optim/lr": 0.0029999839452163122, "optim/total_tokens": 1332215808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 6.491988182067871, "created_at": "2025-01-14T20:10:20.158570+00:00"} {"global_step": 5083, "acc_step": 0, "speed/wps": 12972.47606675724, "speed/FLOPS": 203750562678574.72, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6503509283065796, "optim/lr": 0.0029999835463556264, "optim/total_tokens": 1332477952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448361, "loss/out": 6.536684989929199, "created_at": "2025-01-14T20:10:25.216678+00:00"} {"global_step": 5084, "acc_step": 0, "speed/wps": 12971.91249696804, "speed/FLOPS": 203741711040609.3, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.44687557220459, "optim/lr": 0.0029999831426009813, "optim/total_tokens": 1332740096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321779, "loss/out": 6.526966094970703, "created_at": "2025-01-14T20:10:30.281897+00:00"} {"global_step": 5085, "acc_step": 0, "speed/wps": 12975.691693222789, "speed/FLOPS": 203801068510950.0, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.557908535003662, "optim/lr": 0.002999982733952377, "optim/total_tokens": 1333002240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 389966, "loss/out": 6.391716003417969, "created_at": "2025-01-14T20:10:35.338883+00:00"} {"global_step": 5086, "acc_step": 0, "speed/wps": 12966.454659414272, "speed/FLOPS": 203655988201972.66, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.061983585357666, "optim/lr": 0.0029999823204098154, "optim/total_tokens": 1333264384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456330, "loss/out": 6.382569313049316, "created_at": "2025-01-14T20:10:40.398258+00:00"} {"global_step": 5087, "acc_step": 0, "speed/wps": 12973.177285307507, "speed/FLOPS": 203761576279481.94, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.242518424987793, "optim/lr": 0.0029999819019732974, "optim/total_tokens": 1333526528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 6.365817546844482, "created_at": "2025-01-14T20:10:45.455755+00:00"} {"global_step": 5088, "acc_step": 0, "speed/wps": 12971.434674693699, "speed/FLOPS": 203734206185194.16, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.390986442565918, "optim/lr": 0.0029999814786428242, "optim/total_tokens": 1333788672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440919, "loss/out": 6.397585868835449, "created_at": "2025-01-14T20:10:50.514005+00:00"} {"global_step": 5089, "acc_step": 0, "speed/wps": 12971.483326418238, "speed/FLOPS": 203734970327382.53, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.002983093261719, "optim/lr": 0.0029999810504183983, "optim/total_tokens": 1334050816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 6.493590354919434, "created_at": "2025-01-14T20:10:55.570959+00:00"} {"global_step": 5090, "acc_step": 0, "speed/wps": 12971.845458121028, "speed/FLOPS": 203740658103395.94, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.0423583984375, "optim/lr": 0.0029999806173000203, "optim/total_tokens": 1334312960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462435, "loss/out": 6.527037143707275, "created_at": "2025-01-14T20:11:00.628106+00:00"} {"global_step": 5091, "acc_step": 0, "speed/wps": 12973.857664118852, "speed/FLOPS": 203772262563652.56, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.538813591003418, "optim/lr": 0.0029999801792876916, "optim/total_tokens": 1334575104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323655, "loss/out": 6.510626792907715, "created_at": "2025-01-14T20:11:05.684413+00:00"} {"global_step": 5092, "acc_step": 0, "speed/wps": 12975.22482522553, "speed/FLOPS": 203793735707509.2, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.2614235877990723, "optim/lr": 0.0029999797363814136, "optim/total_tokens": 1334837248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333837, "loss/out": 6.454941749572754, "created_at": "2025-01-14T20:11:10.740818+00:00"} {"global_step": 5093, "acc_step": 0, "speed/wps": 12972.224517098832, "speed/FLOPS": 203746611745523.97, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.702587127685547, "optim/lr": 0.002999979288581188, "optim/total_tokens": 1335099392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453666, "loss/out": 6.420228958129883, "created_at": "2025-01-14T20:11:15.797417+00:00"} {"global_step": 5094, "acc_step": 0, "speed/wps": 12970.251613063936, "speed/FLOPS": 203715624576601.7, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.372888088226318, "optim/lr": 0.0029999788358870162, "optim/total_tokens": 1335361536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441679, "loss/out": 6.333839416503906, "created_at": "2025-01-14T20:11:20.856095+00:00"} {"global_step": 5095, "acc_step": 0, "speed/wps": 12975.034165779172, "speed/FLOPS": 203790741138910.6, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 19.251049041748047, "optim/lr": 0.0029999783782989, "optim/total_tokens": 1335623680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442784, "loss/out": 6.527486801147461, "created_at": "2025-01-14T20:11:25.910635+00:00"} {"global_step": 5096, "acc_step": 0, "speed/wps": 12968.693527953827, "speed/FLOPS": 203691152708914.12, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.56093430519104, "optim/lr": 0.0029999779158168404, "optim/total_tokens": 1335885824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443543, "loss/out": 6.447971343994141, "created_at": "2025-01-14T20:11:30.967951+00:00"} {"global_step": 5097, "acc_step": 0, "speed/wps": 12970.512214095808, "speed/FLOPS": 203719717673910.6, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.644922256469727, "optim/lr": 0.0029999774484408383, "optim/total_tokens": 1336147968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 6.366974353790283, "created_at": "2025-01-14T20:11:36.029714+00:00"} {"global_step": 5098, "acc_step": 0, "speed/wps": 12975.042194741878, "speed/FLOPS": 203790867244802.84, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.777714252471924, "optim/lr": 0.0029999769761708964, "optim/total_tokens": 1336410112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458410, "loss/out": 6.384539604187012, "created_at": "2025-01-14T20:11:41.090067+00:00"} {"global_step": 5099, "acc_step": 0, "speed/wps": 12971.323284254058, "speed/FLOPS": 203732456645273.1, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.676636219024658, "optim/lr": 0.002999976499007016, "optim/total_tokens": 1336672256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435852, "loss/out": 6.374812126159668, "created_at": "2025-01-14T20:11:46.149999+00:00"} {"global_step": 5100, "acc_step": 0, "speed/wps": 12972.434754147667, "speed/FLOPS": 203749913807271.9, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.799098968505859, "optim/lr": 0.0029999760169491982, "optim/total_tokens": 1336934400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 6.470037937164307, "created_at": "2025-01-14T20:11:51.210086+00:00"} {"global_step": 5101, "acc_step": 0, "speed/wps": 12976.995939592585, "speed/FLOPS": 203821553492409.97, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.435086727142334, "optim/lr": 0.0029999755299974454, "optim/total_tokens": 1337196544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431320, "loss/out": 6.339310646057129, "created_at": "2025-01-14T20:11:56.269237+00:00"} {"global_step": 5102, "acc_step": 0, "speed/wps": 12972.924706356263, "speed/FLOPS": 203757609179972.8, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 16.047422409057617, "optim/lr": 0.0029999750381517584, "optim/total_tokens": 1337458688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305392, "loss/out": 6.427050590515137, "created_at": "2025-01-14T20:12:01.330670+00:00"} {"global_step": 5103, "acc_step": 0, "speed/wps": 12971.009432982612, "speed/FLOPS": 203727527179778.2, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0008, "optim/grad_norm": 14.766141891479492, "optim/lr": 0.0029999745414121385, "optim/total_tokens": 1337720832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441971, "loss/out": 6.428675651550293, "created_at": "2025-01-14T20:12:06.386674+00:00"} {"global_step": 5104, "acc_step": 0, "speed/wps": 12968.10152230834, "speed/FLOPS": 203681854446750.75, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.801666021347046, "optim/lr": 0.0029999740397785883, "optim/total_tokens": 1337982976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466683, "loss/out": 6.390936851501465, "created_at": "2025-01-14T20:12:11.451141+00:00"} {"global_step": 5105, "acc_step": 0, "speed/wps": 12970.160309790934, "speed/FLOPS": 203714190533235.34, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.001, "optim/grad_norm": 5.2399444580078125, "optim/lr": 0.002999973533251109, "optim/total_tokens": 1338245120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453666, "loss/out": 6.469296455383301, "created_at": "2025-01-14T20:12:16.507967+00:00"} {"global_step": 5106, "acc_step": 0, "speed/wps": 12966.151553533491, "speed/FLOPS": 203651227507603.84, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.827234268188477, "optim/lr": 0.002999973021829702, "optim/total_tokens": 1338507264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 6.4200544357299805, "created_at": "2025-01-14T20:12:21.571304+00:00"} {"global_step": 5107, "acc_step": 0, "speed/wps": 12971.777059363803, "speed/FLOPS": 203739583806924.0, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.4612796306610107, "optim/lr": 0.002999972505514369, "optim/total_tokens": 1338769408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458173, "loss/out": 6.568851947784424, "created_at": "2025-01-14T20:12:26.628825+00:00"} {"global_step": 5108, "acc_step": 0, "speed/wps": 12975.80646541792, "speed/FLOPS": 203802871165988.03, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.991052627563477, "optim/lr": 0.002999971984305112, "optim/total_tokens": 1339031552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448605, "loss/out": 6.305427551269531, "created_at": "2025-01-14T20:12:31.687967+00:00"} {"global_step": 5109, "acc_step": 0, "speed/wps": 12967.450845453799, "speed/FLOPS": 203671634672626.5, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.234461784362793, "optim/lr": 0.002999971458201933, "optim/total_tokens": 1339293696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310371, "loss/out": 6.456517696380615, "created_at": "2025-01-14T20:12:36.746944+00:00"} {"global_step": 5110, "acc_step": 0, "speed/wps": 12973.166584697725, "speed/FLOPS": 203761408211701.12, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0008, "optim/grad_norm": 10.076410293579102, "optim/lr": 0.0029999709272048336, "optim/total_tokens": 1339555840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294941, "loss/out": 6.511134624481201, "created_at": "2025-01-14T20:12:41.802669+00:00"} {"global_step": 5111, "acc_step": 0, "speed/wps": 12973.91771996501, "speed/FLOPS": 203773205823241.84, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.122664451599121, "optim/lr": 0.002999970391313814, "optim/total_tokens": 1339817984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376973, "loss/out": 6.389015197753906, "created_at": "2025-01-14T20:12:46.857587+00:00"} {"global_step": 5112, "acc_step": 0, "speed/wps": 12969.415269864312, "speed/FLOPS": 203702488657395.62, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.2088541984558105, "optim/lr": 0.0029999698505288786, "optim/total_tokens": 1340080128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437432, "loss/out": 6.412102699279785, "created_at": "2025-01-14T20:12:51.918513+00:00"} {"global_step": 5113, "acc_step": 0, "speed/wps": 12983.244014187849, "speed/FLOPS": 203919688089682.78, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.088820457458496, "optim/lr": 0.0029999693048500266, "optim/total_tokens": 1340342272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450746, "loss/out": 6.404681205749512, "created_at": "2025-01-14T20:12:56.975619+00:00"} {"global_step": 5114, "acc_step": 0, "speed/wps": 12969.36880168073, "speed/FLOPS": 203701758810718.47, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.853214740753174, "optim/lr": 0.0029999687542772617, "optim/total_tokens": 1340604416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 6.389392852783203, "created_at": "2025-01-14T20:13:02.037016+00:00"} {"global_step": 5115, "acc_step": 0, "speed/wps": 12977.030951360002, "speed/FLOPS": 203822103400328.47, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.18186092376709, "optim/lr": 0.0029999681988105847, "optim/total_tokens": 1340866560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443545, "loss/out": 6.455482482910156, "created_at": "2025-01-14T20:13:07.091028+00:00"} {"global_step": 5116, "acc_step": 0, "speed/wps": 12974.804057443318, "speed/FLOPS": 203787126971294.03, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0012, "optim/grad_norm": 124.23702239990234, "optim/lr": 0.0029999676384499974, "optim/total_tokens": 1341128704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 6.367162704467773, "created_at": "2025-01-14T20:13:12.150852+00:00"} {"global_step": 5117, "acc_step": 0, "speed/wps": 12974.81692342841, "speed/FLOPS": 203787329049269.62, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7110111713409424, "optim/lr": 0.0029999670731955023, "optim/total_tokens": 1341390848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447555, "loss/out": 6.474862575531006, "created_at": "2025-01-14T20:13:17.211011+00:00"} {"global_step": 5118, "acc_step": 0, "speed/wps": 12967.735271810623, "speed/FLOPS": 203676101979403.2, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 22.57172966003418, "optim/lr": 0.0029999665030471002, "optim/total_tokens": 1341652992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289838, "loss/out": 6.308084964752197, "created_at": "2025-01-14T20:13:22.273812+00:00"} {"global_step": 5119, "acc_step": 0, "speed/wps": 12977.57982559441, "speed/FLOPS": 203830724224412.06, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.001, "optim/grad_norm": 8.447973251342773, "optim/lr": 0.0029999659280047944, "optim/total_tokens": 1341915136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415534, "loss/out": 6.584550857543945, "created_at": "2025-01-14T20:13:27.331902+00:00"} {"global_step": 5120, "acc_step": 0, "speed/wps": 12972.596344746675, "speed/FLOPS": 203752451809678.1, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.1085150241851807, "optim/lr": 0.0029999653480685855, "optim/total_tokens": 1342177280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451559, "loss/out": 6.4965643882751465, "created_at": "2025-01-14T20:13:32.390174+00:00"} {"global_step": 5121, "acc_step": 0, "speed/wps": 12978.901629791755, "speed/FLOPS": 203851484975680.7, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.932265758514404, "optim/lr": 0.002999964763238476, "optim/total_tokens": 1342439424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 6.37536096572876, "created_at": "2025-01-14T20:13:37.447751+00:00"} {"global_step": 5122, "acc_step": 0, "speed/wps": 12967.605090295145, "speed/FLOPS": 203674057299814.44, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.925631046295166, "optim/lr": 0.002999964173514467, "optim/total_tokens": 1342701568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480954, "loss/out": 6.484508514404297, "created_at": "2025-01-14T20:13:42.505416+00:00"} {"global_step": 5123, "acc_step": 0, "speed/wps": 12973.161337243768, "speed/FLOPS": 203761325793225.97, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.516399383544922, "optim/lr": 0.002999963578896562, "optim/total_tokens": 1342963712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469069, "loss/out": 6.361560821533203, "created_at": "2025-01-14T20:13:47.563514+00:00"} {"global_step": 5124, "acc_step": 0, "speed/wps": 12972.139446602747, "speed/FLOPS": 203745275596483.84, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.117711067199707, "optim/lr": 0.002999962979384761, "optim/total_tokens": 1343225856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 6.420055389404297, "created_at": "2025-01-14T20:13:52.619600+00:00"} {"global_step": 5125, "acc_step": 0, "speed/wps": 12979.195801744972, "speed/FLOPS": 203856105350440.6, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.825986862182617, "optim/lr": 0.0029999623749790674, "optim/total_tokens": 1343488000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 6.374241828918457, "created_at": "2025-01-14T20:13:57.676603+00:00"} {"global_step": 5126, "acc_step": 0, "speed/wps": 12967.246356675863, "speed/FLOPS": 203668422895371.97, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.507874488830566, "optim/lr": 0.0029999617656794836, "optim/total_tokens": 1343750144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435866, "loss/out": 6.470386505126953, "created_at": "2025-01-14T20:14:02.738925+00:00"} {"global_step": 5127, "acc_step": 0, "speed/wps": 12973.573227455026, "speed/FLOPS": 203767795094990.25, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.036713123321533, "optim/lr": 0.0029999611514860097, "optim/total_tokens": 1344012288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295714, "loss/out": 6.3531975746154785, "created_at": "2025-01-14T20:14:07.794383+00:00"} {"global_step": 5128, "acc_step": 0, "speed/wps": 12978.917670968936, "speed/FLOPS": 203851736924411.2, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.457064151763916, "optim/lr": 0.002999960532398649, "optim/total_tokens": 1344274432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361473, "loss/out": 6.287815093994141, "created_at": "2025-01-14T20:14:12.848079+00:00"} {"global_step": 5129, "acc_step": 0, "speed/wps": 12975.448821107162, "speed/FLOPS": 203797253870632.8, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.632452011108398, "optim/lr": 0.0029999599084174033, "optim/total_tokens": 1344536576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450462, "loss/out": 6.394055366516113, "created_at": "2025-01-14T20:14:17.902610+00:00"} {"global_step": 5130, "acc_step": 0, "speed/wps": 12980.899582058211, "speed/FLOPS": 203882865561499.5, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.33278751373291, "optim/lr": 0.0029999592795422747, "optim/total_tokens": 1344798720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430567, "loss/out": 6.376812934875488, "created_at": "2025-01-14T20:14:22.958303+00:00"} {"global_step": 5131, "acc_step": 0, "speed/wps": 12972.360222321788, "speed/FLOPS": 203748743182529.03, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.6920812129974365, "optim/lr": 0.002999958645773265, "optim/total_tokens": 1345060864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 6.38225793838501, "created_at": "2025-01-14T20:14:28.021021+00:00"} {"global_step": 5132, "acc_step": 0, "speed/wps": 12975.498395384022, "speed/FLOPS": 203798032502773.1, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.897822141647339, "optim/lr": 0.0029999580071103766, "optim/total_tokens": 1345323008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419618, "loss/out": 6.370034694671631, "created_at": "2025-01-14T20:14:33.083575+00:00"} {"global_step": 5133, "acc_step": 0, "speed/wps": 12973.79199250226, "speed/FLOPS": 203771231100671.72, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.1423444747924805, "optim/lr": 0.0029999573635536114, "optim/total_tokens": 1345585152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 6.388515949249268, "created_at": "2025-01-14T20:14:38.142622+00:00"} {"global_step": 5134, "acc_step": 0, "speed/wps": 12970.909375520188, "speed/FLOPS": 203725955639837.47, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.217919111251831, "optim/lr": 0.002999956715102971, "optim/total_tokens": 1345847296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 6.380934715270996, "created_at": "2025-01-14T20:14:43.199361+00:00"} {"global_step": 5135, "acc_step": 0, "speed/wps": 12977.651686558296, "speed/FLOPS": 203831852899597.22, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.001, "optim/grad_norm": 5.99604606628418, "optim/lr": 0.0029999560617584584, "optim/total_tokens": 1346109440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 6.333444595336914, "created_at": "2025-01-14T20:14:48.252850+00:00"} {"global_step": 5136, "acc_step": 0, "speed/wps": 12969.097368646757, "speed/FLOPS": 203697495581931.84, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0009, "optim/grad_norm": 2.197378635406494, "optim/lr": 0.0029999554035200753, "optim/total_tokens": 1346371584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 6.478035926818848, "created_at": "2025-01-14T20:14:53.309745+00:00"} {"global_step": 5137, "acc_step": 0, "speed/wps": 12973.982976640993, "speed/FLOPS": 203774230769010.25, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 26.072378158569336, "optim/lr": 0.002999954740387824, "optim/total_tokens": 1346633728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423105, "loss/out": 6.32624626159668, "created_at": "2025-01-14T20:14:58.367766+00:00"} {"global_step": 5138, "acc_step": 0, "speed/wps": 12976.307488201674, "speed/FLOPS": 203810740417284.34, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.688553333282471, "optim/lr": 0.002999954072361706, "optim/total_tokens": 1346895872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293067, "loss/out": 6.619707107543945, "created_at": "2025-01-14T20:15:03.425791+00:00"} {"global_step": 5139, "acc_step": 0, "speed/wps": 12980.138362625168, "speed/FLOPS": 203870909564278.44, "speed/curr_iter_time": 1.272, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.7975893020629883, "optim/lr": 0.0029999533994417246, "optim/total_tokens": 1347158016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370682, "loss/out": 6.489577293395996, "created_at": "2025-01-14T20:15:08.481888+00:00"} {"global_step": 5140, "acc_step": 0, "speed/wps": 12975.329937696806, "speed/FLOPS": 203795386643310.25, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.5463764667510986, "optim/lr": 0.0029999527216278813, "optim/total_tokens": 1347420160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445976, "loss/out": 6.286050796508789, "created_at": "2025-01-14T20:15:13.542298+00:00"} {"global_step": 5141, "acc_step": 0, "speed/wps": 12973.220824582182, "speed/FLOPS": 203762260123619.94, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3861746788024902, "optim/lr": 0.002999952038920178, "optim/total_tokens": 1347682304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 6.424931049346924, "created_at": "2025-01-14T20:15:18.597821+00:00"} {"global_step": 5142, "acc_step": 0, "speed/wps": 12977.601116939746, "speed/FLOPS": 203831058634247.94, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.145303726196289, "optim/lr": 0.002999951351318618, "optim/total_tokens": 1347944448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433481, "loss/out": 6.427123546600342, "created_at": "2025-01-14T20:15:23.652153+00:00"} {"global_step": 5143, "acc_step": 0, "speed/wps": 12978.416293412809, "speed/FLOPS": 203843862100927.25, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4989259243011475, "optim/lr": 0.0029999506588232025, "optim/total_tokens": 1348206592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451523, "loss/out": 6.269449234008789, "created_at": "2025-01-14T20:15:28.712693+00:00"} {"global_step": 5144, "acc_step": 0, "speed/wps": 12976.69476004219, "speed/FLOPS": 203816823053706.94, "speed/curr_iter_time": 1.2717, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.784423351287842, "optim/lr": 0.0029999499614339347, "optim/total_tokens": 1348468736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462701, "loss/out": 6.280935287475586, "created_at": "2025-01-14T20:15:33.778202+00:00"} {"global_step": 5145, "acc_step": 0, "speed/wps": 12972.659835129178, "speed/FLOPS": 203753449013380.22, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.556542873382568, "optim/lr": 0.002999949259150816, "optim/total_tokens": 1348730880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308260, "loss/out": 6.6174821853637695, "created_at": "2025-01-14T20:15:38.833716+00:00"} {"global_step": 5146, "acc_step": 0, "speed/wps": 12976.777576204684, "speed/FLOPS": 203818123795340.25, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5770058631896973, "optim/lr": 0.002999948551973849, "optim/total_tokens": 1348993024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353219, "loss/out": 6.549543380737305, "created_at": "2025-01-14T20:15:43.892688+00:00"} {"global_step": 5147, "acc_step": 0, "speed/wps": 12972.939601513668, "speed/FLOPS": 203757843128887.94, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0009, "optim/grad_norm": 4.118215084075928, "optim/lr": 0.002999947839903036, "optim/total_tokens": 1349255168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453945, "loss/out": 6.408949851989746, "created_at": "2025-01-14T20:15:48.957231+00:00"} {"global_step": 5148, "acc_step": 0, "speed/wps": 12975.290504171038, "speed/FLOPS": 203794767285600.72, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.406584739685059, "optim/lr": 0.0029999471229383793, "optim/total_tokens": 1349517312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439820, "loss/out": 6.401763916015625, "created_at": "2025-01-14T20:15:54.012274+00:00"} {"global_step": 5149, "acc_step": 0, "speed/wps": 12975.968810001294, "speed/FLOPS": 203805421010754.78, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.001, "optim/grad_norm": 9.062265396118164, "optim/lr": 0.002999946401079882, "optim/total_tokens": 1349779456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468540, "loss/out": 6.477101802825928, "created_at": "2025-01-14T20:15:59.069547+00:00"} {"global_step": 5150, "acc_step": 0, "speed/wps": 12976.236108731495, "speed/FLOPS": 203809619304619.53, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.133676528930664, "optim/lr": 0.0029999456743275447, "optim/total_tokens": 1350041600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 6.460725784301758, "created_at": "2025-01-14T20:16:04.123822+00:00"} {"global_step": 5151, "acc_step": 0, "speed/wps": 12970.937798577948, "speed/FLOPS": 203726402063016.97, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.5269579887390137, "optim/lr": 0.0029999449426813714, "optim/total_tokens": 1350303744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434515, "loss/out": 6.479928970336914, "created_at": "2025-01-14T20:16:09.185412+00:00"} {"global_step": 5152, "acc_step": 0, "speed/wps": 12978.552463334663, "speed/FLOPS": 203846000836667.06, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.027859926223755, "optim/lr": 0.002999944206141364, "optim/total_tokens": 1350565888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 6.44658088684082, "created_at": "2025-01-14T20:16:14.243764+00:00"} {"global_step": 5153, "acc_step": 0, "speed/wps": 12973.226947465606, "speed/FLOPS": 203762356291917.84, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.636490345001221, "optim/lr": 0.0029999434647075243, "optim/total_tokens": 1350828032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453911, "loss/out": 6.5298285484313965, "created_at": "2025-01-14T20:16:19.308723+00:00"} {"global_step": 5154, "acc_step": 0, "speed/wps": 12975.590178180952, "speed/FLOPS": 203799474077721.66, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.2844977378845215, "optim/lr": 0.0029999427183798554, "optim/total_tokens": 1351090176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352926, "loss/out": 6.427158355712891, "created_at": "2025-01-14T20:16:24.364327+00:00"} {"global_step": 5155, "acc_step": 0, "speed/wps": 12965.98028042809, "speed/FLOPS": 203648537428129.62, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0006, "optim/grad_norm": 29.228416442871094, "optim/lr": 0.0029999419671583607, "optim/total_tokens": 1351352320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435866, "loss/out": 6.390728950500488, "created_at": "2025-01-14T20:16:29.427334+00:00"} {"global_step": 5156, "acc_step": 0, "speed/wps": 12977.931747278622, "speed/FLOPS": 203836251638053.7, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.40352725982666, "optim/lr": 0.00299994121104304, "optim/total_tokens": 1351614464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 6.4473185539245605, "created_at": "2025-01-14T20:16:34.481938+00:00"} {"global_step": 5157, "acc_step": 0, "speed/wps": 12969.310408569694, "speed/FLOPS": 203700841666669.12, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0014, "optim/grad_norm": 7.492758750915527, "optim/lr": 0.002999940450033898, "optim/total_tokens": 1351876608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453660, "loss/out": 6.38663387298584, "created_at": "2025-01-14T20:16:39.540905+00:00"} {"global_step": 5158, "acc_step": 0, "speed/wps": 12978.863589307588, "speed/FLOPS": 203850887497603.75, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.703360557556152, "optim/lr": 0.002999939684130937, "optim/total_tokens": 1352138752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 6.30313777923584, "created_at": "2025-01-14T20:16:44.595696+00:00"} {"global_step": 5159, "acc_step": 0, "speed/wps": 12970.041549565341, "speed/FLOPS": 203712325240698.78, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.08864688873291, "optim/lr": 0.002999938913334158, "optim/total_tokens": 1352400896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 6.325035572052002, "created_at": "2025-01-14T20:16:49.653429+00:00"} {"global_step": 5160, "acc_step": 0, "speed/wps": 12966.761891569806, "speed/FLOPS": 203660813705156.53, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.545750617980957, "optim/lr": 0.002999938137643565, "optim/total_tokens": 1352663040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448077, "loss/out": 6.403210163116455, "created_at": "2025-01-14T20:16:54.711300+00:00"} {"global_step": 5161, "acc_step": 0, "speed/wps": 12966.003100971991, "speed/FLOPS": 203648895856130.3, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.208443641662598, "optim/lr": 0.00299993735705916, "optim/total_tokens": 1352925184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297061, "loss/out": 6.319229602813721, "created_at": "2025-01-14T20:16:59.775341+00:00"} {"global_step": 5162, "acc_step": 0, "speed/wps": 12971.365677719687, "speed/FLOPS": 203733122492905.53, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6868901252746582, "optim/lr": 0.002999936571580945, "optim/total_tokens": 1353187328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 347724, "loss/out": 6.4705634117126465, "created_at": "2025-01-14T20:17:04.831748+00:00"} {"global_step": 5163, "acc_step": 0, "speed/wps": 12972.051488508345, "speed/FLOPS": 203743894093744.2, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.3921117782592773, "optim/lr": 0.002999935781208924, "optim/total_tokens": 1353449472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295129, "loss/out": 6.277074813842773, "created_at": "2025-01-14T20:17:09.887680+00:00"} {"global_step": 5164, "acc_step": 0, "speed/wps": 12973.23028874466, "speed/FLOPS": 203762408771296.78, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 58.15127182006836, "optim/lr": 0.002999934985943098, "optim/total_tokens": 1353711616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364548, "loss/out": 6.410651206970215, "created_at": "2025-01-14T20:17:14.942860+00:00"} {"global_step": 5165, "acc_step": 0, "speed/wps": 12976.205604596686, "speed/FLOPS": 203809140195265.28, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.899873733520508, "optim/lr": 0.0029999341857834703, "optim/total_tokens": 1353973760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454182, "loss/out": 6.475735664367676, "created_at": "2025-01-14T20:17:19.998063+00:00"} {"global_step": 5166, "acc_step": 0, "speed/wps": 12968.798381663382, "speed/FLOPS": 203692799580506.78, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 43.6777458190918, "optim/lr": 0.002999933380730044, "optim/total_tokens": 1354235904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 6.37668514251709, "created_at": "2025-01-14T20:17:25.061415+00:00"} {"global_step": 5167, "acc_step": 0, "speed/wps": 12974.09177431016, "speed/FLOPS": 203775939585908.2, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.6630687713623047, "optim/lr": 0.002999932570782821, "optim/total_tokens": 1354498048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436666, "loss/out": 6.407687187194824, "created_at": "2025-01-14T20:17:30.117540+00:00"} {"global_step": 5168, "acc_step": 0, "speed/wps": 12975.501945988659, "speed/FLOPS": 203798088269897.97, "speed/curr_iter_time": 1.2722, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6022824048995972, "optim/lr": 0.0029999317559418043, "optim/total_tokens": 1354760192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 6.431281089782715, "created_at": "2025-01-14T20:17:35.175738+00:00"} {"global_step": 5169, "acc_step": 0, "speed/wps": 12966.625199584998, "speed/FLOPS": 203658666769700.66, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0011, "optim/grad_norm": 2.084867477416992, "optim/lr": 0.0029999309362069965, "optim/total_tokens": 1355022336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440921, "loss/out": 6.320664882659912, "created_at": "2025-01-14T20:17:40.238195+00:00"} {"global_step": 5170, "acc_step": 0, "speed/wps": 12972.790354921453, "speed/FLOPS": 203755499006070.9, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.072735786437988, "optim/lr": 0.0029999301115783998, "optim/total_tokens": 1355284480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292724, "loss/out": 6.329774379730225, "created_at": "2025-01-14T20:17:45.294445+00:00"} {"global_step": 5171, "acc_step": 0, "speed/wps": 12977.538966319815, "speed/FLOPS": 203830082473358.03, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.058461666107178, "optim/lr": 0.002999929282056018, "optim/total_tokens": 1355546624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363341, "loss/out": 6.673709392547607, "created_at": "2025-01-14T20:17:50.351355+00:00"} {"global_step": 5172, "acc_step": 0, "speed/wps": 12970.830517010743, "speed/FLOPS": 203724717058583.7, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.910337924957275, "optim/lr": 0.0029999284476398525, "optim/total_tokens": 1355808768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432668, "loss/out": 6.415597915649414, "created_at": "2025-01-14T20:17:55.408874+00:00"} {"global_step": 5173, "acc_step": 0, "speed/wps": 12969.167770864675, "speed/FLOPS": 203698601345507.03, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.431690216064453, "optim/lr": 0.002999927608329907, "optim/total_tokens": 1356070912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 6.400328636169434, "created_at": "2025-01-14T20:18:00.467041+00:00"} {"global_step": 5174, "acc_step": 0, "speed/wps": 12979.447957011205, "speed/FLOPS": 203860065795394.4, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 16.62322425842285, "optim/lr": 0.0029999267641261838, "optim/total_tokens": 1356333056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482522, "loss/out": 6.485597610473633, "created_at": "2025-01-14T20:18:05.525361+00:00"} {"global_step": 5175, "acc_step": 0, "speed/wps": 12969.265406386345, "speed/FLOPS": 203700134845540.12, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 16.415376663208008, "optim/lr": 0.0029999259150286853, "optim/total_tokens": 1356595200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 6.546335220336914, "created_at": "2025-01-14T20:18:10.583562+00:00"} {"global_step": 5176, "acc_step": 0, "speed/wps": 12973.750075916367, "speed/FLOPS": 203770572743090.88, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0009, "optim/grad_norm": 4.4177727699279785, "optim/lr": 0.002999925061037415, "optim/total_tokens": 1356857344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 6.508551120758057, "created_at": "2025-01-14T20:18:15.643271+00:00"} {"global_step": 5177, "acc_step": 0, "speed/wps": 12972.489624046631, "speed/FLOPS": 203750775614434.28, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.904098033905029, "optim/lr": 0.0029999242021523757, "optim/total_tokens": 1357119488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457380, "loss/out": 6.381501197814941, "created_at": "2025-01-14T20:18:20.700558+00:00"} {"global_step": 5178, "acc_step": 0, "speed/wps": 12974.019543725597, "speed/FLOPS": 203774805105322.4, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.358532428741455, "optim/lr": 0.0029999233383735694, "optim/total_tokens": 1357381632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 6.401887893676758, "created_at": "2025-01-14T20:18:25.756681+00:00"} {"global_step": 5179, "acc_step": 0, "speed/wps": 12970.775426842774, "speed/FLOPS": 203723851791793.9, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.214650630950928, "optim/lr": 0.0029999224697009994, "optim/total_tokens": 1357643776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446505, "loss/out": 6.232398986816406, "created_at": "2025-01-14T20:18:30.813135+00:00"} {"global_step": 5180, "acc_step": 0, "speed/wps": 12968.882590912315, "speed/FLOPS": 203694122202476.9, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.959381580352783, "optim/lr": 0.0029999215961346694, "optim/total_tokens": 1357905920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322837, "loss/out": 6.351127624511719, "created_at": "2025-01-14T20:18:35.890271+00:00"} {"global_step": 5181, "acc_step": 0, "speed/wps": 12975.14632573787, "speed/FLOPS": 203792502765185.88, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.072721481323242, "optim/lr": 0.0029999207176745805, "optim/total_tokens": 1358168064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454182, "loss/out": 6.309750556945801, "created_at": "2025-01-14T20:18:40.945143+00:00"} {"global_step": 5182, "acc_step": 0, "speed/wps": 12972.83712505321, "speed/FLOPS": 203756233595259.03, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.114786624908447, "optim/lr": 0.002999919834320737, "optim/total_tokens": 1358430208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 6.279963493347168, "created_at": "2025-01-14T20:18:46.004712+00:00"} {"global_step": 5183, "acc_step": 0, "speed/wps": 12973.880486088174, "speed/FLOPS": 203772621014041.44, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.750532627105713, "optim/lr": 0.0029999189460731407, "optim/total_tokens": 1358692352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452330, "loss/out": 6.2999162673950195, "created_at": "2025-01-14T20:18:51.061332+00:00"} {"global_step": 5184, "acc_step": 0, "speed/wps": 12968.733999324915, "speed/FLOPS": 203691788367410.56, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.250091075897217, "optim/lr": 0.0029999180529317955, "optim/total_tokens": 1358954496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462435, "loss/out": 6.330737113952637, "created_at": "2025-01-14T20:18:56.121826+00:00"} {"global_step": 5185, "acc_step": 0, "speed/wps": 12967.350850053783, "speed/FLOPS": 203670064107461.16, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.2783432006835938, "optim/lr": 0.002999917154896703, "optim/total_tokens": 1359216640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462701, "loss/out": 6.240339279174805, "created_at": "2025-01-14T20:19:01.179339+00:00"} {"global_step": 5186, "acc_step": 0, "speed/wps": 12969.073456829476, "speed/FLOPS": 203697120013983.2, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.0688796043396, "optim/lr": 0.002999916251967868, "optim/total_tokens": 1359478784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331823, "loss/out": 6.518444538116455, "created_at": "2025-01-14T20:19:06.240236+00:00"} {"global_step": 5187, "acc_step": 0, "speed/wps": 12959.761079707148, "speed/FLOPS": 203550856334729.88, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0006, "optim/grad_norm": 21.229806900024414, "optim/lr": 0.0029999153441452923, "optim/total_tokens": 1359740928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309147, "loss/out": 6.349132537841797, "created_at": "2025-01-14T20:19:11.305130+00:00"} {"global_step": 5188, "acc_step": 0, "speed/wps": 12956.06826938702, "speed/FLOPS": 203492855674199.12, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.9477014541625977, "optim/lr": 0.0029999144314289787, "optim/total_tokens": 1360003072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 388341, "loss/out": 6.326808452606201, "created_at": "2025-01-14T20:19:16.373291+00:00"} {"global_step": 5189, "acc_step": 0, "speed/wps": 12958.97936247493, "speed/FLOPS": 203538578391405.84, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.387877464294434, "optim/lr": 0.0029999135138189304, "optim/total_tokens": 1360265216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459237, "loss/out": 6.344095706939697, "created_at": "2025-01-14T20:19:21.436764+00:00"} {"global_step": 5190, "acc_step": 0, "speed/wps": 12967.78402005529, "speed/FLOPS": 203676867637572.78, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 57.05904769897461, "optim/lr": 0.0029999125913151505, "optim/total_tokens": 1360527360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497367, "loss/out": 6.306146621704102, "created_at": "2025-01-14T20:19:26.500056+00:00"} {"global_step": 5191, "acc_step": 0, "speed/wps": 12968.9738403444, "speed/FLOPS": 203695555400197.7, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.769856929779053, "optim/lr": 0.002999911663917642, "optim/total_tokens": 1360789504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440390, "loss/out": 6.4399261474609375, "created_at": "2025-01-14T20:19:31.562819+00:00"} {"global_step": 5192, "acc_step": 0, "speed/wps": 12969.129997759192, "speed/FLOPS": 203698008066979.75, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.103351593017578, "optim/lr": 0.002999910731626408, "optim/total_tokens": 1361051648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 6.260374069213867, "created_at": "2025-01-14T20:19:36.620949+00:00"} {"global_step": 5193, "acc_step": 0, "speed/wps": 12969.283663462225, "speed/FLOPS": 203700421598004.7, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.084159851074219, "optim/lr": 0.0029999097944414512, "optim/total_tokens": 1361313792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 6.38831090927124, "created_at": "2025-01-14T20:19:41.681456+00:00"} {"global_step": 5194, "acc_step": 0, "speed/wps": 12965.660571907367, "speed/FLOPS": 203643515966486.12, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.84630012512207, "optim/lr": 0.0029999088523627758, "optim/total_tokens": 1361575936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290078, "loss/out": 6.321096420288086, "created_at": "2025-01-14T20:19:46.742287+00:00"} {"global_step": 5195, "acc_step": 0, "speed/wps": 12975.33999828462, "speed/FLOPS": 203795544658666.56, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.727267265319824, "optim/lr": 0.002999907905390383, "optim/total_tokens": 1361838080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 385632, "loss/out": 6.399343490600586, "created_at": "2025-01-14T20:19:51.798684+00:00"} {"global_step": 5196, "acc_step": 0, "speed/wps": 12966.707913553633, "speed/FLOPS": 203659965906238.97, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0005, "optim/grad_norm": 11.059599876403809, "optim/lr": 0.002999906953524277, "optim/total_tokens": 1362100224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455808, "loss/out": 6.428606033325195, "created_at": "2025-01-14T20:19:56.860856+00:00"} {"global_step": 5197, "acc_step": 0, "speed/wps": 12971.570913354262, "speed/FLOPS": 203736346000569.84, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.937612056732178, "optim/lr": 0.002999905996764461, "optim/total_tokens": 1362362368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 6.250591278076172, "created_at": "2025-01-14T20:20:01.920700+00:00"} {"global_step": 5198, "acc_step": 0, "speed/wps": 12963.692458032501, "speed/FLOPS": 203612604033609.47, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0013, "optim/grad_norm": 3.671316146850586, "optim/lr": 0.002999905035110938, "optim/total_tokens": 1362624512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457380, "loss/out": 6.223321914672852, "created_at": "2025-01-14T20:20:06.980290+00:00"} {"global_step": 5199, "acc_step": 0, "speed/wps": 12960.626659133297, "speed/FLOPS": 203564451449048.3, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 37.95261001586914, "optim/lr": 0.0029999040685637106, "optim/total_tokens": 1362886656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 6.422651290893555, "created_at": "2025-01-14T20:20:12.041506+00:00"} {"global_step": 5200, "acc_step": 0, "speed/wps": 12965.595293811228, "speed/FLOPS": 203642490684284.75, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.878501892089844, "optim/lr": 0.002999903097122782, "optim/total_tokens": 1363148800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 265442, "loss/out": 6.319660663604736, "created_at": "2025-01-14T20:20:17.111407+00:00"} {"global_step": 5201, "acc_step": 0, "speed/wps": 12976.072778803604, "speed/FLOPS": 203807053983663.16, "speed/curr_iter_time": 1.2721, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.862621307373047, "optim/lr": 0.002999902120788157, "optim/total_tokens": 1363410944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310371, "loss/out": 6.2953667640686035, "created_at": "2025-01-14T20:20:22.165801+00:00"} {"global_step": 5202, "acc_step": 0, "speed/wps": 12968.784289532128, "speed/FLOPS": 203692578244220.84, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.519001483917236, "optim/lr": 0.002999901139559836, "optim/total_tokens": 1363673088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291959, "loss/out": 6.338930130004883, "created_at": "2025-01-14T20:20:27.227528+00:00"} {"global_step": 5203, "acc_step": 0, "speed/wps": 12966.134874938381, "speed/FLOPS": 203650965547348.88, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.1231422424316406, "optim/lr": 0.002999900153437825, "optim/total_tokens": 1363935232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456584, "loss/out": 6.213672637939453, "created_at": "2025-01-14T20:20:32.288469+00:00"} {"global_step": 5204, "acc_step": 0, "speed/wps": 12971.702800517292, "speed/FLOPS": 203738417469697.2, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6614359617233276, "optim/lr": 0.0029998991624221255, "optim/total_tokens": 1364197376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454995, "loss/out": 6.309615612030029, "created_at": "2025-01-14T20:20:37.344502+00:00"} {"global_step": 5205, "acc_step": 0, "speed/wps": 12965.428046325393, "speed/FLOPS": 203639863832697.6, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.355166912078857, "optim/lr": 0.002999898166512741, "optim/total_tokens": 1364459520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 6.410845756530762, "created_at": "2025-01-14T20:20:42.407700+00:00"} {"global_step": 5206, "acc_step": 0, "speed/wps": 12969.64765641442, "speed/FLOPS": 203706138607497.62, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.385784864425659, "optim/lr": 0.002999897165709675, "optim/total_tokens": 1364721664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431842, "loss/out": 6.416605472564697, "created_at": "2025-01-14T20:20:47.467207+00:00"} {"global_step": 5207, "acc_step": 0, "speed/wps": 12967.491411823612, "speed/FLOPS": 203672271823208.44, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 24.805957794189453, "optim/lr": 0.002999896160012931, "optim/total_tokens": 1364983808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462963, "loss/out": 6.427257537841797, "created_at": "2025-01-14T20:20:52.526014+00:00"} {"global_step": 5208, "acc_step": 0, "speed/wps": 12969.014418633828, "speed/FLOPS": 203696192737993.53, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.001, "optim/grad_norm": 2.8029792308807373, "optim/lr": 0.002999895149422512, "optim/total_tokens": 1365245952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467490, "loss/out": 6.753413200378418, "created_at": "2025-01-14T20:20:57.587847+00:00"} {"global_step": 5209, "acc_step": 0, "speed/wps": 12964.408503331051, "speed/FLOPS": 203623850508972.56, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.533712863922119, "optim/lr": 0.0029998941339384215, "optim/total_tokens": 1365508096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305561, "loss/out": 6.481625080108643, "created_at": "2025-01-14T20:21:02.646823+00:00"} {"global_step": 5210, "acc_step": 0, "speed/wps": 12964.710887574389, "speed/FLOPS": 203628599869033.66, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.696056365966797, "optim/lr": 0.002999893113560662, "optim/total_tokens": 1365770240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 395218, "loss/out": 6.386022567749023, "created_at": "2025-01-14T20:21:07.710786+00:00"} {"global_step": 5211, "acc_step": 0, "speed/wps": 12964.693670471546, "speed/FLOPS": 203628329450774.75, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.935466289520264, "optim/lr": 0.002999892088289237, "optim/total_tokens": 1366032384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 6.304481506347656, "created_at": "2025-01-14T20:21:12.772075+00:00"} {"global_step": 5212, "acc_step": 0, "speed/wps": 12956.726860417439, "speed/FLOPS": 203503199751330.75, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.2271792888641357, "optim/lr": 0.0029998910581241515, "optim/total_tokens": 1366294528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446220, "loss/out": 6.458586692810059, "created_at": "2025-01-14T20:21:17.834264+00:00"} {"global_step": 5213, "acc_step": 0, "speed/wps": 12968.40469068503, "speed/FLOPS": 203686616122703.7, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.4982292652130127, "optim/lr": 0.0029998900230654066, "optim/total_tokens": 1366556672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440105, "loss/out": 6.401307582855225, "created_at": "2025-01-14T20:21:22.892461+00:00"} {"global_step": 5214, "acc_step": 0, "speed/wps": 12974.03619237823, "speed/FLOPS": 203775066595289.62, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.266786575317383, "optim/lr": 0.0029998889831130073, "optim/total_tokens": 1366818816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 6.462757110595703, "created_at": "2025-01-14T20:21:27.948271+00:00"} {"global_step": 5215, "acc_step": 0, "speed/wps": 12973.384254932724, "speed/FLOPS": 203764827021852.03, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.225430965423584, "optim/lr": 0.002999887938266956, "optim/total_tokens": 1367080960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 6.521667957305908, "created_at": "2025-01-14T20:21:33.006997+00:00"} {"global_step": 5216, "acc_step": 0, "speed/wps": 12969.22102359666, "speed/FLOPS": 203699437752839.78, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.4684269428253174, "optim/lr": 0.0029998868885272568, "optim/total_tokens": 1367343104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 6.400464057922363, "created_at": "2025-01-14T20:21:38.064955+00:00"} {"global_step": 5217, "acc_step": 0, "speed/wps": 12975.70492704625, "speed/FLOPS": 203801276366332.62, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0009, "optim/grad_norm": 8.063307762145996, "optim/lr": 0.002999885833893913, "optim/total_tokens": 1367605248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312250, "loss/out": 6.375507354736328, "created_at": "2025-01-14T20:21:43.121791+00:00"} {"global_step": 5218, "acc_step": 0, "speed/wps": 12978.637723289065, "speed/FLOPS": 203847339961410.38, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 118.8099136352539, "optim/lr": 0.0029998847743669278, "optim/total_tokens": 1367867392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318405, "loss/out": 6.390893936157227, "created_at": "2025-01-14T20:21:48.179978+00:00"} {"global_step": 5219, "acc_step": 0, "speed/wps": 12977.134752907798, "speed/FLOPS": 203823733746274.88, "speed/curr_iter_time": 1.2719, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.751652240753174, "optim/lr": 0.002999883709946304, "optim/total_tokens": 1368129536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450746, "loss/out": 6.459828853607178, "created_at": "2025-01-14T20:21:53.243762+00:00"} {"global_step": 5220, "acc_step": 0, "speed/wps": 12977.479207222028, "speed/FLOPS": 203829143874609.7, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 141.2381134033203, "optim/lr": 0.0029998826406320463, "optim/total_tokens": 1368391680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 6.385281562805176, "created_at": "2025-01-14T20:21:58.301663+00:00"} {"global_step": 5221, "acc_step": 0, "speed/wps": 12973.929120281226, "speed/FLOPS": 203773384880873.72, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.005079746246338, "optim/lr": 0.002999881566424158, "optim/total_tokens": 1368653824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450462, "loss/out": 6.4282755851745605, "created_at": "2025-01-14T20:22:03.360457+00:00"} {"global_step": 5222, "acc_step": 0, "speed/wps": 12974.551175920342, "speed/FLOPS": 203783155119480.06, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.87712574005127, "optim/lr": 0.002999880487322642, "optim/total_tokens": 1368915968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445933, "loss/out": 6.382401466369629, "created_at": "2025-01-14T20:22:08.421527+00:00"} {"global_step": 5223, "acc_step": 0, "speed/wps": 12966.881829458374, "speed/FLOPS": 203662697494508.44, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0022, "optim/grad_norm": 97.85797119140625, "optim/lr": 0.002999879403327502, "optim/total_tokens": 1369178112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 6.419084548950195, "created_at": "2025-01-14T20:22:13.480078+00:00"} {"global_step": 5224, "acc_step": 0, "speed/wps": 12973.315277877344, "speed/FLOPS": 203763743642412.88, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.335810661315918, "optim/lr": 0.0029998783144387416, "optim/total_tokens": 1369440256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327402, "loss/out": 6.6561198234558105, "created_at": "2025-01-14T20:22:18.538012+00:00"} {"global_step": 5225, "acc_step": 0, "speed/wps": 12969.305905229907, "speed/FLOPS": 203700770935529.53, "speed/curr_iter_time": 1.2723, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.000565528869629, "optim/lr": 0.002999877220656365, "optim/total_tokens": 1369702400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351058, "loss/out": 6.5279364585876465, "created_at": "2025-01-14T20:22:23.601741+00:00"} {"global_step": 5226, "acc_step": 0, "speed/wps": 12969.113825401499, "speed/FLOPS": 203697754057879.03, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.776139736175537, "optim/lr": 0.0029998761219803744, "optim/total_tokens": 1369964544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 6.509139060974121, "created_at": "2025-01-14T20:22:28.660080+00:00"} {"global_step": 5227, "acc_step": 0, "speed/wps": 12976.38865070017, "speed/FLOPS": 203812015185852.2, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9867652654647827, "optim/lr": 0.002999875018410775, "optim/total_tokens": 1370226688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 6.453738212585449, "created_at": "2025-01-14T20:22:33.715269+00:00"} {"global_step": 5228, "acc_step": 0, "speed/wps": 12969.416142341108, "speed/FLOPS": 203702502360842.6, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 20.412559509277344, "optim/lr": 0.002999873909947569, "optim/total_tokens": 1370488832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 6.330507278442383, "created_at": "2025-01-14T20:22:38.773571+00:00"} {"global_step": 5229, "acc_step": 0, "speed/wps": 12973.322426926754, "speed/FLOPS": 203763855928057.7, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.271695137023926, "optim/lr": 0.0029998727965907606, "optim/total_tokens": 1370750976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501889, "loss/out": 6.489678382873535, "created_at": "2025-01-14T20:22:43.828895+00:00"} {"global_step": 5230, "acc_step": 0, "speed/wps": 12971.63391068445, "speed/FLOPS": 203737335460207.84, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0008, "optim/grad_norm": 8.156571388244629, "optim/lr": 0.0029998716783403534, "optim/total_tokens": 1371013120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443022, "loss/out": 6.535077095031738, "created_at": "2025-01-14T20:22:48.888522+00:00"} {"global_step": 5231, "acc_step": 0, "speed/wps": 12976.31593031766, "speed/FLOPS": 203810873012316.6, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.113263130187988, "optim/lr": 0.002999870555196351, "optim/total_tokens": 1371275264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465870, "loss/out": 6.420656681060791, "created_at": "2025-01-14T20:22:53.943526+00:00"} {"global_step": 5232, "acc_step": 0, "speed/wps": 12978.244801458868, "speed/FLOPS": 203841168584135.97, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.232446670532227, "optim/lr": 0.0029998694271587576, "optim/total_tokens": 1371537408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481758, "loss/out": 6.3816022872924805, "created_at": "2025-01-14T20:22:58.997168+00:00"} {"global_step": 5233, "acc_step": 0, "speed/wps": 12972.751711701301, "speed/FLOPS": 203754892061197.06, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.27734661102295, "optim/lr": 0.002999868294227576, "optim/total_tokens": 1371799552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443016, "loss/out": 6.472309112548828, "created_at": "2025-01-14T20:23:04.053798+00:00"} {"global_step": 5234, "acc_step": 0, "speed/wps": 12970.759324497283, "speed/FLOPS": 203723598882331.03, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.518527507781982, "optim/lr": 0.0029998671564028102, "optim/total_tokens": 1372061696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297534, "loss/out": 6.5051164627075195, "created_at": "2025-01-14T20:23:09.115005+00:00"} {"global_step": 5235, "acc_step": 0, "speed/wps": 12972.125631779854, "speed/FLOPS": 203745058615706.78, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 17.106103897094727, "optim/lr": 0.002999866013684464, "optim/total_tokens": 1372323840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 382950, "loss/out": 6.387042999267578, "created_at": "2025-01-14T20:23:14.170968+00:00"} {"global_step": 5236, "acc_step": 0, "speed/wps": 12972.388546514863, "speed/FLOPS": 203749188052902.78, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.3512020111083984, "optim/lr": 0.0029998648660725415, "optim/total_tokens": 1372585984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434515, "loss/out": 6.589639663696289, "created_at": "2025-01-14T20:23:19.230701+00:00"} {"global_step": 5237, "acc_step": 0, "speed/wps": 12975.840597425753, "speed/FLOPS": 203803407256073.38, "speed/curr_iter_time": 1.2723, "speed/data_load_time": 0.0006, "optim/grad_norm": 16.287832260131836, "optim/lr": 0.002999863713567046, "optim/total_tokens": 1372848128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 6.405832290649414, "created_at": "2025-01-14T20:23:24.291650+00:00"} {"global_step": 5238, "acc_step": 0, "speed/wps": 12968.913469345023, "speed/FLOPS": 203694607190694.0, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 28.93815803527832, "optim/lr": 0.0029998625561679813, "optim/total_tokens": 1373110272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447555, "loss/out": 6.4278974533081055, "created_at": "2025-01-14T20:23:29.348702+00:00"} {"global_step": 5239, "acc_step": 0, "speed/wps": 12974.087403997262, "speed/FLOPS": 203775870944138.7, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.69968032836914, "optim/lr": 0.0029998613938753507, "optim/total_tokens": 1373372416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433993, "loss/out": 6.565406799316406, "created_at": "2025-01-14T20:23:34.404265+00:00"} {"global_step": 5240, "acc_step": 0, "speed/wps": 12971.517873261735, "speed/FLOPS": 203735512933032.0, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 38.038963317871094, "optim/lr": 0.0029998602266891592, "optim/total_tokens": 1373634560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 6.470698356628418, "created_at": "2025-01-14T20:23:39.465340+00:00"} {"global_step": 5241, "acc_step": 0, "speed/wps": 12976.357451905138, "speed/FLOPS": 203811525165904.6, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.335930824279785, "optim/lr": 0.0029998590546094096, "optim/total_tokens": 1373896704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 6.427562713623047, "created_at": "2025-01-14T20:23:44.520597+00:00"} {"global_step": 5242, "acc_step": 0, "speed/wps": 12967.89739298474, "speed/FLOPS": 203678648315220.53, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.814384460449219, "optim/lr": 0.002999857877636106, "optim/total_tokens": 1374158848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 6.627974510192871, "created_at": "2025-01-14T20:23:49.579995+00:00"} {"global_step": 5243, "acc_step": 0, "speed/wps": 12970.301630551668, "speed/FLOPS": 203716410169977.78, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.059389114379883, "optim/lr": 0.0029998566957692525, "optim/total_tokens": 1374420992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445404, "loss/out": 6.521535396575928, "created_at": "2025-01-14T20:23:54.643396+00:00"} {"global_step": 5244, "acc_step": 0, "speed/wps": 12978.455975733583, "speed/FLOPS": 203844485366304.25, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.628957509994507, "optim/lr": 0.0029998555090088525, "optim/total_tokens": 1374683136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318405, "loss/out": 6.4510884284973145, "created_at": "2025-01-14T20:23:59.704647+00:00"} {"global_step": 5245, "acc_step": 0, "speed/wps": 12971.445745457546, "speed/FLOPS": 203734380066753.22, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.92589282989502, "optim/lr": 0.0029998543173549103, "optim/total_tokens": 1374945280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449411, "loss/out": 6.381609916687012, "created_at": "2025-01-14T20:24:04.766022+00:00"} {"global_step": 5246, "acc_step": 0, "speed/wps": 12972.028646668818, "speed/FLOPS": 203743535331266.44, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.44476842880249, "optim/lr": 0.0029998531208074294, "optim/total_tokens": 1375207424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 6.569173812866211, "created_at": "2025-01-14T20:24:09.825526+00:00"} {"global_step": 5247, "acc_step": 0, "speed/wps": 12970.053045865548, "speed/FLOPS": 203712505805891.1, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 29.422725677490234, "optim/lr": 0.002999851919366414, "optim/total_tokens": 1375469568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434807, "loss/out": 6.343944549560547, "created_at": "2025-01-14T20:24:14.884451+00:00"} {"global_step": 5248, "acc_step": 0, "speed/wps": 12969.076026317807, "speed/FLOPS": 203697160371328.28, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 18.460092544555664, "optim/lr": 0.0029998507130318684, "optim/total_tokens": 1375731712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 6.4080352783203125, "created_at": "2025-01-14T20:24:19.941775+00:00"} {"global_step": 5249, "acc_step": 0, "speed/wps": 12974.581574194928, "speed/FLOPS": 203783632566154.03, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.645914554595947, "optim/lr": 0.0029998495018037953, "optim/total_tokens": 1375993856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477242, "loss/out": 6.474689960479736, "created_at": "2025-01-14T20:24:25.000445+00:00"} {"global_step": 5250, "acc_step": 0, "speed/wps": 12975.031718335129, "speed/FLOPS": 203790702698438.75, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 26.534305572509766, "optim/lr": 0.0029998482856821995, "optim/total_tokens": 1376256000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299996, "loss/out": 6.322795867919922, "created_at": "2025-01-14T20:24:30.057059+00:00"} {"global_step": 5251, "acc_step": 0, "speed/wps": 12975.898671302293, "speed/FLOPS": 203804319386106.28, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.555633544921875, "optim/lr": 0.0029998470646670854, "optim/total_tokens": 1376518144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307966, "loss/out": 6.4299726486206055, "created_at": "2025-01-14T20:24:35.115482+00:00"} {"global_step": 5252, "acc_step": 0, "speed/wps": 12967.479074133738, "speed/FLOPS": 203672078042835.22, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.881318092346191, "optim/lr": 0.0029998458387584565, "optim/total_tokens": 1376780288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422503, "loss/out": 6.39460563659668, "created_at": "2025-01-14T20:24:40.177872+00:00"} {"global_step": 5253, "acc_step": 0, "speed/wps": 12972.994872700154, "speed/FLOPS": 203758711238822.28, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.636483192443848, "optim/lr": 0.002999844607956316, "optim/total_tokens": 1377042432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 6.351295471191406, "created_at": "2025-01-14T20:24:45.233734+00:00"} {"global_step": 5254, "acc_step": 0, "speed/wps": 12970.119038226381, "speed/FLOPS": 203713542306600.97, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.250462532043457, "optim/lr": 0.0029998433722606696, "optim/total_tokens": 1377304576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471454, "loss/out": 6.417538642883301, "created_at": "2025-01-14T20:24:50.296884+00:00"} {"global_step": 5255, "acc_step": 0, "speed/wps": 12970.410956993417, "speed/FLOPS": 203718127291975.7, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.314906120300293, "optim/lr": 0.0029998421316715196, "optim/total_tokens": 1377566720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 6.354601860046387, "created_at": "2025-01-14T20:24:55.361719+00:00"} {"global_step": 5256, "acc_step": 0, "speed/wps": 12974.928749482862, "speed/FLOPS": 203789085431119.88, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.125028610229492, "optim/lr": 0.002999840886188871, "optim/total_tokens": 1377828864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431605, "loss/out": 6.3444743156433105, "created_at": "2025-01-14T20:25:00.421921+00:00"} {"global_step": 5257, "acc_step": 0, "speed/wps": 12968.164944802442, "speed/FLOPS": 203682850584172.25, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0009, "optim/grad_norm": 207.09849548339844, "optim/lr": 0.0029998396358127282, "optim/total_tokens": 1378091008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 6.289365768432617, "created_at": "2025-01-14T20:25:05.486091+00:00"} {"global_step": 5258, "acc_step": 0, "speed/wps": 12974.680473037553, "speed/FLOPS": 203785185908378.72, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.001, "optim/grad_norm": 3.6446118354797363, "optim/lr": 0.0029998383805430948, "optim/total_tokens": 1378353152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432656, "loss/out": 6.3857035636901855, "created_at": "2025-01-14T20:25:10.543182+00:00"} {"global_step": 5259, "acc_step": 0, "speed/wps": 12972.604600278813, "speed/FLOPS": 203752581474154.62, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.190697193145752, "optim/lr": 0.0029998371203799746, "optim/total_tokens": 1378615296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287143, "loss/out": 6.415162563323975, "created_at": "2025-01-14T20:25:15.602440+00:00"} {"global_step": 5260, "acc_step": 0, "speed/wps": 12976.890567996252, "speed/FLOPS": 203819898486693.8, "speed/curr_iter_time": 1.272, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.462734222412109, "optim/lr": 0.0029998358553233718, "optim/total_tokens": 1378877440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 409931, "loss/out": 6.463441848754883, "created_at": "2025-01-14T20:25:20.656110+00:00"} {"global_step": 5261, "acc_step": 0, "speed/wps": 12978.703362583583, "speed/FLOPS": 203848370916728.66, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4578254222869873, "optim/lr": 0.002999834585373291, "optim/total_tokens": 1379139584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444119, "loss/out": 6.369264602661133, "created_at": "2025-01-14T20:25:25.713343+00:00"} {"global_step": 5262, "acc_step": 0, "speed/wps": 12975.4112195579, "speed/FLOPS": 203796663286631.72, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.7613461017608643, "optim/lr": 0.0029998333105297364, "optim/total_tokens": 1379401728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 6.311558723449707, "created_at": "2025-01-14T20:25:30.767902+00:00"} {"global_step": 5263, "acc_step": 0, "speed/wps": 12976.15167474754, "speed/FLOPS": 203808293152877.06, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.598537445068359, "optim/lr": 0.0029998320307927113, "optim/total_tokens": 1379663872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463987, "loss/out": 6.476153373718262, "created_at": "2025-01-14T20:25:35.825595+00:00"} {"global_step": 5264, "acc_step": 0, "speed/wps": 12970.666503085033, "speed/FLOPS": 203722140994501.66, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.782238006591797, "optim/lr": 0.0029998307461622203, "optim/total_tokens": 1379926016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450756, "loss/out": 6.471925735473633, "created_at": "2025-01-14T20:25:40.882583+00:00"} {"global_step": 5265, "acc_step": 0, "speed/wps": 12966.768627205238, "speed/FLOPS": 203660919497566.72, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0012, "optim/grad_norm": 2.0898923873901367, "optim/lr": 0.002999829456638268, "optim/total_tokens": 1380188160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 6.382741928100586, "created_at": "2025-01-14T20:25:45.941734+00:00"} {"global_step": 5266, "acc_step": 0, "speed/wps": 12967.279387479286, "speed/FLOPS": 203668941689528.8, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.374957084655762, "optim/lr": 0.0029998281622208582, "optim/total_tokens": 1380450304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451278, "loss/out": 6.35012674331665, "created_at": "2025-01-14T20:25:51.001919+00:00"} {"global_step": 5267, "acc_step": 0, "speed/wps": 12970.0352743009, "speed/FLOPS": 203712226679047.62, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.001, "optim/grad_norm": 87.93417358398438, "optim/lr": 0.0029998268629099953, "optim/total_tokens": 1380712448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332731, "loss/out": 6.42568826675415, "created_at": "2025-01-14T20:25:56.059534+00:00"} {"global_step": 5268, "acc_step": 0, "speed/wps": 12969.959335715508, "speed/FLOPS": 203711033959213.47, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.381345748901367, "optim/lr": 0.0029998255587056835, "optim/total_tokens": 1380974592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321253, "loss/out": 6.588940620422363, "created_at": "2025-01-14T20:26:01.116006+00:00"} {"global_step": 5269, "acc_step": 0, "speed/wps": 12969.0651484258, "speed/FLOPS": 203696989519086.5, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.9372549057006836, "optim/lr": 0.0029998242496079275, "optim/total_tokens": 1381236736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312011, "loss/out": 6.524238586425781, "created_at": "2025-01-14T20:26:06.173333+00:00"} {"global_step": 5270, "acc_step": 0, "speed/wps": 12974.005930613368, "speed/FLOPS": 203774591292688.47, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.713768720626831, "optim/lr": 0.0029998229356167304, "optim/total_tokens": 1381498880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335427, "loss/out": 6.442023277282715, "created_at": "2025-01-14T20:26:11.228707+00:00"} {"global_step": 5271, "acc_step": 0, "speed/wps": 12963.992389476343, "speed/FLOPS": 203617314869084.03, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.725388526916504, "optim/lr": 0.0029998216167320974, "optim/total_tokens": 1381761024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450756, "loss/out": 6.395455837249756, "created_at": "2025-01-14T20:26:16.291984+00:00"} {"global_step": 5272, "acc_step": 0, "speed/wps": 12970.701122084063, "speed/FLOPS": 203722684733452.9, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.877999186515808, "optim/lr": 0.002999820292954033, "optim/total_tokens": 1382023168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472789, "loss/out": 6.544624328613281, "created_at": "2025-01-14T20:26:21.353841+00:00"} {"global_step": 5273, "acc_step": 0, "speed/wps": 12972.688701121817, "speed/FLOPS": 203753902393460.66, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.1950511932373047, "optim/lr": 0.0029998189642825406, "optim/total_tokens": 1382285312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437723, "loss/out": 6.491257667541504, "created_at": "2025-01-14T20:26:26.411234+00:00"} {"global_step": 5274, "acc_step": 0, "speed/wps": 12971.64651691919, "speed/FLOPS": 203737533458447.22, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.785109281539917, "optim/lr": 0.0029998176307176254, "optim/total_tokens": 1382547456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450746, "loss/out": 6.368411064147949, "created_at": "2025-01-14T20:26:31.468085+00:00"} {"global_step": 5275, "acc_step": 0, "speed/wps": 12970.267556365216, "speed/FLOPS": 203715874988056.7, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.37083911895752, "optim/lr": 0.002999816292259291, "optim/total_tokens": 1382809600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 6.433081150054932, "created_at": "2025-01-14T20:26:36.526850+00:00"} {"global_step": 5276, "acc_step": 0, "speed/wps": 12976.191214007413, "speed/FLOPS": 203808914171286.0, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.168135166168213, "optim/lr": 0.0029998149489075423, "optim/total_tokens": 1383071744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422290, "loss/out": 6.543015480041504, "created_at": "2025-01-14T20:26:41.582393+00:00"} {"global_step": 5277, "acc_step": 0, "speed/wps": 12973.669742928918, "speed/FLOPS": 203769311003133.88, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.152198553085327, "optim/lr": 0.0029998136006623836, "optim/total_tokens": 1383333888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293593, "loss/out": 6.472856044769287, "created_at": "2025-01-14T20:26:46.641528+00:00"} {"global_step": 5278, "acc_step": 0, "speed/wps": 12978.441450894084, "speed/FLOPS": 203844257233740.7, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.261991500854492, "optim/lr": 0.0029998122475238197, "optim/total_tokens": 1383596032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 405960, "loss/out": 6.3506622314453125, "created_at": "2025-01-14T20:26:51.694915+00:00"} {"global_step": 5279, "acc_step": 0, "speed/wps": 12973.90184184081, "speed/FLOPS": 203772956435482.47, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1662869453430176, "optim/lr": 0.0029998108894918538, "optim/total_tokens": 1383858176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456045, "loss/out": 6.410778999328613, "created_at": "2025-01-14T20:26:56.750238+00:00"} {"global_step": 5280, "acc_step": 0, "speed/wps": 12976.797938799813, "speed/FLOPS": 203818443617877.88, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0012, "optim/grad_norm": 5.188498497009277, "optim/lr": 0.002999809526566491, "optim/total_tokens": 1384120320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447839, "loss/out": 6.276769161224365, "created_at": "2025-01-14T20:27:01.807636+00:00"} {"global_step": 5281, "acc_step": 0, "speed/wps": 12978.162014439531, "speed/FLOPS": 203839868300236.0, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.243088722229004, "optim/lr": 0.0029998081587477364, "optim/total_tokens": 1384382464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474937, "loss/out": 6.334172248840332, "created_at": "2025-01-14T20:27:06.861560+00:00"} {"global_step": 5282, "acc_step": 0, "speed/wps": 12969.977052089305, "speed/FLOPS": 203711312219208.8, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.173572540283203, "optim/lr": 0.0029998067860355936, "optim/total_tokens": 1384644608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 6.282904624938965, "created_at": "2025-01-14T20:27:11.918119+00:00"} {"global_step": 5283, "acc_step": 0, "speed/wps": 12963.50031278876, "speed/FLOPS": 203609586128521.28, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.0320682525634766, "optim/lr": 0.0029998054084300675, "optim/total_tokens": 1384906752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444878, "loss/out": 6.206297874450684, "created_at": "2025-01-14T20:27:16.977776+00:00"} {"global_step": 5284, "acc_step": 0, "speed/wps": 12971.758209420557, "speed/FLOPS": 203739287742662.78, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.755591869354248, "optim/lr": 0.0029998040259311626, "optim/total_tokens": 1385168896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462672, "loss/out": 6.3813581466674805, "created_at": "2025-01-14T20:27:22.042419+00:00"} {"global_step": 5285, "acc_step": 0, "speed/wps": 12980.959882711479, "speed/FLOPS": 203883812666120.84, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.184565544128418, "optim/lr": 0.002999802638538883, "optim/total_tokens": 1385431040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466369, "loss/out": 6.237561225891113, "created_at": "2025-01-14T20:27:27.099099+00:00"} {"global_step": 5286, "acc_step": 0, "speed/wps": 12966.837386520376, "speed/FLOPS": 203661999457096.16, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.923243999481201, "optim/lr": 0.002999801246253234, "optim/total_tokens": 1385693184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338641, "loss/out": 6.264875411987305, "created_at": "2025-01-14T20:27:32.157625+00:00"} {"global_step": 5287, "acc_step": 0, "speed/wps": 12976.768951615551, "speed/FLOPS": 203817988334316.5, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3853102922439575, "optim/lr": 0.0029997998490742194, "optim/total_tokens": 1385955328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289212, "loss/out": 6.267411231994629, "created_at": "2025-01-14T20:27:37.215035+00:00"} {"global_step": 5288, "acc_step": 0, "speed/wps": 12976.080861693457, "speed/FLOPS": 203807180936555.34, "speed/curr_iter_time": 1.2721, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.983438014984131, "optim/lr": 0.0029997984470018437, "optim/total_tokens": 1386217472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434230, "loss/out": 6.260378360748291, "created_at": "2025-01-14T20:27:42.269557+00:00"} {"global_step": 5289, "acc_step": 0, "speed/wps": 12969.086311451996, "speed/FLOPS": 203697321913493.97, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.960646629333496, "optim/lr": 0.0029997970400361122, "optim/total_tokens": 1386479616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449135, "loss/out": 6.361510753631592, "created_at": "2025-01-14T20:27:47.332979+00:00"} {"global_step": 5290, "acc_step": 0, "speed/wps": 12980.057612914085, "speed/FLOPS": 203869641279104.12, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.921150803565979, "optim/lr": 0.0029997956281770286, "optim/total_tokens": 1386741760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 6.281381607055664, "created_at": "2025-01-14T20:27:52.390509+00:00"} {"global_step": 5291, "acc_step": 0, "speed/wps": 12978.38034782597, "speed/FLOPS": 203843297526091.56, "speed/curr_iter_time": 1.2718, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7805335521698, "optim/lr": 0.0029997942114245984, "optim/total_tokens": 1387003904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433179, "loss/out": 6.232541084289551, "created_at": "2025-01-14T20:27:57.446725+00:00"} {"global_step": 5292, "acc_step": 0, "speed/wps": 12976.201584608441, "speed/FLOPS": 203809077055825.84, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.852306842803955, "optim/lr": 0.002999792789778826, "optim/total_tokens": 1387266048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 6.156598091125488, "created_at": "2025-01-14T20:28:02.501586+00:00"} {"global_step": 5293, "acc_step": 0, "speed/wps": 12973.22769608417, "speed/FLOPS": 203762368050001.1, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0009, "optim/grad_norm": 2.030489683151245, "optim/lr": 0.0029997913632397156, "optim/total_tokens": 1387528192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 6.186341762542725, "created_at": "2025-01-14T20:28:07.558239+00:00"} {"global_step": 5294, "acc_step": 0, "speed/wps": 12974.841478439916, "speed/FLOPS": 203787714719467.44, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1976094245910645, "optim/lr": 0.0029997899318072716, "optim/total_tokens": 1387790336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453375, "loss/out": 6.199376583099365, "created_at": "2025-01-14T20:28:12.613048+00:00"} {"global_step": 5295, "acc_step": 0, "speed/wps": 12974.219619823594, "speed/FLOPS": 203777947575374.8, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.93301010131836, "optim/lr": 0.0029997884954814998, "optim/total_tokens": 1388052480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 6.341550350189209, "created_at": "2025-01-14T20:28:17.667974+00:00"} {"global_step": 5296, "acc_step": 0, "speed/wps": 12977.005232171085, "speed/FLOPS": 203821699445124.62, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0009, "optim/grad_norm": 2.0005993843078613, "optim/lr": 0.002999787054262404, "optim/total_tokens": 1388314624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 278077, "loss/out": 6.20134973526001, "created_at": "2025-01-14T20:28:22.730562+00:00"} {"global_step": 5297, "acc_step": 0, "speed/wps": 12977.244357312818, "speed/FLOPS": 203825455234067.94, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6074881553649902, "optim/lr": 0.002999785608149989, "optim/total_tokens": 1388576768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334895, "loss/out": 6.253701686859131, "created_at": "2025-01-14T20:28:27.786076+00:00"} {"global_step": 5298, "acc_step": 0, "speed/wps": 12978.022044617843, "speed/FLOPS": 203837669881847.7, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.4375574588775635, "optim/lr": 0.00299978415714426, "optim/total_tokens": 1388838912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431299, "loss/out": 6.20496940612793, "created_at": "2025-01-14T20:28:32.843649+00:00"} {"global_step": 5299, "acc_step": 0, "speed/wps": 12978.874304105188, "speed/FLOPS": 203851055788223.72, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.001, "optim/grad_norm": 1.7462018728256226, "optim/lr": 0.0029997827012452215, "optim/total_tokens": 1389101056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455517, "loss/out": 6.246993541717529, "created_at": "2025-01-14T20:28:37.899896+00:00"} {"global_step": 5300, "acc_step": 0, "speed/wps": 12979.10476271387, "speed/FLOPS": 203854675457356.4, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3745296001434326, "optim/lr": 0.0029997812404528776, "optim/total_tokens": 1389363200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 6.271346569061279, "created_at": "2025-01-14T20:28:42.955457+00:00"} {"global_step": 5301, "acc_step": 0, "speed/wps": 12966.693780567552, "speed/FLOPS": 203659743928271.8, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.087928295135498, "optim/lr": 0.0029997797747672345, "optim/total_tokens": 1389625344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445404, "loss/out": 6.200108528137207, "created_at": "2025-01-14T20:28:48.014196+00:00"} {"global_step": 5302, "acc_step": 0, "speed/wps": 12966.003356604586, "speed/FLOPS": 203648899871191.47, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 33.840267181396484, "optim/lr": 0.0029997783041882955, "optim/total_tokens": 1389887488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 6.273739337921143, "created_at": "2025-01-14T20:28:53.073685+00:00"} {"global_step": 5303, "acc_step": 0, "speed/wps": 12969.919633619511, "speed/FLOPS": 203710410383239.44, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.976975917816162, "optim/lr": 0.002999776828716066, "optim/total_tokens": 1390149632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458430, "loss/out": 6.422893047332764, "created_at": "2025-01-14T20:28:58.135373+00:00"} {"global_step": 5304, "acc_step": 0, "speed/wps": 12973.111599053267, "speed/FLOPS": 203760544586596.72, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.297846555709839, "optim/lr": 0.002999775348350551, "optim/total_tokens": 1390411776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317214, "loss/out": 6.236907005310059, "created_at": "2025-01-14T20:29:03.190849+00:00"} {"global_step": 5305, "acc_step": 0, "speed/wps": 12974.956647086, "speed/FLOPS": 203789523601312.5, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0009, "optim/grad_norm": 7.278539657592773, "optim/lr": 0.002999773863091755, "optim/total_tokens": 1390673920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299469, "loss/out": 6.260066509246826, "created_at": "2025-01-14T20:29:08.245938+00:00"} {"global_step": 5306, "acc_step": 0, "speed/wps": 12966.573107864035, "speed/FLOPS": 203657848597641.4, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7216086387634277, "optim/lr": 0.0029997723729396834, "optim/total_tokens": 1390936064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 403615, "loss/out": 6.303066253662109, "created_at": "2025-01-14T20:29:13.322153+00:00"} {"global_step": 5307, "acc_step": 0, "speed/wps": 12971.06761186227, "speed/FLOPS": 203728440959029.4, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0009, "optim/grad_norm": 2.3922841548919678, "optim/lr": 0.0029997708778943402, "optim/total_tokens": 1391198208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 6.245087146759033, "created_at": "2025-01-14T20:29:18.383737+00:00"} {"global_step": 5308, "acc_step": 0, "speed/wps": 12971.985486046544, "speed/FLOPS": 203742857434384.72, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 21.957530975341797, "optim/lr": 0.0029997693779557303, "optim/total_tokens": 1391460352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 6.288706302642822, "created_at": "2025-01-14T20:29:23.440837+00:00"} {"global_step": 5309, "acc_step": 0, "speed/wps": 12962.002338347165, "speed/FLOPS": 203586058381483.47, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0011, "optim/grad_norm": 4.222452163696289, "optim/lr": 0.00299976787312386, "optim/total_tokens": 1391722496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443543, "loss/out": 6.344672679901123, "created_at": "2025-01-14T20:29:28.507480+00:00"} {"global_step": 5310, "acc_step": 0, "speed/wps": 12967.769571667288, "speed/FLOPS": 203676640705785.03, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.791438102722168, "optim/lr": 0.0029997663633987326, "optim/total_tokens": 1391984640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 6.34238338470459, "created_at": "2025-01-14T20:29:33.570094+00:00"} {"global_step": 5311, "acc_step": 0, "speed/wps": 12962.75279979669, "speed/FLOPS": 203597845409790.34, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.292914390563965, "optim/lr": 0.002999764848780354, "optim/total_tokens": 1392246784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439871, "loss/out": 6.313944339752197, "created_at": "2025-01-14T20:29:38.632736+00:00"} {"global_step": 5312, "acc_step": 0, "speed/wps": 12968.75632047088, "speed/FLOPS": 203692138951680.4, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.454116106033325, "optim/lr": 0.002999763329268729, "optim/total_tokens": 1392508928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300465, "loss/out": 6.218327522277832, "created_at": "2025-01-14T20:29:43.693817+00:00"} {"global_step": 5313, "acc_step": 0, "speed/wps": 12969.855259271973, "speed/FLOPS": 203709399295651.75, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 21.412609100341797, "optim/lr": 0.0029997618048638623, "optim/total_tokens": 1392771072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421165, "loss/out": 6.331210613250732, "created_at": "2025-01-14T20:29:48.752546+00:00"} {"global_step": 5314, "acc_step": 0, "speed/wps": 12968.348667529452, "speed/FLOPS": 203685736202061.44, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.391750335693359, "optim/lr": 0.0029997602755657583, "optim/total_tokens": 1393033216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453945, "loss/out": 6.358999252319336, "created_at": "2025-01-14T20:29:53.812000+00:00"} {"global_step": 5315, "acc_step": 0, "speed/wps": 12968.497534697975, "speed/FLOPS": 203688074365508.03, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 110.59725189208984, "optim/lr": 0.0029997587413744236, "optim/total_tokens": 1393295360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 6.353841781616211, "created_at": "2025-01-14T20:29:58.874261+00:00"} {"global_step": 5316, "acc_step": 0, "speed/wps": 12975.48297994451, "speed/FLOPS": 203797790382112.53, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.67689037322998, "optim/lr": 0.002999757202289861, "optim/total_tokens": 1393557504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433179, "loss/out": 6.343759536743164, "created_at": "2025-01-14T20:30:03.928661+00:00"} {"global_step": 5317, "acc_step": 0, "speed/wps": 12967.607793023806, "speed/FLOPS": 203674099749882.0, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0009, "optim/grad_norm": 3.8498520851135254, "optim/lr": 0.002999755658312078, "optim/total_tokens": 1393819648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 6.336061477661133, "created_at": "2025-01-14T20:30:08.994827+00:00"} {"global_step": 5318, "acc_step": 0, "speed/wps": 12968.593847993883, "speed/FLOPS": 203689587098168.6, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.6043442487716675, "optim/lr": 0.002999754109441077, "optim/total_tokens": 1394081792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429739, "loss/out": 6.154733180999756, "created_at": "2025-01-14T20:30:14.055357+00:00"} {"global_step": 5319, "acc_step": 0, "speed/wps": 12973.827120933902, "speed/FLOPS": 203771782840962.2, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.107177972793579, "optim/lr": 0.0029997525556768657, "optim/total_tokens": 1394343936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298409, "loss/out": 6.278085231781006, "created_at": "2025-01-14T20:30:19.110740+00:00"} {"global_step": 5320, "acc_step": 0, "speed/wps": 12978.088194690672, "speed/FLOPS": 203838708859641.22, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.9970028400421143, "optim/lr": 0.0029997509970194473, "optim/total_tokens": 1394606080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330862, "loss/out": 6.266251564025879, "created_at": "2025-01-14T20:30:24.167945+00:00"} {"global_step": 5321, "acc_step": 0, "speed/wps": 12977.17497304366, "speed/FLOPS": 203824365458776.84, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0012, "optim/grad_norm": 2.808593988418579, "optim/lr": 0.0029997494334688276, "optim/total_tokens": 1394868224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358542, "loss/out": 6.168434143066406, "created_at": "2025-01-14T20:30:29.226294+00:00"} {"global_step": 5322, "acc_step": 0, "speed/wps": 12969.761654770384, "speed/FLOPS": 203707929108328.22, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.153292179107666, "optim/lr": 0.0029997478650250118, "optim/total_tokens": 1395130368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440108, "loss/out": 6.225673675537109, "created_at": "2025-01-14T20:30:34.284882+00:00"} {"global_step": 5323, "acc_step": 0, "speed/wps": 12974.855066989738, "speed/FLOPS": 203787928146315.0, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.295586585998535, "optim/lr": 0.002999746291688005, "optim/total_tokens": 1395392512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443016, "loss/out": 6.347939968109131, "created_at": "2025-01-14T20:30:39.339548+00:00"} {"global_step": 5324, "acc_step": 0, "speed/wps": 12975.039570255658, "speed/FLOPS": 203790826023640.34, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.233962059020996, "optim/lr": 0.002999744713457812, "optim/total_tokens": 1395654656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 6.2195143699646, "created_at": "2025-01-14T20:30:44.400430+00:00"} {"global_step": 5325, "acc_step": 0, "speed/wps": 12968.84222377979, "speed/FLOPS": 203693488181190.34, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 22.187828063964844, "optim/lr": 0.0029997431303344377, "optim/total_tokens": 1395916800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431320, "loss/out": 6.254790306091309, "created_at": "2025-01-14T20:30:49.458994+00:00"} {"global_step": 5326, "acc_step": 0, "speed/wps": 12969.007477498284, "speed/FLOPS": 203696083717921.72, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 19.552324295043945, "optim/lr": 0.002999741542317888, "optim/total_tokens": 1396178944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 6.166236400604248, "created_at": "2025-01-14T20:30:54.518982+00:00"} {"global_step": 5327, "acc_step": 0, "speed/wps": 12973.991523374238, "speed/FLOPS": 203774365007200.3, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0468204021453857, "optim/lr": 0.002999739949408168, "optim/total_tokens": 1396441088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451275, "loss/out": 6.32557487487793, "created_at": "2025-01-14T20:30:59.578000+00:00"} {"global_step": 5328, "acc_step": 0, "speed/wps": 12969.702997562335, "speed/FLOPS": 203707007816272.38, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7334020137786865, "optim/lr": 0.002999738351605282, "optim/total_tokens": 1396703232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330862, "loss/out": 6.180265426635742, "created_at": "2025-01-14T20:31:04.636938+00:00"} {"global_step": 5329, "acc_step": 0, "speed/wps": 12970.574766022097, "speed/FLOPS": 203720700137868.3, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.7866036891937256, "optim/lr": 0.0029997367489092365, "optim/total_tokens": 1396965376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363341, "loss/out": 6.428006172180176, "created_at": "2025-01-14T20:31:09.698698+00:00"} {"global_step": 5330, "acc_step": 0, "speed/wps": 12966.181859900144, "speed/FLOPS": 203651703510737.44, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.7011775970458984, "optim/lr": 0.0029997351413200354, "optim/total_tokens": 1397227520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455003, "loss/out": 6.225646018981934, "created_at": "2025-01-14T20:31:14.761532+00:00"} {"global_step": 5331, "acc_step": 0, "speed/wps": 12963.955757087495, "speed/FLOPS": 203616739507078.97, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 19.517648696899414, "optim/lr": 0.002999733528837685, "optim/total_tokens": 1397489664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452081, "loss/out": 6.210910797119141, "created_at": "2025-01-14T20:31:19.823977+00:00"} {"global_step": 5332, "acc_step": 0, "speed/wps": 12972.939527372626, "speed/FLOPS": 203757841964401.0, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7275152206420898, "optim/lr": 0.0029997319114621903, "optim/total_tokens": 1397751808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427880, "loss/out": 6.309333324432373, "created_at": "2025-01-14T20:31:24.879752+00:00"} {"global_step": 5333, "acc_step": 0, "speed/wps": 12962.234251590367, "speed/FLOPS": 203589700897650.12, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.768336296081543, "optim/lr": 0.002999730289193556, "optim/total_tokens": 1398013952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 6.28996467590332, "created_at": "2025-01-14T20:31:29.943909+00:00"} {"global_step": 5334, "acc_step": 0, "speed/wps": 12972.012493385508, "speed/FLOPS": 203743281621755.1, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0643835067749023, "optim/lr": 0.002999728662031789, "optim/total_tokens": 1398276096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 6.209006309509277, "created_at": "2025-01-14T20:31:35.003435+00:00"} {"global_step": 5335, "acc_step": 0, "speed/wps": 12975.022586482164, "speed/FLOPS": 203790559270139.47, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.445176601409912, "optim/lr": 0.0029997270299768923, "optim/total_tokens": 1398538240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 6.2645182609558105, "created_at": "2025-01-14T20:31:40.058260+00:00"} {"global_step": 5336, "acc_step": 0, "speed/wps": 12971.670851899526, "speed/FLOPS": 203737915672753.16, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.199217796325684, "optim/lr": 0.002999725393028873, "optim/total_tokens": 1398800384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 6.233949661254883, "created_at": "2025-01-14T20:31:45.114690+00:00"} {"global_step": 5337, "acc_step": 0, "speed/wps": 12969.33712271333, "speed/FLOPS": 203701261249004.0, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.097217321395874, "optim/lr": 0.0029997237511877354, "optim/total_tokens": 1399062528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435329, "loss/out": 6.147089004516602, "created_at": "2025-01-14T20:31:50.176428+00:00"} {"global_step": 5338, "acc_step": 0, "speed/wps": 12975.90927831714, "speed/FLOPS": 203804485983850.03, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.49268913269043, "optim/lr": 0.002999722104453485, "optim/total_tokens": 1399324672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321630, "loss/out": 6.141417503356934, "created_at": "2025-01-14T20:31:55.235598+00:00"} {"global_step": 5339, "acc_step": 0, "speed/wps": 12972.133786276221, "speed/FLOPS": 203745186693277.66, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.27069664001465, "optim/lr": 0.0029997204528261286, "optim/total_tokens": 1399586816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459528, "loss/out": 6.260766983032227, "created_at": "2025-01-14T20:32:00.291525+00:00"} {"global_step": 5340, "acc_step": 0, "speed/wps": 12975.13008813058, "speed/FLOPS": 203792247731250.56, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.001, "optim/grad_norm": 16.289155960083008, "optim/lr": 0.0029997187963056696, "optim/total_tokens": 1399848960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 6.392082691192627, "created_at": "2025-01-14T20:32:05.347257+00:00"} {"global_step": 5341, "acc_step": 0, "speed/wps": 12976.140552878598, "speed/FLOPS": 203808118468642.28, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 21.289478302001953, "optim/lr": 0.002999717134892115, "optim/total_tokens": 1400111104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 6.437381744384766, "created_at": "2025-01-14T20:32:10.404981+00:00"} {"global_step": 5342, "acc_step": 0, "speed/wps": 12965.61917369286, "speed/FLOPS": 203642865750640.2, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.4660279750823975, "optim/lr": 0.0029997154685854686, "optim/total_tokens": 1400373248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447555, "loss/out": 6.394773483276367, "created_at": "2025-01-14T20:32:15.464986+00:00"} {"global_step": 5343, "acc_step": 0, "speed/wps": 12969.612247028732, "speed/FLOPS": 203705582454437.8, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.0243701934814453, "optim/lr": 0.0029997137973857375, "optim/total_tokens": 1400635392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 6.332767486572266, "created_at": "2025-01-14T20:32:20.522745+00:00"} {"global_step": 5344, "acc_step": 0, "speed/wps": 12965.972649707188, "speed/FLOPS": 203648417577172.16, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.739999771118164, "optim/lr": 0.0029997121212929257, "optim/total_tokens": 1400897536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435852, "loss/out": 6.246124744415283, "created_at": "2025-01-14T20:32:25.590215+00:00"} {"global_step": 5345, "acc_step": 0, "speed/wps": 12972.584898942267, "speed/FLOPS": 203752272037591.66, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.97853422164917, "optim/lr": 0.0029997104403070396, "optim/total_tokens": 1401159680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361759, "loss/out": 6.2411909103393555, "created_at": "2025-01-14T20:32:30.647234+00:00"} {"global_step": 5346, "acc_step": 0, "speed/wps": 12961.987644978008, "speed/FLOPS": 203585827601930.03, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.717122554779053, "optim/lr": 0.002999708754428085, "optim/total_tokens": 1401421824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458734, "loss/out": 6.26906156539917, "created_at": "2025-01-14T20:32:35.709874+00:00"} {"global_step": 5347, "acc_step": 0, "speed/wps": 12970.559304900862, "speed/FLOPS": 203720457299713.5, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.4321463108062744, "optim/lr": 0.002999707063656066, "optim/total_tokens": 1401683968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444119, "loss/out": 6.26690149307251, "created_at": "2025-01-14T20:32:40.770574+00:00"} {"global_step": 5348, "acc_step": 0, "speed/wps": 12970.582404551604, "speed/FLOPS": 203720820111470.66, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5402439832687378, "optim/lr": 0.0029997053679909893, "optim/total_tokens": 1401946112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436666, "loss/out": 6.233875751495361, "created_at": "2025-01-14T20:32:45.831329+00:00"} {"global_step": 5349, "acc_step": 0, "speed/wps": 12968.944866592476, "speed/FLOPS": 203695100327609.6, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.684735298156738, "optim/lr": 0.0029997036674328597, "optim/total_tokens": 1402208256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 6.2515668869018555, "created_at": "2025-01-14T20:32:50.892160+00:00"} {"global_step": 5350, "acc_step": 0, "speed/wps": 12972.10199097238, "speed/FLOPS": 203744687304339.5, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.2979068756103516, "optim/lr": 0.002999701961981683, "optim/total_tokens": 1402470400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 6.24857759475708, "created_at": "2025-01-14T20:32:55.951149+00:00"} {"global_step": 5351, "acc_step": 0, "speed/wps": 12967.43020153211, "speed/FLOPS": 203671310431468.66, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.721938133239746, "optim/lr": 0.0029997002516374657, "optim/total_tokens": 1402732544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461100, "loss/out": 6.239759922027588, "created_at": "2025-01-14T20:33:01.008971+00:00"} {"global_step": 5352, "acc_step": 0, "speed/wps": 12973.81841893917, "speed/FLOPS": 203771646164177.16, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.735725402832031, "optim/lr": 0.0029996985364002123, "optim/total_tokens": 1402994688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 6.293067932128906, "created_at": "2025-01-14T20:33:06.069083+00:00"} {"global_step": 5353, "acc_step": 0, "speed/wps": 12968.60022224556, "speed/FLOPS": 203689687214550.34, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.312656879425049, "optim/lr": 0.0029996968162699284, "optim/total_tokens": 1403256832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437432, "loss/out": 6.290878772735596, "created_at": "2025-01-14T20:33:11.134768+00:00"} {"global_step": 5354, "acc_step": 0, "speed/wps": 12968.908354399806, "speed/FLOPS": 203694526853450.7, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7587743997573853, "optim/lr": 0.0029996950912466198, "optim/total_tokens": 1403518976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356143, "loss/out": 6.323208808898926, "created_at": "2025-01-14T20:33:16.191935+00:00"} {"global_step": 5355, "acc_step": 0, "speed/wps": 12975.535282050003, "speed/FLOPS": 203798611858549.9, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.6256747245788574, "optim/lr": 0.002999693361330293, "optim/total_tokens": 1403781120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453382, "loss/out": 6.106078147888184, "created_at": "2025-01-14T20:33:21.251041+00:00"} {"global_step": 5356, "acc_step": 0, "speed/wps": 12973.97230101436, "speed/FLOPS": 203774063093624.12, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.0665700435638428, "optim/lr": 0.0029996916265209515, "optim/total_tokens": 1404043264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461913, "loss/out": 6.284806251525879, "created_at": "2025-01-14T20:33:26.306744+00:00"} {"global_step": 5357, "acc_step": 0, "speed/wps": 12976.675532064239, "speed/FLOPS": 203816521051891.2, "speed/curr_iter_time": 1.2723, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5964637994766235, "optim/lr": 0.0029996898868186036, "optim/total_tokens": 1404305408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 6.124176025390625, "created_at": "2025-01-14T20:33:31.369211+00:00"} {"global_step": 5358, "acc_step": 0, "speed/wps": 12971.8433801311, "speed/FLOPS": 203740625465708.7, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.323496103286743, "optim/lr": 0.002999688142223253, "optim/total_tokens": 1404567552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 6.1422576904296875, "created_at": "2025-01-14T20:33:36.425297+00:00"} {"global_step": 5359, "acc_step": 0, "speed/wps": 12969.498330181663, "speed/FLOPS": 203703793233816.44, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.9652043581008911, "optim/lr": 0.0029996863927349063, "optim/total_tokens": 1404829696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483617, "loss/out": 6.230663299560547, "created_at": "2025-01-14T20:33:41.482993+00:00"} {"global_step": 5360, "acc_step": 0, "speed/wps": 12976.020757106056, "speed/FLOPS": 203806236911417.88, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9306621551513672, "optim/lr": 0.0029996846383535686, "optim/total_tokens": 1405091840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 6.238908767700195, "created_at": "2025-01-14T20:33:46.538690+00:00"} {"global_step": 5361, "acc_step": 0, "speed/wps": 12969.408612750705, "speed/FLOPS": 203702384098278.6, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.6701533794403076, "optim/lr": 0.0029996828790792465, "optim/total_tokens": 1405353984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462417, "loss/out": 6.145100116729736, "created_at": "2025-01-14T20:33:51.600599+00:00"} {"global_step": 5362, "acc_step": 0, "speed/wps": 12965.437480010332, "speed/FLOPS": 203640012001682.8, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.509448766708374, "optim/lr": 0.002999681114911945, "optim/total_tokens": 1405616128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460287, "loss/out": 6.300179958343506, "created_at": "2025-01-14T20:33:56.668633+00:00"} {"global_step": 5363, "acc_step": 0, "speed/wps": 12972.982933509997, "speed/FLOPS": 203758523717434.7, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.396243095397949, "optim/lr": 0.0029996793458516705, "optim/total_tokens": 1405878272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313070, "loss/out": 6.2774248123168945, "created_at": "2025-01-14T20:34:01.725464+00:00"} {"global_step": 5364, "acc_step": 0, "speed/wps": 12975.62791475515, "speed/FLOPS": 203800066782474.94, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.307858467102051, "optim/lr": 0.002999677571898428, "optim/total_tokens": 1406140416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420077, "loss/out": 6.208935737609863, "created_at": "2025-01-14T20:34:06.785812+00:00"} {"global_step": 5365, "acc_step": 0, "speed/wps": 12973.754919607527, "speed/FLOPS": 203770648819916.47, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 27.632600784301758, "optim/lr": 0.002999675793052224, "optim/total_tokens": 1406402560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449411, "loss/out": 6.116093635559082, "created_at": "2025-01-14T20:34:11.848949+00:00"} {"global_step": 5366, "acc_step": 0, "speed/wps": 12966.059005910183, "speed/FLOPS": 203649773920006.03, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.406665802001953, "optim/lr": 0.0029996740093130633, "optim/total_tokens": 1406664704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457124, "loss/out": 6.2758097648620605, "created_at": "2025-01-14T20:34:16.907161+00:00"} {"global_step": 5367, "acc_step": 0, "speed/wps": 12970.565504167715, "speed/FLOPS": 203720554667718.12, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.747405052185059, "optim/lr": 0.002999672220680953, "optim/total_tokens": 1406926848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447839, "loss/out": 6.096062660217285, "created_at": "2025-01-14T20:34:21.968856+00:00"} {"global_step": 5368, "acc_step": 0, "speed/wps": 12976.74854125037, "speed/FLOPS": 203817667761484.53, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9106755256652832, "optim/lr": 0.0029996704271558977, "optim/total_tokens": 1407188992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 6.256641387939453, "created_at": "2025-01-14T20:34:27.025154+00:00"} {"global_step": 5369, "acc_step": 0, "speed/wps": 12967.797667908853, "speed/FLOPS": 203677081995867.12, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3724305629730225, "optim/lr": 0.0029996686287379047, "optim/total_tokens": 1407451136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433718, "loss/out": 6.186466217041016, "created_at": "2025-01-14T20:34:32.082520+00:00"} {"global_step": 5370, "acc_step": 0, "speed/wps": 12972.290046103062, "speed/FLOPS": 203747640968581.72, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.6137115955352783, "optim/lr": 0.0029996668254269783, "optim/total_tokens": 1407713280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306165, "loss/out": 6.178683280944824, "created_at": "2025-01-14T20:34:37.139218+00:00"} {"global_step": 5371, "acc_step": 0, "speed/wps": 12966.630601837236, "speed/FLOPS": 203658751619495.53, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.60292387008667, "optim/lr": 0.0029996650172231256, "optim/total_tokens": 1407975424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401934, "loss/out": 6.064486026763916, "created_at": "2025-01-14T20:34:42.197090+00:00"} {"global_step": 5372, "acc_step": 0, "speed/wps": 12960.64704984813, "speed/FLOPS": 203564771713244.56, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.286505103111267, "optim/lr": 0.0029996632041263517, "optim/total_tokens": 1408237568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444641, "loss/out": 6.106838226318359, "created_at": "2025-01-14T20:34:47.260149+00:00"} {"global_step": 5373, "acc_step": 0, "speed/wps": 12965.270924575898, "speed/FLOPS": 203637396019713.84, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.500246047973633, "optim/lr": 0.002999661386136663, "optim/total_tokens": 1408499712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 6.220414161682129, "created_at": "2025-01-14T20:34:52.322926+00:00"} {"global_step": 5374, "acc_step": 0, "speed/wps": 12970.61198092708, "speed/FLOPS": 203721284649089.72, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.1682329177856445, "optim/lr": 0.0029996595632540654, "optim/total_tokens": 1408761856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456330, "loss/out": 6.31499719619751, "created_at": "2025-01-14T20:34:57.389831+00:00"} {"global_step": 5375, "acc_step": 0, "speed/wps": 12971.128827806333, "speed/FLOPS": 203729402439550.4, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9290894269943237, "optim/lr": 0.002999657735478565, "optim/total_tokens": 1409024000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 6.19719123840332, "created_at": "2025-01-14T20:35:02.454060+00:00"} {"global_step": 5376, "acc_step": 0, "speed/wps": 12970.590865518576, "speed/FLOPS": 203720953002583.53, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.578848123550415, "optim/lr": 0.0029996559028101664, "optim/total_tokens": 1409286144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 6.30571174621582, "created_at": "2025-01-14T20:35:07.512184+00:00"} {"global_step": 5377, "acc_step": 0, "speed/wps": 12971.929379340856, "speed/FLOPS": 203741976201473.2, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7417030334472656, "optim/lr": 0.0029996540652488775, "optim/total_tokens": 1409548288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448605, "loss/out": 6.167779445648193, "created_at": "2025-01-14T20:35:12.569204+00:00"} {"global_step": 5378, "acc_step": 0, "speed/wps": 12968.461538873107, "speed/FLOPS": 203687509001615.03, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.8315768241882324, "optim/lr": 0.0029996522227947035, "optim/total_tokens": 1409810432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319090, "loss/out": 6.065136909484863, "created_at": "2025-01-14T20:35:17.630164+00:00"} {"global_step": 5379, "acc_step": 0, "speed/wps": 12966.740380114077, "speed/FLOPS": 203660475838184.47, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.287568092346191, "optim/lr": 0.00299965037544765, "optim/total_tokens": 1410072576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361998, "loss/out": 6.113170623779297, "created_at": "2025-01-14T20:35:22.692554+00:00"} {"global_step": 5380, "acc_step": 0, "speed/wps": 12975.765028707845, "speed/FLOPS": 203802220345516.3, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.305983781814575, "optim/lr": 0.002999648523207724, "optim/total_tokens": 1410334720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446505, "loss/out": 6.222799301147461, "created_at": "2025-01-14T20:35:27.752774+00:00"} {"global_step": 5381, "acc_step": 0, "speed/wps": 12964.813749251234, "speed/FLOPS": 203630215453015.28, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0012, "optim/grad_norm": 4.41326379776001, "optim/lr": 0.00299964666607493, "optim/total_tokens": 1410596864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457095, "loss/out": 6.117324352264404, "created_at": "2025-01-14T20:35:32.811411+00:00"} {"global_step": 5382, "acc_step": 0, "speed/wps": 12964.56313679641, "speed/FLOPS": 203626279240034.7, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.941686749458313, "optim/lr": 0.0029996448040492756, "optim/total_tokens": 1410859008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 6.193451881408691, "created_at": "2025-01-14T20:35:37.873518+00:00"} {"global_step": 5383, "acc_step": 0, "speed/wps": 12968.694044211385, "speed/FLOPS": 203691160817448.47, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9443451166152954, "optim/lr": 0.0029996429371307665, "optim/total_tokens": 1411121152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445933, "loss/out": 6.177738189697266, "created_at": "2025-01-14T20:35:42.936078+00:00"} {"global_step": 5384, "acc_step": 0, "speed/wps": 12970.03666561615, "speed/FLOPS": 203712248531565.47, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 26.527917861938477, "optim/lr": 0.002999641065319408, "optim/total_tokens": 1411383296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 6.076752662658691, "created_at": "2025-01-14T20:35:48.002404+00:00"} {"global_step": 5385, "acc_step": 0, "speed/wps": 12973.75784974344, "speed/FLOPS": 203770694841727.4, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.992944717407227, "optim/lr": 0.002999639188615207, "optim/total_tokens": 1411645440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457902, "loss/out": 6.0951313972473145, "created_at": "2025-01-14T20:35:53.060051+00:00"} {"global_step": 5386, "acc_step": 0, "speed/wps": 12969.422957225548, "speed/FLOPS": 203702609397967.38, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.128193378448486, "optim/lr": 0.0029996373070181694, "optim/total_tokens": 1411907584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308260, "loss/out": 6.187819480895996, "created_at": "2025-01-14T20:35:58.119832+00:00"} {"global_step": 5387, "acc_step": 0, "speed/wps": 12969.348056039002, "speed/FLOPS": 203701432971907.62, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.250981569290161, "optim/lr": 0.002999635420528301, "optim/total_tokens": 1412169728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345973, "loss/out": 6.142277717590332, "created_at": "2025-01-14T20:36:03.177717+00:00"} {"global_step": 5388, "acc_step": 0, "speed/wps": 12969.160650295973, "speed/FLOPS": 203698489507190.9, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0009, "optim/grad_norm": 3.9372825622558594, "optim/lr": 0.002999633529145609, "optim/total_tokens": 1412431872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330091, "loss/out": 6.072803974151611, "created_at": "2025-01-14T20:36:08.239417+00:00"} {"global_step": 5389, "acc_step": 0, "speed/wps": 12964.02826018472, "speed/FLOPS": 203617878267850.6, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4423828125, "optim/lr": 0.0029996316328700983, "optim/total_tokens": 1412694016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456045, "loss/out": 6.176753044128418, "created_at": "2025-01-14T20:36:13.298406+00:00"} {"global_step": 5390, "acc_step": 0, "speed/wps": 12965.975992028732, "speed/FLOPS": 203648470072924.84, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.0587193965911865, "optim/lr": 0.002999629731701776, "optim/total_tokens": 1412956160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441679, "loss/out": 6.314352035522461, "created_at": "2025-01-14T20:36:18.361048+00:00"} {"global_step": 5391, "acc_step": 0, "speed/wps": 12964.9235755932, "speed/FLOPS": 203631940426633.06, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8113583326339722, "optim/lr": 0.002999627825640648, "optim/total_tokens": 1413218304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 6.072600841522217, "created_at": "2025-01-14T20:36:23.422817+00:00"} {"global_step": 5392, "acc_step": 0, "speed/wps": 12971.701179290327, "speed/FLOPS": 203738392006099.9, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2215304374694824, "optim/lr": 0.00299962591468672, "optim/total_tokens": 1413480448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436388, "loss/out": 6.051831245422363, "created_at": "2025-01-14T20:36:28.484466+00:00"} {"global_step": 5393, "acc_step": 0, "speed/wps": 12969.926702429872, "speed/FLOPS": 203710521408619.72, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.420848846435547, "optim/lr": 0.002999623998839999, "optim/total_tokens": 1413742592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 6.015170574188232, "created_at": "2025-01-14T20:36:33.552340+00:00"} {"global_step": 5394, "acc_step": 0, "speed/wps": 12969.6825520476, "speed/FLOPS": 203706686691368.34, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.1855785846710205, "optim/lr": 0.0029996220781004913, "optim/total_tokens": 1414004736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447839, "loss/out": 6.10675048828125, "created_at": "2025-01-14T20:36:38.611768+00:00"} {"global_step": 5395, "acc_step": 0, "speed/wps": 12970.759582708299, "speed/FLOPS": 203723602937889.84, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1596245765686035, "optim/lr": 0.002999620152468202, "optim/total_tokens": 1414266880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289311, "loss/out": 6.124269485473633, "created_at": "2025-01-14T20:36:43.672941+00:00"} {"global_step": 5396, "acc_step": 0, "speed/wps": 12962.808617012164, "speed/FLOPS": 203598722095860.25, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9937583208084106, "optim/lr": 0.0029996182219431387, "optim/total_tokens": 1414529024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466392, "loss/out": 6.099587917327881, "created_at": "2025-01-14T20:36:48.732208+00:00"} {"global_step": 5397, "acc_step": 0, "speed/wps": 12969.072787567566, "speed/FLOPS": 203697109502305.22, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8619576692581177, "optim/lr": 0.002999616286525307, "optim/total_tokens": 1414791168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464007, "loss/out": 6.085070610046387, "created_at": "2025-01-14T20:36:53.790358+00:00"} {"global_step": 5398, "acc_step": 0, "speed/wps": 12964.290106367025, "speed/FLOPS": 203621990921958.06, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8296396732330322, "optim/lr": 0.002999614346214714, "optim/total_tokens": 1415053312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463770, "loss/out": 6.168357849121094, "created_at": "2025-01-14T20:36:58.853531+00:00"} {"global_step": 5399, "acc_step": 0, "speed/wps": 12967.570005108062, "speed/FLOPS": 203673506238739.22, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.251798391342163, "optim/lr": 0.0029996124010113648, "optim/total_tokens": 1415315456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464007, "loss/out": 6.150997161865234, "created_at": "2025-01-14T20:37:03.915719+00:00"} {"global_step": 5400, "acc_step": 0, "speed/wps": 12964.93916899213, "speed/FLOPS": 203632185342390.84, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6617884635925293, "optim/lr": 0.0029996104509152667, "optim/total_tokens": 1415577600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 6.194088459014893, "created_at": "2025-01-14T20:37:08.974182+00:00"} {"global_step": 5401, "acc_step": 0, "speed/wps": 12974.637136877172, "speed/FLOPS": 203784505254429.8, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 19.2296085357666, "optim/lr": 0.0029996084959264258, "optim/total_tokens": 1415839744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 6.056037902832031, "created_at": "2025-01-14T20:37:14.047912+00:00"} {"global_step": 5402, "acc_step": 0, "speed/wps": 12973.068196838283, "speed/FLOPS": 203759862895169.5, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.351529836654663, "optim/lr": 0.002999606536044848, "optim/total_tokens": 1416101888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307679, "loss/out": 6.237861633300781, "created_at": "2025-01-14T20:37:19.103405+00:00"} {"global_step": 5403, "acc_step": 0, "speed/wps": 12969.513090781644, "speed/FLOPS": 203704025069322.34, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.3491244316101074, "optim/lr": 0.0029996045712705408, "optim/total_tokens": 1416364032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353744, "loss/out": 6.148006916046143, "created_at": "2025-01-14T20:37:24.161214+00:00"} {"global_step": 5404, "acc_step": 0, "speed/wps": 12972.527861542767, "speed/FLOPS": 203751376186854.94, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.6343188285827637, "optim/lr": 0.0029996026016035093, "optim/total_tokens": 1416626176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446505, "loss/out": 6.090181350708008, "created_at": "2025-01-14T20:37:29.216828+00:00"} {"global_step": 5405, "acc_step": 0, "speed/wps": 12971.785347330311, "speed/FLOPS": 203739713980826.84, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.2165679931640625, "optim/lr": 0.002999600627043761, "optim/total_tokens": 1416888320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 6.144410133361816, "created_at": "2025-01-14T20:37:34.278410+00:00"} {"global_step": 5406, "acc_step": 0, "speed/wps": 12973.062279790194, "speed/FLOPS": 203759769959798.4, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.725587248802185, "optim/lr": 0.002999598647591302, "optim/total_tokens": 1417150464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450462, "loss/out": 6.061907768249512, "created_at": "2025-01-14T20:37:39.334322+00:00"} {"global_step": 5407, "acc_step": 0, "speed/wps": 12971.343555612695, "speed/FLOPS": 203732775034816.25, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4344252347946167, "optim/lr": 0.002999596663246138, "optim/total_tokens": 1417412608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 6.034617900848389, "created_at": "2025-01-14T20:37:44.393106+00:00"} {"global_step": 5408, "acc_step": 0, "speed/wps": 12965.909427833758, "speed/FLOPS": 203647424590773.97, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.948267936706543, "optim/lr": 0.0029995946740082767, "optim/total_tokens": 1417674752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 6.094027996063232, "created_at": "2025-01-14T20:37:49.452022+00:00"} {"global_step": 5409, "acc_step": 0, "speed/wps": 12974.514790040386, "speed/FLOPS": 203782583629235.9, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.28874933719635, "optim/lr": 0.002999592679877724, "optim/total_tokens": 1417936896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445170, "loss/out": 6.186431407928467, "created_at": "2025-01-14T20:37:54.510772+00:00"} {"global_step": 5410, "acc_step": 0, "speed/wps": 12979.981427634451, "speed/FLOPS": 203868444684597.97, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3097094297409058, "optim/lr": 0.0029995906808544864, "optim/total_tokens": 1418199040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328746, "loss/out": 6.07049560546875, "created_at": "2025-01-14T20:37:59.567736+00:00"} {"global_step": 5411, "acc_step": 0, "speed/wps": 12967.466325451163, "speed/FLOPS": 203671877807256.84, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.3255079984664917, "optim/lr": 0.00299958867693857, "optim/total_tokens": 1418461184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360416, "loss/out": 6.1373162269592285, "created_at": "2025-01-14T20:38:04.627694+00:00"} {"global_step": 5412, "acc_step": 0, "speed/wps": 12963.167802442593, "speed/FLOPS": 203604363596617.62, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0009, "optim/grad_norm": 2.5882017612457275, "optim/lr": 0.0029995866681299826, "optim/total_tokens": 1418723328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441971, "loss/out": 6.018697738647461, "created_at": "2025-01-14T20:38:09.689462+00:00"} {"global_step": 5413, "acc_step": 0, "speed/wps": 12967.639721630436, "speed/FLOPS": 203674601232523.62, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.9602261781692505, "optim/lr": 0.0029995846544287295, "optim/total_tokens": 1418985472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 6.110045433044434, "created_at": "2025-01-14T20:38:14.747351+00:00"} {"global_step": 5414, "acc_step": 0, "speed/wps": 12964.0880507127, "speed/FLOPS": 203618817360253.3, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.702322006225586, "optim/lr": 0.0029995826358348177, "optim/total_tokens": 1419247616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 6.0796308517456055, "created_at": "2025-01-14T20:38:19.806362+00:00"} {"global_step": 5415, "acc_step": 0, "speed/wps": 12968.557258906647, "speed/FLOPS": 203689012416274.88, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7636851072311401, "optim/lr": 0.002999580612348254, "optim/total_tokens": 1419509760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483617, "loss/out": 6.0795087814331055, "created_at": "2025-01-14T20:38:24.863358+00:00"} {"global_step": 5416, "acc_step": 0, "speed/wps": 12970.338129073742, "speed/FLOPS": 203716983429421.16, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.378820776939392, "optim/lr": 0.0029995785839690447, "optim/total_tokens": 1419771904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451278, "loss/out": 6.15009069442749, "created_at": "2025-01-14T20:38:29.925325+00:00"} {"global_step": 5417, "acc_step": 0, "speed/wps": 12970.436346428593, "speed/FLOPS": 203718526067943.88, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1977555751800537, "optim/lr": 0.002999576550697196, "optim/total_tokens": 1420034048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 6.095255374908447, "created_at": "2025-01-14T20:38:34.981900+00:00"} {"global_step": 5418, "acc_step": 0, "speed/wps": 12979.507734241928, "speed/FLOPS": 203861004678945.44, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.6672327518463135, "optim/lr": 0.0029995745125327164, "optim/total_tokens": 1420296192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457902, "loss/out": 6.10561990737915, "created_at": "2025-01-14T20:38:40.035480+00:00"} {"global_step": 5419, "acc_step": 0, "speed/wps": 12971.545710767707, "speed/FLOPS": 203735950159316.0, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.4157257080078125, "optim/lr": 0.00299957246947561, "optim/total_tokens": 1420558336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436381, "loss/out": 6.266422271728516, "created_at": "2025-01-14T20:38:45.091468+00:00"} {"global_step": 5420, "acc_step": 0, "speed/wps": 12959.271994794042, "speed/FLOPS": 203543174584096.44, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.327458381652832, "optim/lr": 0.0029995704215258854, "optim/total_tokens": 1420820480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 385871, "loss/out": 6.164161682128906, "created_at": "2025-01-14T20:38:50.156135+00:00"} {"global_step": 5421, "acc_step": 0, "speed/wps": 12971.273849471376, "speed/FLOPS": 203731680204080.56, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.816856861114502, "optim/lr": 0.002999568368683548, "optim/total_tokens": 1421082624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434531, "loss/out": 5.979621887207031, "created_at": "2025-01-14T20:38:55.216985+00:00"} {"global_step": 5422, "acc_step": 0, "speed/wps": 12972.869261179356, "speed/FLOPS": 203756738337279.8, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9838993549346924, "optim/lr": 0.0029995663109486054, "optim/total_tokens": 1421344768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 6.122470378875732, "created_at": "2025-01-14T20:39:00.272763+00:00"} {"global_step": 5423, "acc_step": 0, "speed/wps": 12966.68737951772, "speed/FLOPS": 203659643390988.22, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 35.291297912597656, "optim/lr": 0.002999564248321064, "optim/total_tokens": 1421606912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464317, "loss/out": 6.1152143478393555, "created_at": "2025-01-14T20:39:05.335973+00:00"} {"global_step": 5424, "acc_step": 0, "speed/wps": 12966.968453443156, "speed/FLOPS": 203664058043223.25, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.206415891647339, "optim/lr": 0.0029995621808009305, "optim/total_tokens": 1421869056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 6.082347869873047, "created_at": "2025-01-14T20:39:10.395927+00:00"} {"global_step": 5425, "acc_step": 0, "speed/wps": 12967.850089698572, "speed/FLOPS": 203677905352109.72, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.5550897121429443, "optim/lr": 0.0029995601083882112, "optim/total_tokens": 1422131200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462417, "loss/out": 6.055595397949219, "created_at": "2025-01-14T20:39:15.453843+00:00"} {"global_step": 5426, "acc_step": 0, "speed/wps": 12973.964012645054, "speed/FLOPS": 203773932913394.75, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6080477237701416, "optim/lr": 0.002999558031082914, "optim/total_tokens": 1422393344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440628, "loss/out": 6.107504844665527, "created_at": "2025-01-14T20:39:20.514280+00:00"} {"global_step": 5427, "acc_step": 0, "speed/wps": 12974.470012483485, "speed/FLOPS": 203781880336173.97, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.9011430740356445, "optim/lr": 0.0029995559488850444, "optim/total_tokens": 1422655488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331967, "loss/out": 6.018750190734863, "created_at": "2025-01-14T20:39:25.570558+00:00"} {"global_step": 5428, "acc_step": 0, "speed/wps": 12972.628553817402, "speed/FLOPS": 203752957697393.75, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.819960594177246, "optim/lr": 0.0029995538617946104, "optim/total_tokens": 1422917632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334076, "loss/out": 6.1021528244018555, "created_at": "2025-01-14T20:39:30.631037+00:00"} {"global_step": 5429, "acc_step": 0, "speed/wps": 12972.840475697909, "speed/FLOPS": 203756286221738.3, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 16.979476928710938, "optim/lr": 0.0029995517698116177, "optim/total_tokens": 1423179776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294127, "loss/out": 6.145725250244141, "created_at": "2025-01-14T20:39:35.687571+00:00"} {"global_step": 5430, "acc_step": 0, "speed/wps": 12972.565601830505, "speed/FLOPS": 203751968949934.44, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.7666473388671875, "optim/lr": 0.002999549672936073, "optim/total_tokens": 1423441920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 394163, "loss/out": 6.140586853027344, "created_at": "2025-01-14T20:39:40.746928+00:00"} {"global_step": 5431, "acc_step": 0, "speed/wps": 12975.42733106295, "speed/FLOPS": 203796916339958.06, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0009, "optim/grad_norm": 3.1161646842956543, "optim/lr": 0.002999547571167985, "optim/total_tokens": 1423704064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443545, "loss/out": 6.105374336242676, "created_at": "2025-01-14T20:39:45.806125+00:00"} {"global_step": 5432, "acc_step": 0, "speed/wps": 12974.269390432022, "speed/FLOPS": 203778729291172.16, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.399460315704346, "optim/lr": 0.0029995454645073585, "optim/total_tokens": 1423966208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 6.074070453643799, "created_at": "2025-01-14T20:39:50.860991+00:00"} {"global_step": 5433, "acc_step": 0, "speed/wps": 12967.759931460718, "speed/FLOPS": 203676489293093.84, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8687111139297485, "optim/lr": 0.0029995433529542014, "optim/total_tokens": 1424228352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437717, "loss/out": 6.101322174072266, "created_at": "2025-01-14T20:39:55.918250+00:00"} {"global_step": 5434, "acc_step": 0, "speed/wps": 12974.190904896677, "speed/FLOPS": 203777496567989.12, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.3046274185180664, "optim/lr": 0.0029995412365085208, "optim/total_tokens": 1424490496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 6.127546310424805, "created_at": "2025-01-14T20:40:00.978225+00:00"} {"global_step": 5435, "acc_step": 0, "speed/wps": 12959.605897260222, "speed/FLOPS": 203548418981157.88, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.748331546783447, "optim/lr": 0.0029995391151703228, "optim/total_tokens": 1424752640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443307, "loss/out": 5.9617204666137695, "created_at": "2025-01-14T20:40:06.038737+00:00"} {"global_step": 5436, "acc_step": 0, "speed/wps": 12971.355926958948, "speed/FLOPS": 203732969343809.16, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.253595352172852, "optim/lr": 0.002999536988939615, "optim/total_tokens": 1425014784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318692, "loss/out": 6.042926788330078, "created_at": "2025-01-14T20:40:11.094708+00:00"} {"global_step": 5437, "acc_step": 0, "speed/wps": 12973.285737007634, "speed/FLOPS": 203763279662460.7, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 20.177518844604492, "optim/lr": 0.0029995348578164033, "optim/total_tokens": 1425276928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363872, "loss/out": 6.163358688354492, "created_at": "2025-01-14T20:40:16.155577+00:00"} {"global_step": 5438, "acc_step": 0, "speed/wps": 12971.891102330097, "speed/FLOPS": 203741375008421.1, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.719707489013672, "optim/lr": 0.0029995327218006957, "optim/total_tokens": 1425539072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430791, "loss/out": 6.080827713012695, "created_at": "2025-01-14T20:40:21.216666+00:00"} {"global_step": 5439, "acc_step": 0, "speed/wps": 12976.15721176083, "speed/FLOPS": 203808380119279.44, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.959634304046631, "optim/lr": 0.0029995305808924993, "optim/total_tokens": 1425801216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 6.196746826171875, "created_at": "2025-01-14T20:40:26.275993+00:00"} {"global_step": 5440, "acc_step": 0, "speed/wps": 12969.86583011322, "speed/FLOPS": 203709565325239.38, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.133575916290283, "optim/lr": 0.0029995284350918206, "optim/total_tokens": 1426063360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 6.138369083404541, "created_at": "2025-01-14T20:40:31.335738+00:00"} {"global_step": 5441, "acc_step": 0, "speed/wps": 12969.639170383221, "speed/FLOPS": 203706005322716.62, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7162054777145386, "optim/lr": 0.002999526284398666, "optim/total_tokens": 1426325504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440921, "loss/out": 6.023155212402344, "created_at": "2025-01-14T20:40:36.398951+00:00"} {"global_step": 5442, "acc_step": 0, "speed/wps": 12973.96988267521, "speed/FLOPS": 203774025110284.62, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 74.37645721435547, "optim/lr": 0.0029995241288130435, "optim/total_tokens": 1426587648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456567, "loss/out": 6.066810607910156, "created_at": "2025-01-14T20:40:41.453912+00:00"} {"global_step": 5443, "acc_step": 0, "speed/wps": 12977.318064552188, "speed/FLOPS": 203826612907546.6, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0008, "optim/grad_norm": 11.615015029907227, "optim/lr": 0.00299952196833496, "optim/total_tokens": 1426849792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463248, "loss/out": 6.061042785644531, "created_at": "2025-01-14T20:40:46.513456+00:00"} {"global_step": 5444, "acc_step": 0, "speed/wps": 12976.289187416527, "speed/FLOPS": 203810452978305.72, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.7594075202941895, "optim/lr": 0.0029995198029644224, "optim/total_tokens": 1427111936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 6.091960906982422, "created_at": "2025-01-14T20:40:51.570904+00:00"} {"global_step": 5445, "acc_step": 0, "speed/wps": 12977.805836694646, "speed/FLOPS": 203834274039313.06, "speed/curr_iter_time": 1.2722, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.465989589691162, "optim/lr": 0.0029995176327014374, "optim/total_tokens": 1427374080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430567, "loss/out": 6.127617835998535, "created_at": "2025-01-14T20:40:56.625405+00:00"} {"global_step": 5446, "acc_step": 0, "speed/wps": 12966.151374347668, "speed/FLOPS": 203651224693244.28, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0009, "optim/grad_norm": 3.6651463508605957, "optim/lr": 0.0029995154575460127, "optim/total_tokens": 1427636224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293783, "loss/out": 6.0983476638793945, "created_at": "2025-01-14T20:41:01.693388+00:00"} {"global_step": 5447, "acc_step": 0, "speed/wps": 12968.778223423998, "speed/FLOPS": 203692482967656.94, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0011, "optim/grad_norm": 3.0999064445495605, "optim/lr": 0.0029995132774981548, "optim/total_tokens": 1427898368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438536, "loss/out": 6.133925437927246, "created_at": "2025-01-14T20:41:06.754852+00:00"} {"global_step": 5448, "acc_step": 0, "speed/wps": 12972.584047562721, "speed/FLOPS": 203752258665506.0, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.6107640266418457, "optim/lr": 0.0029995110925578714, "optim/total_tokens": 1428160512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 5.892131328582764, "created_at": "2025-01-14T20:41:11.810528+00:00"} {"global_step": 5449, "acc_step": 0, "speed/wps": 12962.558242053772, "speed/FLOPS": 203594789613087.47, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.0557756423950195, "optim/lr": 0.002999508902725169, "optim/total_tokens": 1428422656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443543, "loss/out": 6.098776817321777, "created_at": "2025-01-14T20:41:16.872315+00:00"} {"global_step": 5450, "acc_step": 0, "speed/wps": 12971.604047492072, "speed/FLOPS": 203736866417735.22, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.001, "optim/grad_norm": 2.082094669342041, "optim/lr": 0.002999506708000055, "optim/total_tokens": 1428684800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457095, "loss/out": 5.984328269958496, "created_at": "2025-01-14T20:41:21.932839+00:00"} {"global_step": 5451, "acc_step": 0, "speed/wps": 12972.248753809432, "speed/FLOPS": 203746992416368.7, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 32.40046691894531, "optim/lr": 0.002999504508382537, "optim/total_tokens": 1428946944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 5.964804649353027, "created_at": "2025-01-14T20:41:26.990169+00:00"} {"global_step": 5452, "acc_step": 0, "speed/wps": 12976.469153760256, "speed/FLOPS": 203813279597033.66, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.3987081050872803, "optim/lr": 0.0029995023038726218, "optim/total_tokens": 1429209088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439583, "loss/out": 6.101479530334473, "created_at": "2025-01-14T20:41:32.044016+00:00"} {"global_step": 5453, "acc_step": 0, "speed/wps": 12969.176155888128, "speed/FLOPS": 203698733043822.6, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.963566780090332, "optim/lr": 0.0029995000944703165, "optim/total_tokens": 1429471232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297001, "loss/out": 6.02018928527832, "created_at": "2025-01-14T20:41:37.100789+00:00"} {"global_step": 5454, "acc_step": 0, "speed/wps": 12974.392168017834, "speed/FLOPS": 203780657681871.4, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7171714305877686, "optim/lr": 0.002999497880175629, "optim/total_tokens": 1429733376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297001, "loss/out": 6.063426494598389, "created_at": "2025-01-14T20:41:42.156698+00:00"} {"global_step": 5455, "acc_step": 0, "speed/wps": 12976.651922530418, "speed/FLOPS": 203816150231719.66, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5919249057769775, "optim/lr": 0.002999495660988565, "optim/total_tokens": 1429995520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317346, "loss/out": 5.985323905944824, "created_at": "2025-01-14T20:41:47.217253+00:00"} {"global_step": 5456, "acc_step": 0, "speed/wps": 12972.55553837004, "speed/FLOPS": 203751810889459.2, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.800553560256958, "optim/lr": 0.0029994934369091328, "optim/total_tokens": 1430257664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453131, "loss/out": 6.010141372680664, "created_at": "2025-01-14T20:41:52.280718+00:00"} {"global_step": 5457, "acc_step": 0, "speed/wps": 12966.719095718297, "speed/FLOPS": 203660141537500.94, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.69362211227417, "optim/lr": 0.00299949120793734, "optim/total_tokens": 1430519808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 5.948729515075684, "created_at": "2025-01-14T20:41:57.341270+00:00"} {"global_step": 5458, "acc_step": 0, "speed/wps": 12969.540142805698, "speed/FLOPS": 203704449958533.47, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.59276008605957, "optim/lr": 0.002999488974073193, "optim/total_tokens": 1430781952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445404, "loss/out": 6.191162586212158, "created_at": "2025-01-14T20:42:02.402088+00:00"} {"global_step": 5459, "acc_step": 0, "speed/wps": 12968.528540358198, "speed/FLOPS": 203688561352008.06, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.440171957015991, "optim/lr": 0.0029994867353167, "optim/total_tokens": 1431044096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467205, "loss/out": 6.008536338806152, "created_at": "2025-01-14T20:42:07.466682+00:00"} {"global_step": 5460, "acc_step": 0, "speed/wps": 12967.637766862035, "speed/FLOPS": 203674570530199.78, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.643472671508789, "optim/lr": 0.0029994844916678676, "optim/total_tokens": 1431306240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 6.078401565551758, "created_at": "2025-01-14T20:42:12.529317+00:00"} {"global_step": 5461, "acc_step": 0, "speed/wps": 12964.854093698026, "speed/FLOPS": 203630849117991.22, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.305746555328369, "optim/lr": 0.002999482243126703, "optim/total_tokens": 1431568384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299119, "loss/out": 6.03767204284668, "created_at": "2025-01-14T20:42:17.588699+00:00"} {"global_step": 5462, "acc_step": 0, "speed/wps": 12968.942321057842, "speed/FLOPS": 203695060346490.25, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.0181145668029785, "optim/lr": 0.002999479989693214, "optim/total_tokens": 1431830528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366652, "loss/out": 6.108671188354492, "created_at": "2025-01-14T20:42:22.646675+00:00"} {"global_step": 5463, "acc_step": 0, "speed/wps": 12971.945764371336, "speed/FLOPS": 203742233550892.25, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5944840908050537, "optim/lr": 0.002999477731367408, "optim/total_tokens": 1432092672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441971, "loss/out": 6.030335426330566, "created_at": "2025-01-14T20:42:27.706977+00:00"} {"global_step": 5464, "acc_step": 0, "speed/wps": 12972.944262834404, "speed/FLOPS": 203757916341335.44, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.978720188140869, "optim/lr": 0.002999475468149292, "optim/total_tokens": 1432354816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 5.956092834472656, "created_at": "2025-01-14T20:42:32.762315+00:00"} {"global_step": 5465, "acc_step": 0, "speed/wps": 12966.879800846666, "speed/FLOPS": 203662665632373.94, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7643132209777832, "optim/lr": 0.002999473200038874, "optim/total_tokens": 1432616960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447791, "loss/out": 6.088057518005371, "created_at": "2025-01-14T20:42:37.820545+00:00"} {"global_step": 5466, "acc_step": 0, "speed/wps": 12966.509435729444, "speed/FLOPS": 203656848539272.97, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.6040393114089966, "optim/lr": 0.0029994709270361603, "optim/total_tokens": 1432879104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473310, "loss/out": 6.110445976257324, "created_at": "2025-01-14T20:42:42.879307+00:00"} {"global_step": 5467, "acc_step": 0, "speed/wps": 12972.517879451547, "speed/FLOPS": 203751219404395.47, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.0247039794921875, "optim/lr": 0.0029994686491411587, "optim/total_tokens": 1433141248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 5.961045742034912, "created_at": "2025-01-14T20:42:47.937782+00:00"} {"global_step": 5468, "acc_step": 0, "speed/wps": 12971.427128412803, "speed/FLOPS": 203734087660483.0, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1107516288757324, "optim/lr": 0.0029994663663538775, "optim/total_tokens": 1433403392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 6.063053131103516, "created_at": "2025-01-14T20:42:52.998560+00:00"} {"global_step": 5469, "acc_step": 0, "speed/wps": 12964.954953524837, "speed/FLOPS": 203632433260167.22, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.6642141342163086, "optim/lr": 0.0029994640786743243, "optim/total_tokens": 1433665536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301104, "loss/out": 6.039973735809326, "created_at": "2025-01-14T20:42:58.059051+00:00"} {"global_step": 5470, "acc_step": 0, "speed/wps": 12975.232933376437, "speed/FLOPS": 203793863057160.94, "speed/curr_iter_time": 1.272, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.388821601867676, "optim/lr": 0.002999461786102505, "optim/total_tokens": 1433927680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298593, "loss/out": 6.119256019592285, "created_at": "2025-01-14T20:43:03.117043+00:00"} {"global_step": 5471, "acc_step": 0, "speed/wps": 12971.55787686178, "speed/FLOPS": 203736141244541.4, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.154301643371582, "optim/lr": 0.002999459488638428, "optim/total_tokens": 1434189824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434505, "loss/out": 6.098716735839844, "created_at": "2025-01-14T20:43:08.177933+00:00"} {"global_step": 5472, "acc_step": 0, "speed/wps": 12970.068608575153, "speed/FLOPS": 203712750239630.84, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9212826490402222, "optim/lr": 0.0029994571862821, "optim/total_tokens": 1434451968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 6.036881446838379, "created_at": "2025-01-14T20:43:13.244615+00:00"} {"global_step": 5473, "acc_step": 0, "speed/wps": 12967.154997738358, "speed/FLOPS": 203666987977718.12, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4007272720336914, "optim/lr": 0.0029994548790335303, "optim/total_tokens": 1434714112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430791, "loss/out": 5.97426700592041, "created_at": "2025-01-14T20:43:18.302300+00:00"} {"global_step": 5474, "acc_step": 0, "speed/wps": 12966.23215898596, "speed/FLOPS": 203652493526998.38, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.574718475341797, "optim/lr": 0.0029994525668927247, "optim/total_tokens": 1434976256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460815, "loss/out": 6.017463684082031, "created_at": "2025-01-14T20:43:23.361497+00:00"} {"global_step": 5475, "acc_step": 0, "speed/wps": 12968.194075192747, "speed/FLOPS": 203683308116981.34, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.177487850189209, "optim/lr": 0.002999450249859692, "optim/total_tokens": 1435238400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 6.036856651306152, "created_at": "2025-01-14T20:43:28.423793+00:00"} {"global_step": 5476, "acc_step": 0, "speed/wps": 12967.186460547287, "speed/FLOPS": 203667482144366.78, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.517271518707275, "optim/lr": 0.002999447927934438, "optim/total_tokens": 1435500544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326058, "loss/out": 6.061032772064209, "created_at": "2025-01-14T20:43:33.485300+00:00"} {"global_step": 5477, "acc_step": 0, "speed/wps": 12966.729995958058, "speed/FLOPS": 203660312740744.88, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0008, "optim/grad_norm": 9.39222240447998, "optim/lr": 0.0029994456011169725, "optim/total_tokens": 1435762688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359074, "loss/out": 6.037295341491699, "created_at": "2025-01-14T20:43:38.544410+00:00"} {"global_step": 5478, "acc_step": 0, "speed/wps": 12960.503347043878, "speed/FLOPS": 203562514663235.4, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.060709476470947, "optim/lr": 0.0029994432694073017, "optim/total_tokens": 1436024832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432656, "loss/out": 6.1504926681518555, "created_at": "2025-01-14T20:43:43.607325+00:00"} {"global_step": 5479, "acc_step": 0, "speed/wps": 12961.326523202035, "speed/FLOPS": 203575443775961.84, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.001, "optim/grad_norm": 2.97869873046875, "optim/lr": 0.0029994409328054335, "optim/total_tokens": 1436286976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 6.204799175262451, "created_at": "2025-01-14T20:43:48.671689+00:00"} {"global_step": 5480, "acc_step": 0, "speed/wps": 12968.197084073892, "speed/FLOPS": 203683355375594.38, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.1979684829711914, "optim/lr": 0.002999438591311376, "optim/total_tokens": 1436549120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446220, "loss/out": 6.15844202041626, "created_at": "2025-01-14T20:43:53.733600+00:00"} {"global_step": 5481, "acc_step": 0, "speed/wps": 12965.181241734099, "speed/FLOPS": 203635987427446.25, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.010907173156738, "optim/lr": 0.002999436244925135, "optim/total_tokens": 1436811264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436666, "loss/out": 6.089968204498291, "created_at": "2025-01-14T20:43:58.795504+00:00"} {"global_step": 5482, "acc_step": 0, "speed/wps": 12967.632292079837, "speed/FLOPS": 203674484541222.22, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 8.867664337158203, "optim/lr": 0.002999433893646721, "optim/total_tokens": 1437073408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 6.140950679779053, "created_at": "2025-01-14T20:44:03.853983+00:00"} {"global_step": 5483, "acc_step": 0, "speed/wps": 12964.849774990953, "speed/FLOPS": 203630781286762.12, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0009, "optim/grad_norm": 4.696540832519531, "optim/lr": 0.0029994315374761394, "optim/total_tokens": 1437335552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456045, "loss/out": 6.071770668029785, "created_at": "2025-01-14T20:44:08.917662+00:00"} {"global_step": 5484, "acc_step": 0, "speed/wps": 12955.278458766235, "speed/FLOPS": 203480450613086.16, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.487007141113281, "optim/lr": 0.0029994291764133987, "optim/total_tokens": 1437597696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287957, "loss/out": 6.091219902038574, "created_at": "2025-01-14T20:44:13.982136+00:00"} {"global_step": 5485, "acc_step": 0, "speed/wps": 12970.555811988366, "speed/FLOPS": 203720402438723.0, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.001, "optim/grad_norm": 1.435599684715271, "optim/lr": 0.0029994268104585075, "optim/total_tokens": 1437859840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366413, "loss/out": 6.131593704223633, "created_at": "2025-01-14T20:44:19.038372+00:00"} {"global_step": 5486, "acc_step": 0, "speed/wps": 12968.85916747602, "speed/FLOPS": 203693754305222.72, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.4010658264160156, "optim/lr": 0.002999424439611472, "optim/total_tokens": 1438121984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443016, "loss/out": 6.002575874328613, "created_at": "2025-01-14T20:44:24.101362+00:00"} {"global_step": 5487, "acc_step": 0, "speed/wps": 12963.844843749623, "speed/FLOPS": 203614997460696.78, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.1401171684265137, "optim/lr": 0.0029994220638723005, "optim/total_tokens": 1438384128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 6.148618221282959, "created_at": "2025-01-14T20:44:29.167969+00:00"} {"global_step": 5488, "acc_step": 0, "speed/wps": 12966.422970788975, "speed/FLOPS": 203655490488567.62, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.393032550811768, "optim/lr": 0.0029994196832410004, "optim/total_tokens": 1438646272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445933, "loss/out": 6.08089542388916, "created_at": "2025-01-14T20:44:34.226297+00:00"} {"global_step": 5489, "acc_step": 0, "speed/wps": 12968.32799923952, "speed/FLOPS": 203685411578166.84, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.1204094886779785, "optim/lr": 0.0029994172977175805, "optim/total_tokens": 1438908416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 6.102603912353516, "created_at": "2025-01-14T20:44:39.290774+00:00"} {"global_step": 5490, "acc_step": 0, "speed/wps": 12966.749583940447, "speed/FLOPS": 203660620396925.0, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.1949310302734375, "optim/lr": 0.0029994149073020475, "optim/total_tokens": 1439170560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443545, "loss/out": 6.107670783996582, "created_at": "2025-01-14T20:44:44.352683+00:00"} {"global_step": 5491, "acc_step": 0, "speed/wps": 12965.287326233467, "speed/FLOPS": 203637653630284.22, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.8592678308486938, "optim/lr": 0.00299941251199441, "optim/total_tokens": 1439432704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304284, "loss/out": 6.015871047973633, "created_at": "2025-01-14T20:44:49.410915+00:00"} {"global_step": 5492, "acc_step": 0, "speed/wps": 12967.275224867142, "speed/FLOPS": 203668876309985.06, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.752009153366089, "optim/lr": 0.002999410111794675, "optim/total_tokens": 1439694848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 381895, "loss/out": 6.2042765617370605, "created_at": "2025-01-14T20:44:54.472970+00:00"} {"global_step": 5493, "acc_step": 0, "speed/wps": 12974.096900521103, "speed/FLOPS": 203776020100095.2, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.6633450984954834, "optim/lr": 0.0029994077067028516, "optim/total_tokens": 1439956992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448083, "loss/out": 6.170996189117432, "created_at": "2025-01-14T20:44:59.529149+00:00"} {"global_step": 5494, "acc_step": 0, "speed/wps": 12974.644265853023, "speed/FLOPS": 203784617224791.78, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.994917869567871, "optim/lr": 0.0029994052967189464, "optim/total_tokens": 1440219136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433196, "loss/out": 6.011974334716797, "created_at": "2025-01-14T20:45:04.584833+00:00"} {"global_step": 5495, "acc_step": 0, "speed/wps": 12970.059505222205, "speed/FLOPS": 203712607258963.47, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0009, "optim/grad_norm": 2.1384475231170654, "optim/lr": 0.0029994028818429675, "optim/total_tokens": 1440481280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455517, "loss/out": 6.110057830810547, "created_at": "2025-01-14T20:45:09.642379+00:00"} {"global_step": 5496, "acc_step": 0, "speed/wps": 12963.990982739111, "speed/FLOPS": 203617292774342.72, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.335350513458252, "optim/lr": 0.002999400462074923, "optim/total_tokens": 1440743424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 6.094992637634277, "created_at": "2025-01-14T20:45:14.702375+00:00"} {"global_step": 5497, "acc_step": 0, "speed/wps": 12971.152443547997, "speed/FLOPS": 203729773357224.7, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.87217903137207, "optim/lr": 0.0029993980374148216, "optim/total_tokens": 1441005568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438531, "loss/out": 6.1131391525268555, "created_at": "2025-01-14T20:45:19.762679+00:00"} {"global_step": 5498, "acc_step": 0, "speed/wps": 12974.6503422346, "speed/FLOPS": 203784712662714.25, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.9261486530303955, "optim/lr": 0.0029993956078626697, "optim/total_tokens": 1441267712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 6.202882766723633, "created_at": "2025-01-14T20:45:24.821893+00:00"} {"global_step": 5499, "acc_step": 0, "speed/wps": 12977.13270433442, "speed/FLOPS": 203823701570614.94, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.4558801651000977, "optim/lr": 0.002999393173418476, "optim/total_tokens": 1441529856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450990, "loss/out": 6.021395206451416, "created_at": "2025-01-14T20:45:29.879896+00:00"} {"global_step": 5500, "acc_step": 0, "speed/wps": 12969.744923858074, "speed/FLOPS": 203707666326359.7, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6322565078735352, "optim/lr": 0.0029993907340822483, "optim/total_tokens": 1441792000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294941, "loss/out": 6.024024963378906, "created_at": "2025-01-14T20:45:34.941100+00:00"} {"global_step": 5501, "acc_step": 0, "speed/wps": 12973.5913746751, "speed/FLOPS": 203768080122018.4, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0009, "optim/grad_norm": 3.325143575668335, "optim/lr": 0.0029993882898539945, "optim/total_tokens": 1442054144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371881, "loss/out": 6.0388970375061035, "created_at": "2025-01-14T20:45:39.996371+00:00"} {"global_step": 5502, "acc_step": 0, "speed/wps": 12972.04520173356, "speed/FLOPS": 203743795351307.25, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 31.584482192993164, "optim/lr": 0.002999385840733723, "optim/total_tokens": 1442316288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447555, "loss/out": 6.077596664428711, "created_at": "2025-01-14T20:45:45.052607+00:00"} {"global_step": 5503, "acc_step": 0, "speed/wps": 12967.620142907044, "speed/FLOPS": 203674293721768.78, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.680396556854248, "optim/lr": 0.0029993833867214413, "optim/total_tokens": 1442578432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461622, "loss/out": 6.248278617858887, "created_at": "2025-01-14T20:45:50.112911+00:00"} {"global_step": 5504, "acc_step": 0, "speed/wps": 12968.15554299631, "speed/FLOPS": 203682702915887.53, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0012, "optim/grad_norm": 3.9488821029663086, "optim/lr": 0.0029993809278171576, "optim/total_tokens": 1442840576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 6.0240654945373535, "created_at": "2025-01-14T20:45:55.177976+00:00"} {"global_step": 5505, "acc_step": 0, "speed/wps": 12975.826736435198, "speed/FLOPS": 203803189550169.7, "speed/curr_iter_time": 1.2721, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.916546821594238, "optim/lr": 0.00299937846402088, "optim/total_tokens": 1443102720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456045, "loss/out": 6.113152027130127, "created_at": "2025-01-14T20:46:00.236615+00:00"} {"global_step": 5506, "acc_step": 0, "speed/wps": 12974.52473453708, "speed/FLOPS": 203782739821221.66, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.6345624923706055, "optim/lr": 0.0029993759953326165, "optim/total_tokens": 1443364864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 6.224649429321289, "created_at": "2025-01-14T20:46:05.291380+00:00"} {"global_step": 5507, "acc_step": 0, "speed/wps": 12974.38566607007, "speed/FLOPS": 203780555559847.16, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.461297988891602, "optim/lr": 0.0029993735217523753, "optim/total_tokens": 1443627008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 5.9798736572265625, "created_at": "2025-01-14T20:46:10.348893+00:00"} {"global_step": 5508, "acc_step": 0, "speed/wps": 12971.453396968442, "speed/FLOPS": 203734500244246.1, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.6272270679473877, "optim/lr": 0.002999371043280164, "optim/total_tokens": 1443889152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 6.084657669067383, "created_at": "2025-01-14T20:46:15.406518+00:00"} {"global_step": 5509, "acc_step": 0, "speed/wps": 12967.133774108566, "speed/FLOPS": 203666654631447.94, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.19801664352417, "optim/lr": 0.002999368559915991, "optim/total_tokens": 1444151296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454182, "loss/out": 6.160671234130859, "created_at": "2025-01-14T20:46:20.464103+00:00"} {"global_step": 5510, "acc_step": 0, "speed/wps": 12975.660509742267, "speed/FLOPS": 203800578731537.8, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.9094443321228027, "optim/lr": 0.0029993660716598645, "optim/total_tokens": 1444413440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439583, "loss/out": 6.055989742279053, "created_at": "2025-01-14T20:46:25.521239+00:00"} {"global_step": 5511, "acc_step": 0, "speed/wps": 12975.619842000115, "speed/FLOPS": 203799939988763.97, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.5461392402648926, "optim/lr": 0.0029993635785117924, "optim/total_tokens": 1444675584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316527, "loss/out": 6.040460109710693, "created_at": "2025-01-14T20:46:30.576882+00:00"} {"global_step": 5512, "acc_step": 0, "speed/wps": 12973.564023506744, "speed/FLOPS": 203767650534334.94, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.459061861038208, "optim/lr": 0.0029993610804717828, "optim/total_tokens": 1444937728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 404906, "loss/out": 5.926868438720703, "created_at": "2025-01-14T20:46:35.633140+00:00"} {"global_step": 5513, "acc_step": 0, "speed/wps": 12970.264838183773, "speed/FLOPS": 203715832295281.97, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7679808139801025, "optim/lr": 0.0029993585775398442, "optim/total_tokens": 1445199872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436381, "loss/out": 6.077449321746826, "created_at": "2025-01-14T20:46:40.691996+00:00"} {"global_step": 5514, "acc_step": 0, "speed/wps": 12967.838119350114, "speed/FLOPS": 203677717341338.22, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.8629655838012695, "optim/lr": 0.002999356069715984, "optim/total_tokens": 1445462016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 6.012751579284668, "created_at": "2025-01-14T20:46:45.755603+00:00"} {"global_step": 5515, "acc_step": 0, "speed/wps": 12969.29663323602, "speed/FLOPS": 203700625306124.47, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5107035636901855, "optim/lr": 0.0029993535570002113, "optim/total_tokens": 1445724160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443016, "loss/out": 5.963057518005371, "created_at": "2025-01-14T20:46:50.812599+00:00"} {"global_step": 5516, "acc_step": 0, "speed/wps": 12974.36034250488, "speed/FLOPS": 203780157818457.62, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.568840503692627, "optim/lr": 0.0029993510393925343, "optim/total_tokens": 1445986304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 5.928152084350586, "created_at": "2025-01-14T20:46:55.872610+00:00"} {"global_step": 5517, "acc_step": 0, "speed/wps": 12972.59614864204, "speed/FLOPS": 203752448729585.3, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.6051955223083496, "optim/lr": 0.0029993485168929605, "optim/total_tokens": 1446248448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449696, "loss/out": 5.96829080581665, "created_at": "2025-01-14T20:47:00.933409+00:00"} {"global_step": 5518, "acc_step": 0, "speed/wps": 12974.059519700171, "speed/FLOPS": 203775432982935.8, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.303075313568115, "optim/lr": 0.002999345989501498, "optim/total_tokens": 1446510592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318931, "loss/out": 6.237704277038574, "created_at": "2025-01-14T20:47:05.993086+00:00"} {"global_step": 5519, "acc_step": 0, "speed/wps": 12964.788885827047, "speed/FLOPS": 203629824938772.5, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.821613311767578, "optim/lr": 0.0029993434572181564, "optim/total_tokens": 1446772736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376973, "loss/out": 6.115327835083008, "created_at": "2025-01-14T20:47:11.056551+00:00"} {"global_step": 5520, "acc_step": 0, "speed/wps": 12973.238817743957, "speed/FLOPS": 203762542730950.8, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.101988315582275, "optim/lr": 0.0029993409200429425, "optim/total_tokens": 1447034880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455769, "loss/out": 6.023842811584473, "created_at": "2025-01-14T20:47:16.114066+00:00"} {"global_step": 5521, "acc_step": 0, "speed/wps": 12967.580498144727, "speed/FLOPS": 203673671046298.97, "speed/curr_iter_time": 1.2723, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.3473215103149414, "optim/lr": 0.0029993383779758653, "optim/total_tokens": 1447297024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439871, "loss/out": 6.018167495727539, "created_at": "2025-01-14T20:47:21.172779+00:00"} {"global_step": 5522, "acc_step": 0, "speed/wps": 12955.810388081136, "speed/FLOPS": 203488805293924.16, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.584658145904541, "optim/lr": 0.0029993358310169333, "optim/total_tokens": 1447559168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451797, "loss/out": 6.063536643981934, "created_at": "2025-01-14T20:47:26.234789+00:00"} {"global_step": 5523, "acc_step": 0, "speed/wps": 12969.579522263764, "speed/FLOPS": 203705068467035.47, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.044638156890869, "optim/lr": 0.0029993332791661536, "optim/total_tokens": 1447821312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 5.99414587020874, "created_at": "2025-01-14T20:47:31.294282+00:00"} {"global_step": 5524, "acc_step": 0, "speed/wps": 12971.01509950078, "speed/FLOPS": 203727616180232.66, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 42.561920166015625, "optim/lr": 0.0029993307224235365, "optim/total_tokens": 1448083456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444357, "loss/out": 6.013470649719238, "created_at": "2025-01-14T20:47:36.350459+00:00"} {"global_step": 5525, "acc_step": 0, "speed/wps": 12971.402416556299, "speed/FLOPS": 203733699526818.97, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.7610244750976562, "optim/lr": 0.002999328160789088, "optim/total_tokens": 1448345600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443545, "loss/out": 5.962746620178223, "created_at": "2025-01-14T20:47:41.417711+00:00"} {"global_step": 5526, "acc_step": 0, "speed/wps": 12970.59517131584, "speed/FLOPS": 203721020631046.34, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.4281415939331055, "optim/lr": 0.0029993255942628184, "optim/total_tokens": 1448607744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441442, "loss/out": 6.156578063964844, "created_at": "2025-01-14T20:47:46.475738+00:00"} {"global_step": 5527, "acc_step": 0, "speed/wps": 12969.080536668625, "speed/FLOPS": 203697231212585.75, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.98159646987915, "optim/lr": 0.002999323022844735, "optim/total_tokens": 1448869888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310132, "loss/out": 5.978649139404297, "created_at": "2025-01-14T20:47:51.533906+00:00"} {"global_step": 5528, "acc_step": 0, "speed/wps": 12967.86575466191, "speed/FLOPS": 203678151391884.88, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.431595802307129, "optim/lr": 0.002999320446534847, "optim/total_tokens": 1449132032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348945, "loss/out": 6.055197238922119, "created_at": "2025-01-14T20:47:56.598547+00:00"} {"global_step": 5529, "acc_step": 0, "speed/wps": 12971.323064277407, "speed/FLOPS": 203732453190237.5, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.7952170372009277, "optim/lr": 0.002999317865333162, "optim/total_tokens": 1449394176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449420, "loss/out": 6.083498001098633, "created_at": "2025-01-14T20:48:01.656186+00:00"} {"global_step": 5530, "acc_step": 0, "speed/wps": 12978.131024930675, "speed/FLOPS": 203839381567415.38, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.887869358062744, "optim/lr": 0.0029993152792396885, "optim/total_tokens": 1449656320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 5.972533226013184, "created_at": "2025-01-14T20:48:06.737473+00:00"} {"global_step": 5531, "acc_step": 0, "speed/wps": 12971.329082554346, "speed/FLOPS": 203732547715546.8, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.774142026901245, "optim/lr": 0.002999312688254436, "optim/total_tokens": 1449918464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 6.034748077392578, "created_at": "2025-01-14T20:48:11.794608+00:00"} {"global_step": 5532, "acc_step": 0, "speed/wps": 12976.154029293668, "speed/FLOPS": 203808330134259.47, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.602899551391602, "optim/lr": 0.002999310092377412, "optim/total_tokens": 1450180608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457106, "loss/out": 5.930041790008545, "created_at": "2025-01-14T20:48:16.854048+00:00"} {"global_step": 5533, "acc_step": 0, "speed/wps": 12972.926107858055, "speed/FLOPS": 203757631192484.3, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0008, "optim/grad_norm": 15.850139617919922, "optim/lr": 0.0029993074916086246, "optim/total_tokens": 1450442752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331435, "loss/out": 5.932432174682617, "created_at": "2025-01-14T20:48:21.914599+00:00"} {"global_step": 5534, "acc_step": 0, "speed/wps": 12963.288402219203, "speed/FLOPS": 203606257781830.75, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.9205620288848877, "optim/lr": 0.002999304885948083, "optim/total_tokens": 1450704896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333025, "loss/out": 6.00648307800293, "created_at": "2025-01-14T20:48:26.973900+00:00"} {"global_step": 5535, "acc_step": 0, "speed/wps": 12963.818386278966, "speed/FLOPS": 203614581909763.75, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.126547336578369, "optim/lr": 0.0029993022753957957, "optim/total_tokens": 1450967040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356429, "loss/out": 5.958189487457275, "created_at": "2025-01-14T20:48:32.037026+00:00"} {"global_step": 5536, "acc_step": 0, "speed/wps": 12972.943012002497, "speed/FLOPS": 203757896695301.5, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.3819210529327393, "optim/lr": 0.0029992996599517707, "optim/total_tokens": 1451229184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455003, "loss/out": 5.975165843963623, "created_at": "2025-01-14T20:48:37.092328+00:00"} {"global_step": 5537, "acc_step": 0, "speed/wps": 12973.093754529988, "speed/FLOPS": 203760264313837.7, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.1215181350708, "optim/lr": 0.0029992970396160177, "optim/total_tokens": 1451491328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 5.940244197845459, "created_at": "2025-01-14T20:48:42.147775+00:00"} {"global_step": 5538, "acc_step": 0, "speed/wps": 12969.291773750836, "speed/FLOPS": 203700548981232.03, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.697060585021973, "optim/lr": 0.0029992944143885434, "optim/total_tokens": 1451753472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 5.933767795562744, "created_at": "2025-01-14T20:48:47.209366+00:00"} {"global_step": 5539, "acc_step": 0, "speed/wps": 12969.59155319724, "speed/FLOPS": 203705257429377.94, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.68253755569458, "optim/lr": 0.002999291784269358, "optim/total_tokens": 1452015616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437432, "loss/out": 6.031314849853516, "created_at": "2025-01-14T20:48:52.267508+00:00"} {"global_step": 5540, "acc_step": 0, "speed/wps": 12976.231921224107, "speed/FLOPS": 203809553534061.75, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.1982200145721436, "optim/lr": 0.0029992891492584693, "optim/total_tokens": 1452277760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450462, "loss/out": 5.992972373962402, "created_at": "2025-01-14T20:48:57.321692+00:00"} {"global_step": 5541, "acc_step": 0, "speed/wps": 12972.125079378859, "speed/FLOPS": 203745049939490.06, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.303256034851074, "optim/lr": 0.002999286509355886, "optim/total_tokens": 1452539904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 6.0532708168029785, "created_at": "2025-01-14T20:49:02.379322+00:00"} {"global_step": 5542, "acc_step": 0, "speed/wps": 12977.485449005213, "speed/FLOPS": 203829241910391.72, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.170039176940918, "optim/lr": 0.002999283864561617, "optim/total_tokens": 1452802048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436679, "loss/out": 6.1201324462890625, "created_at": "2025-01-14T20:49:07.437043+00:00"} {"global_step": 5543, "acc_step": 0, "speed/wps": 12971.017390018811, "speed/FLOPS": 203727652155965.8, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 44.777278900146484, "optim/lr": 0.0029992812148756706, "optim/total_tokens": 1453064192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310904, "loss/out": 6.097203731536865, "created_at": "2025-01-14T20:49:12.494310+00:00"} {"global_step": 5544, "acc_step": 0, "speed/wps": 12975.590343272004, "speed/FLOPS": 203799476670703.5, "speed/curr_iter_time": 1.2722, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.043434143066406, "optim/lr": 0.0029992785602980555, "optim/total_tokens": 1453326336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362577, "loss/out": 6.015946865081787, "created_at": "2025-01-14T20:49:17.556690+00:00"} {"global_step": 5545, "acc_step": 0, "speed/wps": 12975.920842447604, "speed/FLOPS": 203804667614409.75, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.400009870529175, "optim/lr": 0.00299927590082878, "optim/total_tokens": 1453588480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441679, "loss/out": 6.184119701385498, "created_at": "2025-01-14T20:49:22.616111+00:00"} {"global_step": 5546, "acc_step": 0, "speed/wps": 12974.237744930788, "speed/FLOPS": 203778232255089.78, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.671075820922852, "optim/lr": 0.002999273236467854, "optim/total_tokens": 1453850624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 6.002048015594482, "created_at": "2025-01-14T20:49:27.671069+00:00"} {"global_step": 5547, "acc_step": 0, "speed/wps": 12973.720724421275, "speed/FLOPS": 203770111737527.25, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.1162078380584717, "optim/lr": 0.0029992705672152852, "optim/total_tokens": 1454112768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 5.996540069580078, "created_at": "2025-01-14T20:49:32.727213+00:00"} {"global_step": 5548, "acc_step": 0, "speed/wps": 12968.407529975628, "speed/FLOPS": 203686660717664.1, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0013, "optim/grad_norm": 2.581066131591797, "optim/lr": 0.002999267893071082, "optim/total_tokens": 1454374912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482002, "loss/out": 6.0042009353637695, "created_at": "2025-01-14T20:49:37.785442+00:00"} {"global_step": 5549, "acc_step": 0, "speed/wps": 12973.488141660526, "speed/FLOPS": 203766458705667.5, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.416540622711182, "optim/lr": 0.0029992652140352536, "optim/total_tokens": 1454637056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435053, "loss/out": 6.008332252502441, "created_at": "2025-01-14T20:49:42.846345+00:00"} {"global_step": 5550, "acc_step": 0, "speed/wps": 12977.136026086904, "speed/FLOPS": 203823753743302.25, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.32512903213501, "optim/lr": 0.002999262530107809, "optim/total_tokens": 1454899200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 6.082006454467773, "created_at": "2025-01-14T20:49:47.901062+00:00"} {"global_step": 5551, "acc_step": 0, "speed/wps": 12974.87488766318, "speed/FLOPS": 203788239457227.88, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.74189567565918, "optim/lr": 0.002999259841288757, "optim/total_tokens": 1455161344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 6.028200149536133, "created_at": "2025-01-14T20:49:52.960680+00:00"} {"global_step": 5552, "acc_step": 0, "speed/wps": 12970.336156756923, "speed/FLOPS": 203716952451475.25, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.6678879261016846, "optim/lr": 0.002999257147578106, "optim/total_tokens": 1455423488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425201, "loss/out": 6.007184028625488, "created_at": "2025-01-14T20:49:58.020879+00:00"} {"global_step": 5553, "acc_step": 0, "speed/wps": 12974.542544721413, "speed/FLOPS": 203783019554640.38, "speed/curr_iter_time": 1.272, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.342265605926514, "optim/lr": 0.002999254448975864, "optim/total_tokens": 1455685632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294835, "loss/out": 6.000369548797607, "created_at": "2025-01-14T20:50:03.081588+00:00"} {"global_step": 5554, "acc_step": 0, "speed/wps": 12971.869681458955, "speed/FLOPS": 203741038564204.44, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0009, "optim/grad_norm": 2.422646999359131, "optim/lr": 0.0029992517454820416, "optim/total_tokens": 1455947776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402512, "loss/out": 6.082269191741943, "created_at": "2025-01-14T20:50:08.137535+00:00"} {"global_step": 5555, "acc_step": 0, "speed/wps": 12971.404857849204, "speed/FLOPS": 203733737870678.72, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.126738548278809, "optim/lr": 0.0029992490370966463, "optim/total_tokens": 1456209920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437432, "loss/out": 5.917349815368652, "created_at": "2025-01-14T20:50:13.195314+00:00"} {"global_step": 5556, "acc_step": 0, "speed/wps": 12975.304237212513, "speed/FLOPS": 203794982981888.25, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.785740613937378, "optim/lr": 0.0029992463238196878, "optim/total_tokens": 1456472064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 6.037235260009766, "created_at": "2025-01-14T20:50:18.254080+00:00"} {"global_step": 5557, "acc_step": 0, "speed/wps": 12971.684008178268, "speed/FLOPS": 203738122310189.44, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.531782150268555, "optim/lr": 0.002999243605651174, "optim/total_tokens": 1456734208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 6.052285194396973, "created_at": "2025-01-14T20:50:23.311192+00:00"} {"global_step": 5558, "acc_step": 0, "speed/wps": 12972.088312535223, "speed/FLOPS": 203744472465687.8, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.935383319854736, "optim/lr": 0.0029992408825911146, "optim/total_tokens": 1456996352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467734, "loss/out": 6.0818023681640625, "created_at": "2025-01-14T20:50:28.370357+00:00"} {"global_step": 5559, "acc_step": 0, "speed/wps": 12965.005508240265, "speed/FLOPS": 203633227291442.47, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.372719764709473, "optim/lr": 0.002999238154639518, "optim/total_tokens": 1457258496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 6.035006523132324, "created_at": "2025-01-14T20:50:33.431799+00:00"} {"global_step": 5560, "acc_step": 0, "speed/wps": 12966.770925781258, "speed/FLOPS": 203660955599861.7, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.069762706756592, "optim/lr": 0.0029992354217963937, "optim/total_tokens": 1457520640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335952, "loss/out": 5.951637268066406, "created_at": "2025-01-14T20:50:38.490024+00:00"} {"global_step": 5561, "acc_step": 0, "speed/wps": 12972.424650235445, "speed/FLOPS": 203749755111446.25, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.651878595352173, "optim/lr": 0.0029992326840617488, "optim/total_tokens": 1457782784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320435, "loss/out": 5.991673469543457, "created_at": "2025-01-14T20:50:43.552778+00:00"} {"global_step": 5562, "acc_step": 0, "speed/wps": 12965.659759659864, "speed/FLOPS": 203643503209022.94, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.920180320739746, "optim/lr": 0.0029992299414355952, "optim/total_tokens": 1458044928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362284, "loss/out": 5.924263954162598, "created_at": "2025-01-14T20:50:48.616361+00:00"} {"global_step": 5563, "acc_step": 0, "speed/wps": 12969.273389966489, "speed/FLOPS": 203700260238636.9, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.559237480163574, "optim/lr": 0.002999227193917939, "optim/total_tokens": 1458307072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428118, "loss/out": 6.063755989074707, "created_at": "2025-01-14T20:50:53.704819+00:00"} {"global_step": 5564, "acc_step": 0, "speed/wps": 12972.541067389971, "speed/FLOPS": 203751583602832.0, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 17.47069549560547, "optim/lr": 0.002999224441508791, "optim/total_tokens": 1458569216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455517, "loss/out": 5.909048080444336, "created_at": "2025-01-14T20:50:58.764907+00:00"} {"global_step": 5565, "acc_step": 0, "speed/wps": 12972.076815018052, "speed/FLOPS": 203744291881381.4, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.7351202964782715, "optim/lr": 0.0029992216842081596, "optim/total_tokens": 1458831360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 5.966011047363281, "created_at": "2025-01-14T20:51:03.824385+00:00"} {"global_step": 5566, "acc_step": 0, "speed/wps": 12973.08796421875, "speed/FLOPS": 203760173369042.97, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.858360290527344, "optim/lr": 0.002999218922016054, "optim/total_tokens": 1459093504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440390, "loss/out": 5.918761730194092, "created_at": "2025-01-14T20:51:08.879611+00:00"} {"global_step": 5567, "acc_step": 0, "speed/wps": 12971.932779733444, "speed/FLOPS": 203742029609311.4, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.096501350402832, "optim/lr": 0.002999216154932482, "optim/total_tokens": 1459355648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300997, "loss/out": 6.158695220947266, "created_at": "2025-01-14T20:51:13.944606+00:00"} {"global_step": 5568, "acc_step": 0, "speed/wps": 12976.754933093085, "speed/FLOPS": 203817768154157.7, "speed/curr_iter_time": 1.2722, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.961353302001953, "optim/lr": 0.0029992133829574545, "optim/total_tokens": 1459617792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429438, "loss/out": 6.094585418701172, "created_at": "2025-01-14T20:51:19.004137+00:00"} {"global_step": 5569, "acc_step": 0, "speed/wps": 12969.343488134407, "speed/FLOPS": 203701361226689.03, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.4422738552093506, "optim/lr": 0.0029992106060909795, "optim/total_tokens": 1459879936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461622, "loss/out": 6.033170700073242, "created_at": "2025-01-14T20:51:24.061462+00:00"} {"global_step": 5570, "acc_step": 0, "speed/wps": 12965.213937140978, "speed/FLOPS": 203636500953739.5, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.001, "optim/grad_norm": 5.2479963302612305, "optim/lr": 0.0029992078243330665, "optim/total_tokens": 1460142080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 5.945277690887451, "created_at": "2025-01-14T20:51:29.122896+00:00"} {"global_step": 5571, "acc_step": 0, "speed/wps": 12974.78062218163, "speed/FLOPS": 203786758888305.78, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.5185418128967285, "optim/lr": 0.0029992050376837238, "optim/total_tokens": 1460404224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434230, "loss/out": 5.989265441894531, "created_at": "2025-01-14T20:51:34.177897+00:00"} {"global_step": 5572, "acc_step": 0, "speed/wps": 12971.29284146931, "speed/FLOPS": 203731978499505.88, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.721855163574219, "optim/lr": 0.002999202246142961, "optim/total_tokens": 1460666368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 5.987349033355713, "created_at": "2025-01-14T20:51:39.237299+00:00"} {"global_step": 5573, "acc_step": 0, "speed/wps": 12970.889712804808, "speed/FLOPS": 203725646809873.03, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.001, "optim/grad_norm": 5.603277206420898, "optim/lr": 0.0029991994497107874, "optim/total_tokens": 1460928512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445976, "loss/out": 6.012391090393066, "created_at": "2025-01-14T20:51:44.294301+00:00"} {"global_step": 5574, "acc_step": 0, "speed/wps": 12975.373472281382, "speed/FLOPS": 203796070413783.8, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.141427040100098, "optim/lr": 0.002999196648387212, "optim/total_tokens": 1461190656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460271, "loss/out": 6.105499267578125, "created_at": "2025-01-14T20:51:49.348661+00:00"} {"global_step": 5575, "acc_step": 0, "speed/wps": 12969.608011206168, "speed/FLOPS": 203705515925023.8, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.287306785583496, "optim/lr": 0.0029991938421722437, "optim/total_tokens": 1461452800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453660, "loss/out": 5.9741082191467285, "created_at": "2025-01-14T20:51:54.407186+00:00"} {"global_step": 5576, "acc_step": 0, "speed/wps": 12964.56150302226, "speed/FLOPS": 203626253579366.6, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.03297758102417, "optim/lr": 0.0029991910310658916, "optim/total_tokens": 1461714944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309845, "loss/out": 5.99009370803833, "created_at": "2025-01-14T20:51:59.465681+00:00"} {"global_step": 5577, "acc_step": 0, "speed/wps": 12966.795655834687, "speed/FLOPS": 203661344019333.44, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.83768367767334, "optim/lr": 0.0029991882150681654, "optim/total_tokens": 1461977088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357724, "loss/out": 5.973997116088867, "created_at": "2025-01-14T20:52:04.524272+00:00"} {"global_step": 5578, "acc_step": 0, "speed/wps": 12976.070113222835, "speed/FLOPS": 203807012117054.28, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.300341606140137, "optim/lr": 0.002999185394179074, "optim/total_tokens": 1462239232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454182, "loss/out": 5.9054646492004395, "created_at": "2025-01-14T20:52:09.580037+00:00"} {"global_step": 5579, "acc_step": 0, "speed/wps": 12971.944100040608, "speed/FLOPS": 203742207410291.16, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0009, "optim/grad_norm": 27.83970069885254, "optim/lr": 0.0029991825683986268, "optim/total_tokens": 1462501376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456851, "loss/out": 6.012167930603027, "created_at": "2025-01-14T20:52:14.636823+00:00"} {"global_step": 5580, "acc_step": 0, "speed/wps": 12968.842747219594, "speed/FLOPS": 203693496402531.78, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 17.040969848632812, "optim/lr": 0.002999179737726832, "optim/total_tokens": 1462763520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460034, "loss/out": 5.992984294891357, "created_at": "2025-01-14T20:52:19.701764+00:00"} {"global_step": 5581, "acc_step": 0, "speed/wps": 12960.293895172124, "speed/FLOPS": 203559224933772.5, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 16.51456642150879, "optim/lr": 0.0029991769021637, "optim/total_tokens": 1463025664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471738, "loss/out": 6.018937110900879, "created_at": "2025-01-14T20:52:24.764960+00:00"} {"global_step": 5582, "acc_step": 0, "speed/wps": 12964.89503556621, "speed/FLOPS": 203631492166291.06, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.0632805824279785, "optim/lr": 0.00299917406170924, "optim/total_tokens": 1463287808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449696, "loss/out": 5.999603271484375, "created_at": "2025-01-14T20:52:29.830749+00:00"} {"global_step": 5583, "acc_step": 0, "speed/wps": 12967.344436407288, "speed/FLOPS": 203669963372329.7, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.732354164123535, "optim/lr": 0.0029991712163634607, "optim/total_tokens": 1463549952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307679, "loss/out": 5.992801666259766, "created_at": "2025-01-14T20:52:34.897155+00:00"} {"global_step": 5584, "acc_step": 0, "speed/wps": 12969.62436646862, "speed/FLOPS": 203705772806895.12, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.757735729217529, "optim/lr": 0.0029991683661263722, "optim/total_tokens": 1463812096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 405430, "loss/out": 6.097577095031738, "created_at": "2025-01-14T20:52:39.953848+00:00"} {"global_step": 5585, "acc_step": 0, "speed/wps": 12960.284239820334, "speed/FLOPS": 203559073283204.78, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.9644012451171875, "optim/lr": 0.0029991655109979825, "optim/total_tokens": 1464074240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445170, "loss/out": 5.918748378753662, "created_at": "2025-01-14T20:52:45.016934+00:00"} {"global_step": 5586, "acc_step": 0, "speed/wps": 12962.351137768608, "speed/FLOPS": 203591536755697.84, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.784185409545898, "optim/lr": 0.002999162650978302, "optim/total_tokens": 1464336384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 6.037914752960205, "created_at": "2025-01-14T20:52:50.080729+00:00"} {"global_step": 5587, "acc_step": 0, "speed/wps": 12967.947583501313, "speed/FLOPS": 203679436626252.3, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.987611770629883, "optim/lr": 0.00299915978606734, "optim/total_tokens": 1464598528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461100, "loss/out": 5.975625038146973, "created_at": "2025-01-14T20:52:55.138043+00:00"} {"global_step": 5588, "acc_step": 0, "speed/wps": 12962.576358471746, "speed/FLOPS": 203595074156326.3, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.479430198669434, "optim/lr": 0.002999156916265105, "optim/total_tokens": 1464860672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454466, "loss/out": 6.00537109375, "created_at": "2025-01-14T20:53:00.208173+00:00"} {"global_step": 5589, "acc_step": 0, "speed/wps": 12968.441646969562, "speed/FLOPS": 203687196571935.5, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.173060417175293, "optim/lr": 0.002999154041571607, "optim/total_tokens": 1465122816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 5.907132148742676, "created_at": "2025-01-14T20:53:05.269736+00:00"} {"global_step": 5590, "acc_step": 0, "speed/wps": 12962.421132474317, "speed/FLOPS": 203592636118733.8, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.447229385375977, "optim/lr": 0.002999151161986855, "optim/total_tokens": 1465384960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427351, "loss/out": 5.96144962310791, "created_at": "2025-01-14T20:53:10.334700+00:00"} {"global_step": 5591, "acc_step": 0, "speed/wps": 12966.61356839112, "speed/FLOPS": 203658484085817.84, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0008, "optim/grad_norm": 9.367291450500488, "optim/lr": 0.002999148277510859, "optim/total_tokens": 1465647104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352115, "loss/out": 6.134564399719238, "created_at": "2025-01-14T20:53:15.399934+00:00"} {"global_step": 5592, "acc_step": 0, "speed/wps": 12963.507749546288, "speed/FLOPS": 203609702933017.44, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0013, "optim/grad_norm": 5.282706260681152, "optim/lr": 0.002999145388143628, "optim/total_tokens": 1465909248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 5.9675397872924805, "created_at": "2025-01-14T20:53:20.461131+00:00"} {"global_step": 5593, "acc_step": 0, "speed/wps": 12965.29072553355, "speed/FLOPS": 203637707020963.16, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.001, "optim/grad_norm": 6.489266872406006, "optim/lr": 0.0029991424938851717, "optim/total_tokens": 1466171392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 6.010480880737305, "created_at": "2025-01-14T20:53:25.526675+00:00"} {"global_step": 5594, "acc_step": 0, "speed/wps": 12968.828167405218, "speed/FLOPS": 203693267406512.22, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.621076822280884, "optim/lr": 0.0029991395947354996, "optim/total_tokens": 1466433536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 5.967550754547119, "created_at": "2025-01-14T20:53:30.588263+00:00"} {"global_step": 5595, "acc_step": 0, "speed/wps": 12963.656066161138, "speed/FLOPS": 203612032449262.06, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.383413314819336, "optim/lr": 0.00299913669069462, "optim/total_tokens": 1466695680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 5.956839561462402, "created_at": "2025-01-14T20:53:35.649928+00:00"} {"global_step": 5596, "acc_step": 0, "speed/wps": 12970.436913030115, "speed/FLOPS": 203718534967199.38, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.763044834136963, "optim/lr": 0.002999133781762544, "optim/total_tokens": 1466957824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442208, "loss/out": 6.151264190673828, "created_at": "2025-01-14T20:53:40.711005+00:00"} {"global_step": 5597, "acc_step": 0, "speed/wps": 12962.859878533554, "speed/FLOPS": 203599527228494.4, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0012, "optim/grad_norm": 12.099514961242676, "optim/lr": 0.00299913086793928, "optim/total_tokens": 1467219968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329658, "loss/out": 6.008502960205078, "created_at": "2025-01-14T20:53:45.775167+00:00"} {"global_step": 5598, "acc_step": 0, "speed/wps": 12968.508471388148, "speed/FLOPS": 203688246141255.78, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.570851802825928, "optim/lr": 0.002999127949224839, "optim/total_tokens": 1467482112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439586, "loss/out": 5.9688720703125, "created_at": "2025-01-14T20:53:50.834099+00:00"} {"global_step": 5599, "acc_step": 0, "speed/wps": 12968.65291343136, "speed/FLOPS": 203690514802028.6, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.410901069641113, "optim/lr": 0.0029991250256192277, "optim/total_tokens": 1467744256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472504, "loss/out": 6.14556884765625, "created_at": "2025-01-14T20:53:55.891591+00:00"} {"global_step": 5600, "acc_step": 0, "speed/wps": 12970.0889859798, "speed/FLOPS": 203713070294772.3, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 22.903074264526367, "optim/lr": 0.0029991220971224584, "optim/total_tokens": 1468006400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 6.066061973571777, "created_at": "2025-01-14T20:54:00.948213+00:00"} {"global_step": 5601, "acc_step": 0, "speed/wps": 12973.27200823858, "speed/FLOPS": 203763064033277.4, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 23.668048858642578, "optim/lr": 0.002999119163734539, "optim/total_tokens": 1468268544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 6.073678970336914, "created_at": "2025-01-14T20:54:06.003505+00:00"} {"global_step": 5602, "acc_step": 0, "speed/wps": 12964.881260055792, "speed/FLOPS": 203631275802970.44, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 10.007107734680176, "optim/lr": 0.0029991162254554803, "optim/total_tokens": 1468530688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 6.147580146789551, "created_at": "2025-01-14T20:54:11.069559+00:00"} {"global_step": 5603, "acc_step": 0, "speed/wps": 12964.241938505485, "speed/FLOPS": 203621234379503.0, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.001, "optim/grad_norm": 4.823336124420166, "optim/lr": 0.002999113282285291, "optim/total_tokens": 1468792832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 6.11267614364624, "created_at": "2025-01-14T20:54:16.129799+00:00"} {"global_step": 5604, "acc_step": 0, "speed/wps": 12968.900856471048, "speed/FLOPS": 203694409088176.3, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.870984077453613, "optim/lr": 0.0029991103342239807, "optim/total_tokens": 1469054976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 6.050999641418457, "created_at": "2025-01-14T20:54:21.186960+00:00"} {"global_step": 5605, "acc_step": 0, "speed/wps": 12972.568646221695, "speed/FLOPS": 203752016766281.5, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.156070709228516, "optim/lr": 0.00299910738127156, "optim/total_tokens": 1469317120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 6.205142498016357, "created_at": "2025-01-14T20:54:26.246488+00:00"} {"global_step": 5606, "acc_step": 0, "speed/wps": 12968.009106965254, "speed/FLOPS": 203680402936795.6, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.962204933166504, "optim/lr": 0.002999104423428037, "optim/total_tokens": 1469579264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299996, "loss/out": 6.070199489593506, "created_at": "2025-01-14T20:54:31.303807+00:00"} {"global_step": 5607, "acc_step": 0, "speed/wps": 12973.4486978023, "speed/FLOPS": 203765839185672.5, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.477466106414795, "optim/lr": 0.0029991014606934223, "optim/total_tokens": 1469841408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 392360, "loss/out": 6.111654758453369, "created_at": "2025-01-14T20:54:36.364996+00:00"} {"global_step": 5608, "acc_step": 0, "speed/wps": 12965.087274710242, "speed/FLOPS": 203634511546212.38, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.82542085647583, "optim/lr": 0.0029990984930677255, "optim/total_tokens": 1470103552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 6.149785995483398, "created_at": "2025-01-14T20:54:41.424317+00:00"} {"global_step": 5609, "acc_step": 0, "speed/wps": 12974.643576877834, "speed/FLOPS": 203784606403489.7, "speed/curr_iter_time": 1.2721, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.7185492515563965, "optim/lr": 0.0029990955205509563, "optim/total_tokens": 1470365696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 6.001194000244141, "created_at": "2025-01-14T20:54:46.479534+00:00"} {"global_step": 5610, "acc_step": 0, "speed/wps": 12968.410641724833, "speed/FLOPS": 203686709591961.38, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 24.592723846435547, "optim/lr": 0.0029990925431431243, "optim/total_tokens": 1470627840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436381, "loss/out": 6.0355143547058105, "created_at": "2025-01-14T20:54:51.540955+00:00"} {"global_step": 5611, "acc_step": 0, "speed/wps": 12965.941639543722, "speed/FLOPS": 203647930519942.47, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.330059051513672, "optim/lr": 0.0029990895608442395, "optim/total_tokens": 1470889984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 6.107730865478516, "created_at": "2025-01-14T20:54:56.608163+00:00"} {"global_step": 5612, "acc_step": 0, "speed/wps": 12966.92664306266, "speed/FLOPS": 203663401353744.06, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 14.39912223815918, "optim/lr": 0.002999086573654311, "optim/total_tokens": 1471152128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448890, "loss/out": 6.012511730194092, "created_at": "2025-01-14T20:55:01.671067+00:00"} {"global_step": 5613, "acc_step": 0, "speed/wps": 12972.72373030795, "speed/FLOPS": 203754452574964.0, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.7322540283203125, "optim/lr": 0.002999083581573349, "optim/total_tokens": 1471414272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315344, "loss/out": 5.927933216094971, "created_at": "2025-01-14T20:55:06.727659+00:00"} {"global_step": 5614, "acc_step": 0, "speed/wps": 12969.903179600751, "speed/FLOPS": 203710151950264.62, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0009, "optim/grad_norm": 21.932292938232422, "optim/lr": 0.002999080584601363, "optim/total_tokens": 1471676416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305632, "loss/out": 6.062936782836914, "created_at": "2025-01-14T20:55:11.784816+00:00"} {"global_step": 5615, "acc_step": 0, "speed/wps": 12967.403765269064, "speed/FLOPS": 203670895213630.28, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.142355918884277, "optim/lr": 0.0029990775827383634, "optim/total_tokens": 1471938560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352115, "loss/out": 6.036890983581543, "created_at": "2025-01-14T20:55:16.843519+00:00"} {"global_step": 5616, "acc_step": 0, "speed/wps": 12971.671349264721, "speed/FLOPS": 203737923484557.0, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 15.953439712524414, "optim/lr": 0.0029990745759843596, "optim/total_tokens": 1472200704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444357, "loss/out": 5.937342643737793, "created_at": "2025-01-14T20:55:21.904561+00:00"} {"global_step": 5617, "acc_step": 0, "speed/wps": 12966.740289318346, "speed/FLOPS": 203660474412112.75, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.001, "optim/grad_norm": 9.226838111877441, "optim/lr": 0.0029990715643393607, "optim/total_tokens": 1472462848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 6.070119857788086, "created_at": "2025-01-14T20:55:26.964656+00:00"} {"global_step": 5618, "acc_step": 0, "speed/wps": 12969.16654705372, "speed/FLOPS": 203698582123874.28, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.641414165496826, "optim/lr": 0.0029990685478033774, "optim/total_tokens": 1472724992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 6.0181450843811035, "created_at": "2025-01-14T20:55:32.022563+00:00"} {"global_step": 5619, "acc_step": 0, "speed/wps": 12967.623254278396, "speed/FLOPS": 203674342590131.34, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.680891036987305, "optim/lr": 0.002999065526376419, "optim/total_tokens": 1472987136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473073, "loss/out": 6.062505722045898, "created_at": "2025-01-14T20:55:37.083035+00:00"} {"global_step": 5620, "acc_step": 0, "speed/wps": 12967.951775222828, "speed/FLOPS": 203679502462998.72, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.62366247177124, "optim/lr": 0.002999062500058496, "optim/total_tokens": 1473249280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443022, "loss/out": 6.037537574768066, "created_at": "2025-01-14T20:55:42.142560+00:00"} {"global_step": 5621, "acc_step": 0, "speed/wps": 12967.282173706817, "speed/FLOPS": 203668985451060.88, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 19.85944175720215, "optim/lr": 0.002999059468849618, "optim/total_tokens": 1473511424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316287, "loss/out": 6.075558662414551, "created_at": "2025-01-14T20:55:47.200112+00:00"} {"global_step": 5622, "acc_step": 0, "speed/wps": 12976.301371981648, "speed/FLOPS": 203810644353644.28, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.817816734313965, "optim/lr": 0.002999056432749795, "optim/total_tokens": 1473773568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361998, "loss/out": 6.088282108306885, "created_at": "2025-01-14T20:55:52.257532+00:00"} {"global_step": 5623, "acc_step": 0, "speed/wps": 12969.595927653054, "speed/FLOPS": 203705326136217.62, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.013832092285156, "optim/lr": 0.0029990533917590365, "optim/total_tokens": 1474035712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467490, "loss/out": 5.968075752258301, "created_at": "2025-01-14T20:55:57.314858+00:00"} {"global_step": 5624, "acc_step": 0, "speed/wps": 12971.35578110413, "speed/FLOPS": 203732967052958.84, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.74364185333252, "optim/lr": 0.002999050345877353, "optim/total_tokens": 1474297856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464535, "loss/out": 5.956480979919434, "created_at": "2025-01-14T20:56:02.371259+00:00"} {"global_step": 5625, "acc_step": 0, "speed/wps": 12973.669264545788, "speed/FLOPS": 203769303489469.47, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5362188816070557, "optim/lr": 0.002999047295104754, "optim/total_tokens": 1474560000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 6.111581802368164, "created_at": "2025-01-14T20:56:07.430852+00:00"} {"global_step": 5626, "acc_step": 0, "speed/wps": 12968.337470604261, "speed/FLOPS": 203685560338965.12, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.673603057861328, "optim/lr": 0.002999044239441249, "optim/total_tokens": 1474822144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447554, "loss/out": 6.064886093139648, "created_at": "2025-01-14T20:56:12.494114+00:00"} {"global_step": 5627, "acc_step": 0, "speed/wps": 12964.930318873705, "speed/FLOPS": 203632046339119.62, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.80917739868164, "optim/lr": 0.002999041178886849, "optim/total_tokens": 1475084288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445695, "loss/out": 6.176980972290039, "created_at": "2025-01-14T20:56:17.553123+00:00"} {"global_step": 5628, "acc_step": 0, "speed/wps": 12971.282191788685, "speed/FLOPS": 203731811231637.44, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.119235038757324, "optim/lr": 0.0029990381134415633, "optim/total_tokens": 1475346432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463248, "loss/out": 6.113242149353027, "created_at": "2025-01-14T20:56:22.610378+00:00"} {"global_step": 5629, "acc_step": 0, "speed/wps": 12968.348239727686, "speed/FLOPS": 203685729482846.84, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.315633296966553, "optim/lr": 0.002999035043105402, "optim/total_tokens": 1475608576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 6.049080848693848, "created_at": "2025-01-14T20:56:27.685914+00:00"} {"global_step": 5630, "acc_step": 0, "speed/wps": 12964.355249508693, "speed/FLOPS": 203623014084514.0, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.590696096420288, "optim/lr": 0.002999031967878376, "optim/total_tokens": 1475870720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294941, "loss/out": 6.01738166809082, "created_at": "2025-01-14T20:56:32.750894+00:00"} {"global_step": 5631, "acc_step": 0, "speed/wps": 12964.210000355803, "speed/FLOPS": 203620732746974.6, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 19.063161849975586, "optim/lr": 0.002999028887760494, "optim/total_tokens": 1476132864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441679, "loss/out": 5.958440780639648, "created_at": "2025-01-14T20:56:37.809821+00:00"} {"global_step": 5632, "acc_step": 0, "speed/wps": 12958.29419580533, "speed/FLOPS": 203527816907342.12, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.522297859191895, "optim/lr": 0.0029990258027517668, "optim/total_tokens": 1476395008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 6.181033134460449, "created_at": "2025-01-14T20:56:42.875565+00:00"} {"global_step": 5633, "acc_step": 0, "speed/wps": 12963.52742809806, "speed/FLOPS": 203610012011714.2, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0009, "optim/grad_norm": 4.990942001342773, "optim/lr": 0.002999022712852204, "optim/total_tokens": 1476657152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448605, "loss/out": 6.087907314300537, "created_at": "2025-01-14T20:56:47.939329+00:00"} {"global_step": 5634, "acc_step": 0, "speed/wps": 12963.556640349647, "speed/FLOPS": 203610470830267.2, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 24.013395309448242, "optim/lr": 0.0029990196180618156, "optim/total_tokens": 1476919296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 6.109817981719971, "created_at": "2025-01-14T20:56:53.001564+00:00"} {"global_step": 5635, "acc_step": 0, "speed/wps": 12963.048408462222, "speed/FLOPS": 203602488350091.4, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.844338893890381, "optim/lr": 0.002999016518380613, "optim/total_tokens": 1477181440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324181, "loss/out": 6.1135101318359375, "created_at": "2025-01-14T20:56:58.060774+00:00"} {"global_step": 5636, "acc_step": 0, "speed/wps": 12968.834491694444, "speed/FLOPS": 203693366738165.0, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0011, "optim/grad_norm": 14.938074111938477, "optim/lr": 0.0029990134138086046, "optim/total_tokens": 1477443584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360941, "loss/out": 6.206696510314941, "created_at": "2025-01-14T20:57:03.120079+00:00"} {"global_step": 5637, "acc_step": 0, "speed/wps": 12962.801307615975, "speed/FLOPS": 203598607291748.94, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 16.905492782592773, "optim/lr": 0.002999010304345802, "optim/total_tokens": 1477705728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454995, "loss/out": 6.178022384643555, "created_at": "2025-01-14T20:57:08.183228+00:00"} {"global_step": 5638, "acc_step": 0, "speed/wps": 12955.743644368562, "speed/FLOPS": 203487756992202.72, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.500938415527344, "optim/lr": 0.002999007189992214, "optim/total_tokens": 1477967872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 6.002659797668457, "created_at": "2025-01-14T20:57:13.246928+00:00"} {"global_step": 5639, "acc_step": 0, "speed/wps": 12961.583079463218, "speed/FLOPS": 203579473344589.0, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.474096298217773, "optim/lr": 0.0029990040707478516, "optim/total_tokens": 1478230016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451275, "loss/out": 6.082341194152832, "created_at": "2025-01-14T20:57:18.313000+00:00"} {"global_step": 5640, "acc_step": 0, "speed/wps": 12963.457624437175, "speed/FLOPS": 203608915649299.7, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 11.945518493652344, "optim/lr": 0.0029990009466127248, "optim/total_tokens": 1478492160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444067, "loss/out": 6.151954174041748, "created_at": "2025-01-14T20:57:23.373112+00:00"} {"global_step": 5641, "acc_step": 0, "speed/wps": 12963.313964377654, "speed/FLOPS": 203606659270655.3, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.278755187988281, "optim/lr": 0.0029989978175868435, "optim/total_tokens": 1478754304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440628, "loss/out": 6.1843791007995605, "created_at": "2025-01-14T20:57:28.434557+00:00"} {"global_step": 5642, "acc_step": 0, "speed/wps": 12957.845978182328, "speed/FLOPS": 203520777033657.3, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.663857460021973, "optim/lr": 0.0029989946836702186, "optim/total_tokens": 1479016448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 6.158987998962402, "created_at": "2025-01-14T20:57:33.499327+00:00"} {"global_step": 5643, "acc_step": 0, "speed/wps": 12966.746329626514, "speed/FLOPS": 203660569283452.78, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.70760726928711, "optim/lr": 0.0029989915448628596, "optim/total_tokens": 1479278592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319909, "loss/out": 6.099842071533203, "created_at": "2025-01-14T20:57:38.557850+00:00"} {"global_step": 5644, "acc_step": 0, "speed/wps": 12967.05357560426, "speed/FLOPS": 203665395003733.47, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 35.523590087890625, "optim/lr": 0.0029989884011647774, "optim/total_tokens": 1479540736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354800, "loss/out": 6.00441837310791, "created_at": "2025-01-14T20:57:43.617097+00:00"} {"global_step": 5645, "acc_step": 0, "speed/wps": 12956.240926717192, "speed/FLOPS": 203495567494825.5, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.113391876220703, "optim/lr": 0.0029989852525759814, "optim/total_tokens": 1479802880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430791, "loss/out": 6.210635185241699, "created_at": "2025-01-14T20:57:48.680264+00:00"} {"global_step": 5646, "acc_step": 0, "speed/wps": 12968.472807712536, "speed/FLOPS": 203687685994223.28, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0009, "optim/grad_norm": 13.844037055969238, "optim/lr": 0.002998982099096483, "optim/total_tokens": 1480065024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 6.123671531677246, "created_at": "2025-01-14T20:57:53.744057+00:00"} {"global_step": 5647, "acc_step": 0, "speed/wps": 12961.693877702788, "speed/FLOPS": 203581213583194.16, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.26098108291626, "optim/lr": 0.002998978940726292, "optim/total_tokens": 1480327168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443016, "loss/out": 6.088089466094971, "created_at": "2025-01-14T20:57:58.807789+00:00"} {"global_step": 5648, "acc_step": 0, "speed/wps": 12967.47618028813, "speed/FLOPS": 203672032591013.34, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.31116008758545, "optim/lr": 0.0029989757774654185, "optim/total_tokens": 1480589312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 6.194663047790527, "created_at": "2025-01-14T20:58:03.865238+00:00"} {"global_step": 5649, "acc_step": 0, "speed/wps": 12967.742123164411, "speed/FLOPS": 203676209589329.16, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.356916427612305, "optim/lr": 0.002998972609313872, "optim/total_tokens": 1480851456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432941, "loss/out": 6.122128009796143, "created_at": "2025-01-14T20:58:08.932003+00:00"} {"global_step": 5650, "acc_step": 0, "speed/wps": 12962.158927226452, "speed/FLOPS": 203588517825008.3, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 49.84702682495117, "optim/lr": 0.0029989694362716644, "optim/total_tokens": 1481113600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357724, "loss/out": 6.061223030090332, "created_at": "2025-01-14T20:58:13.996785+00:00"} {"global_step": 5651, "acc_step": 0, "speed/wps": 12959.542229090664, "speed/FLOPS": 203547418985065.75, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.968542098999023, "optim/lr": 0.0029989662583388053, "optim/total_tokens": 1481375744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430268, "loss/out": 6.091917991638184, "created_at": "2025-01-14T20:58:19.063341+00:00"} {"global_step": 5652, "acc_step": 0, "speed/wps": 12960.556282775928, "speed/FLOPS": 203563346091648.6, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 15.972463607788086, "optim/lr": 0.002998963075515305, "optim/total_tokens": 1481637888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461385, "loss/out": 6.142433166503906, "created_at": "2025-01-14T20:58:24.131650+00:00"} {"global_step": 5653, "acc_step": 0, "speed/wps": 12961.676229313009, "speed/FLOPS": 203580936390981.25, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 31.414424896240234, "optim/lr": 0.0029989598878011744, "optim/total_tokens": 1481900032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 6.131045341491699, "created_at": "2025-01-14T20:58:29.197057+00:00"} {"global_step": 5654, "acc_step": 0, "speed/wps": 12964.192220915593, "speed/FLOPS": 203620453496434.6, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.780698776245117, "optim/lr": 0.0029989566951964227, "optim/total_tokens": 1482162176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 6.033258438110352, "created_at": "2025-01-14T20:58:34.255735+00:00"} {"global_step": 5655, "acc_step": 0, "speed/wps": 12953.969631463171, "speed/FLOPS": 203459893681774.34, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0015, "optim/grad_norm": 25.1491756439209, "optim/lr": 0.002998953497701062, "optim/total_tokens": 1482424320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463770, "loss/out": 6.1148786544799805, "created_at": "2025-01-14T20:58:39.318489+00:00"} {"global_step": 5656, "acc_step": 0, "speed/wps": 12965.835986306862, "speed/FLOPS": 203646271090674.84, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 21.246423721313477, "optim/lr": 0.0029989502953151016, "optim/total_tokens": 1482686464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421736, "loss/out": 6.132171154022217, "created_at": "2025-01-14T20:58:44.377726+00:00"} {"global_step": 5657, "acc_step": 0, "speed/wps": 12957.787633905924, "speed/FLOPS": 203519860656622.94, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0008, "optim/grad_norm": 22.036617279052734, "optim/lr": 0.0029989470880385526, "optim/total_tokens": 1482948608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320810, "loss/out": 6.119237899780273, "created_at": "2025-01-14T20:58:49.438900+00:00"} {"global_step": 5658, "acc_step": 0, "speed/wps": 12958.224990163699, "speed/FLOPS": 203526729937641.78, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 103.64201354980469, "optim/lr": 0.0029989438758714244, "optim/total_tokens": 1483210752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461913, "loss/out": 6.084141731262207, "created_at": "2025-01-14T20:58:54.500610+00:00"} {"global_step": 5659, "acc_step": 0, "speed/wps": 12960.939919821716, "speed/FLOPS": 203569371638625.25, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 47.70273208618164, "optim/lr": 0.002998940658813729, "optim/total_tokens": 1483472896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 6.173145294189453, "created_at": "2025-01-14T20:58:59.568403+00:00"} {"global_step": 5660, "acc_step": 0, "speed/wps": 12961.374504672358, "speed/FLOPS": 203576197390886.72, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 26.946317672729492, "optim/lr": 0.0029989374368654753, "optim/total_tokens": 1483735040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 6.212737560272217, "created_at": "2025-01-14T20:59:04.633728+00:00"} {"global_step": 5661, "acc_step": 0, "speed/wps": 12961.074231217148, "speed/FLOPS": 203571481183653.7, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.335552215576172, "optim/lr": 0.002998934210026675, "optim/total_tokens": 1483997184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 6.127969741821289, "created_at": "2025-01-14T20:59:09.694807+00:00"} {"global_step": 5662, "acc_step": 0, "speed/wps": 12955.239977047037, "speed/FLOPS": 203479846204806.72, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0012, "optim/grad_norm": 26.531545639038086, "optim/lr": 0.002998930978297338, "optim/total_tokens": 1484259328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322598, "loss/out": 6.149690628051758, "created_at": "2025-01-14T20:59:14.759740+00:00"} {"global_step": 5663, "acc_step": 0, "speed/wps": 12959.735617482618, "speed/FLOPS": 203550456415504.88, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 21.293128967285156, "optim/lr": 0.0029989277416774753, "optim/total_tokens": 1484521472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352687, "loss/out": 6.0479936599731445, "created_at": "2025-01-14T20:59:19.821462+00:00"} {"global_step": 5664, "acc_step": 0, "speed/wps": 12958.332750555017, "speed/FLOPS": 203528422462665.72, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 35.86290740966797, "optim/lr": 0.0029989245001670974, "optim/total_tokens": 1484783616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449135, "loss/out": 6.088872909545898, "created_at": "2025-01-14T20:59:24.890780+00:00"} {"global_step": 5665, "acc_step": 0, "speed/wps": 12958.78495378205, "speed/FLOPS": 203535524935738.62, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 18.686798095703125, "optim/lr": 0.002998921253766214, "optim/total_tokens": 1485045760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 6.106457233428955, "created_at": "2025-01-14T20:59:29.951727+00:00"} {"global_step": 5666, "acc_step": 0, "speed/wps": 12955.078484639294, "speed/FLOPS": 203477309744629.3, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.173186302185059, "optim/lr": 0.002998918002474837, "optim/total_tokens": 1485307904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434515, "loss/out": 6.0380048751831055, "created_at": "2025-01-14T20:59:35.014229+00:00"} {"global_step": 5667, "acc_step": 0, "speed/wps": 12956.853884404096, "speed/FLOPS": 203505194837590.47, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0008, "optim/grad_norm": 62.011871337890625, "optim/lr": 0.002998914746292976, "optim/total_tokens": 1485570048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471976, "loss/out": 6.066929817199707, "created_at": "2025-01-14T20:59:40.081071+00:00"} {"global_step": 5668, "acc_step": 0, "speed/wps": 12950.044583263581, "speed/FLOPS": 203398245406218.38, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.064713001251221, "optim/lr": 0.0029989114852206422, "optim/total_tokens": 1485832192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452325, "loss/out": 6.067410469055176, "created_at": "2025-01-14T20:59:45.146142+00:00"} {"global_step": 5669, "acc_step": 0, "speed/wps": 12951.510481923819, "speed/FLOPS": 203421269359033.25, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.989605903625488, "optim/lr": 0.002998908219257846, "optim/total_tokens": 1486094336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318931, "loss/out": 6.137128829956055, "created_at": "2025-01-14T20:59:50.210767+00:00"} {"global_step": 5670, "acc_step": 0, "speed/wps": 12957.4611905279, "speed/FLOPS": 203514733414790.88, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.07575511932373, "optim/lr": 0.002998904948404598, "optim/total_tokens": 1486356480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390782, "loss/out": 6.152322769165039, "created_at": "2025-01-14T20:59:55.274712+00:00"} {"global_step": 5671, "acc_step": 0, "speed/wps": 12958.9113642313, "speed/FLOPS": 203537510385550.25, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.001, "optim/grad_norm": 9.874942779541016, "optim/lr": 0.002998901672660909, "optim/total_tokens": 1486618624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474123, "loss/out": 6.057775020599365, "created_at": "2025-01-14T21:00:00.336591+00:00"} {"global_step": 5672, "acc_step": 0, "speed/wps": 12963.039491673017, "speed/FLOPS": 203602348299663.97, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 11.070786476135254, "optim/lr": 0.00299889839202679, "optim/total_tokens": 1486880768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 5.958428382873535, "created_at": "2025-01-14T21:00:05.395765+00:00"} {"global_step": 5673, "acc_step": 0, "speed/wps": 12957.429414619239, "speed/FLOPS": 203514234330480.66, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.947110176086426, "optim/lr": 0.0029988951065022516, "optim/total_tokens": 1487142912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 6.173559665679932, "created_at": "2025-01-14T21:00:10.463671+00:00"} {"global_step": 5674, "acc_step": 0, "speed/wps": 12955.695427884715, "speed/FLOPS": 203486999686067.5, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 36.40364456176758, "optim/lr": 0.0029988918160873033, "optim/total_tokens": 1487405056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 6.161759853363037, "created_at": "2025-01-14T21:00:15.531050+00:00"} {"global_step": 5675, "acc_step": 0, "speed/wps": 12964.110695020458, "speed/FLOPS": 203619173020223.2, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 12.288021087646484, "optim/lr": 0.002998888520781958, "optim/total_tokens": 1487667200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303511, "loss/out": 6.091497421264648, "created_at": "2025-01-14T21:00:20.594484+00:00"} {"global_step": 5676, "acc_step": 0, "speed/wps": 12969.456527388678, "speed/FLOPS": 203703136663509.56, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.054900169372559, "optim/lr": 0.0029988852205862244, "optim/total_tokens": 1487929344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443015, "loss/out": 6.140678405761719, "created_at": "2025-01-14T21:00:25.651198+00:00"} {"global_step": 5677, "acc_step": 0, "speed/wps": 12957.264095560786, "speed/FLOPS": 203511637767495.47, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.405580043792725, "optim/lr": 0.0029988819155001146, "optim/total_tokens": 1488191488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 6.033703327178955, "created_at": "2025-01-14T21:00:30.715796+00:00"} {"global_step": 5678, "acc_step": 0, "speed/wps": 12965.603941777408, "speed/FLOPS": 203642626512477.16, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.5142879486084, "optim/lr": 0.002998878605523639, "optim/total_tokens": 1488453632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463222, "loss/out": 6.101118087768555, "created_at": "2025-01-14T21:00:35.775375+00:00"} {"global_step": 5679, "acc_step": 0, "speed/wps": 12958.011775001678, "speed/FLOPS": 203523381100533.9, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 35.05925369262695, "optim/lr": 0.002998875290656808, "optim/total_tokens": 1488715776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433708, "loss/out": 6.2636237144470215, "created_at": "2025-01-14T21:00:40.836683+00:00"} {"global_step": 5680, "acc_step": 0, "speed/wps": 12964.097900415583, "speed/FLOPS": 203618972063372.0, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.677022933959961, "optim/lr": 0.002998871970899633, "optim/total_tokens": 1488977920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 6.083194255828857, "created_at": "2025-01-14T21:00:45.900530+00:00"} {"global_step": 5681, "acc_step": 0, "speed/wps": 12966.035038400723, "speed/FLOPS": 203649397477335.16, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 145.6902313232422, "optim/lr": 0.0029988686462521243, "optim/total_tokens": 1489240064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 6.123917102813721, "created_at": "2025-01-14T21:00:50.959300+00:00"} {"global_step": 5682, "acc_step": 0, "speed/wps": 12960.511959574003, "speed/FLOPS": 203562649934855.9, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.440326690673828, "optim/lr": 0.002998865316714293, "optim/total_tokens": 1489502208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446742, "loss/out": 6.232031345367432, "created_at": "2025-01-14T21:00:56.020251+00:00"} {"global_step": 5683, "acc_step": 0, "speed/wps": 12956.922536222939, "speed/FLOPS": 203506273108742.88, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.921558380126953, "optim/lr": 0.0029988619822861507, "optim/total_tokens": 1489764352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308663, "loss/out": 6.159920692443848, "created_at": "2025-01-14T21:01:01.084514+00:00"} {"global_step": 5684, "acc_step": 0, "speed/wps": 12967.00401565632, "speed/FLOPS": 203664616596648.6, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.4984636306762695, "optim/lr": 0.0029988586429677065, "optim/total_tokens": 1490026496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368425, "loss/out": 6.161731719970703, "created_at": "2025-01-14T21:01:06.184200+00:00"} {"global_step": 5685, "acc_step": 0, "speed/wps": 12957.792687589435, "speed/FLOPS": 203519940031666.97, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.4833984375, "optim/lr": 0.002998855298758973, "optim/total_tokens": 1490288640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475696, "loss/out": 6.1101484298706055, "created_at": "2025-01-14T21:01:11.250426+00:00"} {"global_step": 5686, "acc_step": 0, "speed/wps": 12963.9749330592, "speed/FLOPS": 203617040692065.16, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.2098870277404785, "optim/lr": 0.0029988519496599608, "optim/total_tokens": 1490550784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440390, "loss/out": 6.002498626708984, "created_at": "2025-01-14T21:01:16.310452+00:00"} {"global_step": 5687, "acc_step": 0, "speed/wps": 12966.360654817248, "speed/FLOPS": 203654511730600.53, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.033574104309082, "optim/lr": 0.00299884859567068, "optim/total_tokens": 1490812928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320435, "loss/out": 6.080657005310059, "created_at": "2025-01-14T21:01:21.372946+00:00"} {"global_step": 5688, "acc_step": 0, "speed/wps": 12953.981480835408, "speed/FLOPS": 203460079792448.06, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.124090194702148, "optim/lr": 0.002998845236791142, "optim/total_tokens": 1491075072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 412410, "loss/out": 6.100803852081299, "created_at": "2025-01-14T21:01:26.443615+00:00"} {"global_step": 5689, "acc_step": 0, "speed/wps": 12961.502713247624, "speed/FLOPS": 203578211082738.34, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.114557266235352, "optim/lr": 0.0029988418730213583, "optim/total_tokens": 1491337216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463222, "loss/out": 6.119723796844482, "created_at": "2025-01-14T21:01:31.507835+00:00"} {"global_step": 5690, "acc_step": 0, "speed/wps": 12958.404577493864, "speed/FLOPS": 203529550603440.88, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.233293056488037, "optim/lr": 0.002998838504361339, "optim/total_tokens": 1491599360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426015, "loss/out": 6.130823135375977, "created_at": "2025-01-14T21:01:36.576761+00:00"} {"global_step": 5691, "acc_step": 0, "speed/wps": 12963.731496291493, "speed/FLOPS": 203613217183111.16, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.083723545074463, "optim/lr": 0.002998835130811095, "optim/total_tokens": 1491861504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 6.085959434509277, "created_at": "2025-01-14T21:01:41.638112+00:00"} {"global_step": 5692, "acc_step": 0, "speed/wps": 12962.138355066876, "speed/FLOPS": 203588194710973.06, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.849159240722656, "optim/lr": 0.0029988317523706384, "optim/total_tokens": 1492123648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419618, "loss/out": 6.07143497467041, "created_at": "2025-01-14T21:01:46.700875+00:00"} {"global_step": 5693, "acc_step": 0, "speed/wps": 12954.841697287087, "speed/FLOPS": 203473590673882.97, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0008, "optim/grad_norm": 9.127645492553711, "optim/lr": 0.00299882836903998, "optim/total_tokens": 1492385792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302343, "loss/out": 6.216261386871338, "created_at": "2025-01-14T21:01:51.764066+00:00"} {"global_step": 5694, "acc_step": 0, "speed/wps": 12963.350772247964, "speed/FLOPS": 203607237388837.88, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 32.3209342956543, "optim/lr": 0.0029988249808191297, "optim/total_tokens": 1492647936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305736, "loss/out": 6.093096733093262, "created_at": "2025-01-14T21:01:56.824764+00:00"} {"global_step": 5695, "acc_step": 0, "speed/wps": 12959.067608758942, "speed/FLOPS": 203539964420559.34, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.202202320098877, "optim/lr": 0.0029988215877080992, "optim/total_tokens": 1492910080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365501, "loss/out": 6.154733657836914, "created_at": "2025-01-14T21:02:01.885552+00:00"} {"global_step": 5696, "acc_step": 0, "speed/wps": 12956.98552018286, "speed/FLOPS": 203507262358382.47, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 12.510642051696777, "optim/lr": 0.0029988181897069002, "optim/total_tokens": 1493172224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423105, "loss/out": 6.152542591094971, "created_at": "2025-01-14T21:02:06.947431+00:00"} {"global_step": 5697, "acc_step": 0, "speed/wps": 12958.85801714588, "speed/FLOPS": 203536672496267.16, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 14.52722454071045, "optim/lr": 0.0029988147868155427, "optim/total_tokens": 1493434368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 6.041361331939697, "created_at": "2025-01-14T21:02:12.011278+00:00"} {"global_step": 5698, "acc_step": 0, "speed/wps": 12956.330781722476, "speed/FLOPS": 203496978791157.2, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.732195854187012, "optim/lr": 0.002998811379034039, "optim/total_tokens": 1493696512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438240, "loss/out": 6.136453151702881, "created_at": "2025-01-14T21:02:17.076335+00:00"} {"global_step": 5699, "acc_step": 0, "speed/wps": 12959.160395496781, "speed/FLOPS": 203541421763779.47, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0013, "optim/grad_norm": 175.26092529296875, "optim/lr": 0.0029988079663623994, "optim/total_tokens": 1493958656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 6.22072696685791, "created_at": "2025-01-14T21:02:22.137981+00:00"} {"global_step": 5700, "acc_step": 0, "speed/wps": 12959.229950026564, "speed/FLOPS": 203542514213247.62, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 15.87899398803711, "optim/lr": 0.002998804548800635, "optim/total_tokens": 1494220800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433993, "loss/out": 6.231026649475098, "created_at": "2025-01-14T21:02:27.204732+00:00"} {"global_step": 5701, "acc_step": 0, "speed/wps": 12959.764392572364, "speed/FLOPS": 203550908367830.44, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 19.99089241027832, "optim/lr": 0.002998801126348757, "optim/total_tokens": 1494482944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352401, "loss/out": 6.138989448547363, "created_at": "2025-01-14T21:02:32.272988+00:00"} {"global_step": 5702, "acc_step": 0, "speed/wps": 12956.878357157657, "speed/FLOPS": 203505579215814.22, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 21.027311325073242, "optim/lr": 0.002998797699006777, "optim/total_tokens": 1494745088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454196, "loss/out": 6.129268646240234, "created_at": "2025-01-14T21:02:37.335284+00:00"} {"global_step": 5703, "acc_step": 0, "speed/wps": 12959.079130957587, "speed/FLOPS": 203540145392522.25, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.182158470153809, "optim/lr": 0.002998794266774706, "optim/total_tokens": 1495007232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465083, "loss/out": 6.290319442749023, "created_at": "2025-01-14T21:02:42.400904+00:00"} {"global_step": 5704, "acc_step": 0, "speed/wps": 12958.506098235954, "speed/FLOPS": 203531145126199.5, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 15.587409973144531, "optim/lr": 0.002998790829652555, "optim/total_tokens": 1495269376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467490, "loss/out": 6.156197547912598, "created_at": "2025-01-14T21:02:47.462020+00:00"} {"global_step": 5705, "acc_step": 0, "speed/wps": 12955.45378555541, "speed/FLOPS": 203483204361234.4, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.327683925628662, "optim/lr": 0.0029987873876403353, "optim/total_tokens": 1495531520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 6.136960029602051, "created_at": "2025-01-14T21:02:52.528662+00:00"} {"global_step": 5706, "acc_step": 0, "speed/wps": 12953.915392663306, "speed/FLOPS": 203459041786890.56, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.25717544555664, "optim/lr": 0.0029987839407380583, "optim/total_tokens": 1495793664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295129, "loss/out": 6.1245222091674805, "created_at": "2025-01-14T21:02:57.595035+00:00"} {"global_step": 5707, "acc_step": 0, "speed/wps": 12955.048637999364, "speed/FLOPS": 203476840962135.62, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.781249046325684, "optim/lr": 0.0029987804889457345, "optim/total_tokens": 1496055808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 392297, "loss/out": 6.159666061401367, "created_at": "2025-01-14T21:03:02.660593+00:00"} {"global_step": 5708, "acc_step": 0, "speed/wps": 12958.705669891284, "speed/FLOPS": 203534279673286.66, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.30246114730835, "optim/lr": 0.0029987770322633766, "optim/total_tokens": 1496317952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448361, "loss/out": 6.2492995262146, "created_at": "2025-01-14T21:03:07.722595+00:00"} {"global_step": 5709, "acc_step": 0, "speed/wps": 12960.377316101094, "speed/FLOPS": 203560535174094.34, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 19.462881088256836, "optim/lr": 0.0029987735706909946, "optim/total_tokens": 1496580096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 6.001592636108398, "created_at": "2025-01-14T21:03:12.786659+00:00"} {"global_step": 5710, "acc_step": 0, "speed/wps": 12951.756837827937, "speed/FLOPS": 203425138717035.9, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 46.51494216918945, "optim/lr": 0.0029987701042285997, "optim/total_tokens": 1496842240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428688, "loss/out": 6.167786598205566, "created_at": "2025-01-14T21:03:17.852707+00:00"} {"global_step": 5711, "acc_step": 0, "speed/wps": 12957.468648989474, "speed/FLOPS": 203514850560178.88, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.03685188293457, "optim/lr": 0.0029987666328762045, "optim/total_tokens": 1497104384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460556, "loss/out": 6.265162467956543, "created_at": "2025-01-14T21:03:22.918129+00:00"} {"global_step": 5712, "acc_step": 0, "speed/wps": 12956.734714055747, "speed/FLOPS": 203503323103512.4, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0008, "optim/grad_norm": 18.084030151367188, "optim/lr": 0.0029987631566338194, "optim/total_tokens": 1497366528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335952, "loss/out": 6.23038911819458, "created_at": "2025-01-14T21:03:27.980703+00:00"} {"global_step": 5713, "acc_step": 0, "speed/wps": 12953.623004590037, "speed/FLOPS": 203454449430416.56, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 18.700420379638672, "optim/lr": 0.0029987596755014557, "optim/total_tokens": 1497628672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376441, "loss/out": 6.22288179397583, "created_at": "2025-01-14T21:03:33.043913+00:00"} {"global_step": 5714, "acc_step": 0, "speed/wps": 12955.12196929803, "speed/FLOPS": 203477992730948.94, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 101.61124420166016, "optim/lr": 0.0029987561894791255, "optim/total_tokens": 1497890816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439054, "loss/out": 6.210328102111816, "created_at": "2025-01-14T21:03:38.111823+00:00"} {"global_step": 5715, "acc_step": 0, "speed/wps": 12943.806783379378, "speed/FLOPS": 203300272187401.66, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.001, "optim/grad_norm": 190.63861083984375, "optim/lr": 0.0029987526985668393, "optim/total_tokens": 1498152960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 6.338367938995361, "created_at": "2025-01-14T21:03:43.186376+00:00"} {"global_step": 5716, "acc_step": 0, "speed/wps": 12955.969067274165, "speed/FLOPS": 203491297568698.75, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 18.20053482055664, "optim/lr": 0.0029987492027646087, "optim/total_tokens": 1498415104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463750, "loss/out": 6.507530689239502, "created_at": "2025-01-14T21:03:48.248751+00:00"} {"global_step": 5717, "acc_step": 0, "speed/wps": 12952.015852266328, "speed/FLOPS": 203429206894713.97, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 12.274130821228027, "optim/lr": 0.0029987457020724455, "optim/total_tokens": 1498677248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319909, "loss/out": 6.433497905731201, "created_at": "2025-01-14T21:03:53.313847+00:00"} {"global_step": 5718, "acc_step": 0, "speed/wps": 12952.00097889499, "speed/FLOPS": 203428973287978.97, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.242348670959473, "optim/lr": 0.002998742196490361, "optim/total_tokens": 1498939392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 409163, "loss/out": 6.279646396636963, "created_at": "2025-01-14T21:03:58.377509+00:00"} {"global_step": 5719, "acc_step": 0, "speed/wps": 12951.854363944063, "speed/FLOPS": 203426670498704.6, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 57.676780700683594, "optim/lr": 0.002998738686018366, "optim/total_tokens": 1499201536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445407, "loss/out": 6.15997838973999, "created_at": "2025-01-14T21:04:03.445109+00:00"} {"global_step": 5720, "acc_step": 0, "speed/wps": 12961.552426923276, "speed/FLOPS": 203578991904328.2, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 15.692890167236328, "optim/lr": 0.0029987351706564722, "optim/total_tokens": 1499463680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 6.295498371124268, "created_at": "2025-01-14T21:04:08.505203+00:00"} {"global_step": 5721, "acc_step": 0, "speed/wps": 12952.319861543934, "speed/FLOPS": 203433981778171.3, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.2948479652404785, "optim/lr": 0.002998731650404692, "optim/total_tokens": 1499725824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480143, "loss/out": 6.275104522705078, "created_at": "2025-01-14T21:04:13.569764+00:00"} {"global_step": 5722, "acc_step": 0, "speed/wps": 12954.544283019237, "speed/FLOPS": 203468919374116.38, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 10.601037979125977, "optim/lr": 0.0029987281252630356, "optim/total_tokens": 1499987968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435567, "loss/out": 6.1229658126831055, "created_at": "2025-01-14T21:04:18.636204+00:00"} {"global_step": 5723, "acc_step": 0, "speed/wps": 12947.33194783664, "speed/FLOPS": 203355639739286.56, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.149065017700195, "optim/lr": 0.002998724595231516, "optim/total_tokens": 1500250112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 6.090306282043457, "created_at": "2025-01-14T21:04:23.704544+00:00"} {"global_step": 5724, "acc_step": 0, "speed/wps": 12943.993583293477, "speed/FLOPS": 203303206136742.6, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 32.58576583862305, "optim/lr": 0.002998721060310143, "optim/total_tokens": 1500512256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300465, "loss/out": 6.287585258483887, "created_at": "2025-01-14T21:04:28.773066+00:00"} {"global_step": 5725, "acc_step": 0, "speed/wps": 12950.568396376359, "speed/FLOPS": 203406472610949.2, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 23.45151710510254, "optim/lr": 0.0029987175204989292, "optim/total_tokens": 1500774400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453660, "loss/out": 6.148675918579102, "created_at": "2025-01-14T21:04:33.840463+00:00"} {"global_step": 5726, "acc_step": 0, "speed/wps": 12955.358227847924, "speed/FLOPS": 203481703496128.3, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.300848960876465, "optim/lr": 0.002998713975797886, "optim/total_tokens": 1501036544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 6.212918281555176, "created_at": "2025-01-14T21:04:38.908750+00:00"} {"global_step": 5727, "acc_step": 0, "speed/wps": 12950.371129710218, "speed/FLOPS": 203403374266884.66, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.749654769897461, "optim/lr": 0.0029987104262070245, "optim/total_tokens": 1501298688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461622, "loss/out": 6.205369472503662, "created_at": "2025-01-14T21:04:43.972844+00:00"} {"global_step": 5728, "acc_step": 0, "speed/wps": 12949.711316286573, "speed/FLOPS": 203393010990387.66, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0008, "optim/grad_norm": 22.73108673095703, "optim/lr": 0.002998706871726357, "optim/total_tokens": 1501560832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 6.018333435058594, "created_at": "2025-01-14T21:04:49.042043+00:00"} {"global_step": 5729, "acc_step": 0, "speed/wps": 12953.274066706468, "speed/FLOPS": 203448968881464.2, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.708385944366455, "optim/lr": 0.0029987033123558948, "optim/total_tokens": 1501822976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333837, "loss/out": 6.095934867858887, "created_at": "2025-01-14T21:04:54.109002+00:00"} {"global_step": 5730, "acc_step": 0, "speed/wps": 12954.691340464233, "speed/FLOPS": 203471229113368.12, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.736135005950928, "optim/lr": 0.0029986997480956493, "optim/total_tokens": 1502085120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449135, "loss/out": 6.166076183319092, "created_at": "2025-01-14T21:04:59.171463+00:00"} {"global_step": 5731, "acc_step": 0, "speed/wps": 12962.495022431709, "speed/FLOPS": 203593796662050.1, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.171242713928223, "optim/lr": 0.0029986961789456325, "optim/total_tokens": 1502347264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 6.017196178436279, "created_at": "2025-01-14T21:05:04.231223+00:00"} {"global_step": 5732, "acc_step": 0, "speed/wps": 12958.516800924326, "speed/FLOPS": 203531313226627.44, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.147073268890381, "optim/lr": 0.0029986926049058557, "optim/total_tokens": 1502609408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442256, "loss/out": 6.068506240844727, "created_at": "2025-01-14T21:05:09.292374+00:00"} {"global_step": 5733, "acc_step": 0, "speed/wps": 12956.462652586657, "speed/FLOPS": 203499050004288.72, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0008, "optim/grad_norm": 20.095809936523438, "optim/lr": 0.0029986890259763306, "optim/total_tokens": 1502871552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439583, "loss/out": 6.065696716308594, "created_at": "2025-01-14T21:05:14.354443+00:00"} {"global_step": 5734, "acc_step": 0, "speed/wps": 12955.616747101194, "speed/FLOPS": 203485763896243.6, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0013, "optim/grad_norm": 16.293949127197266, "optim/lr": 0.0029986854421570693, "optim/total_tokens": 1503133696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435329, "loss/out": 6.164212226867676, "created_at": "2025-01-14T21:05:19.417262+00:00"} {"global_step": 5735, "acc_step": 0, "speed/wps": 12956.161370998838, "speed/FLOPS": 203494317962947.75, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 32.4300651550293, "optim/lr": 0.0029986818534480826, "optim/total_tokens": 1503395840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 6.13443660736084, "created_at": "2025-01-14T21:05:24.487686+00:00"} {"global_step": 5736, "acc_step": 0, "speed/wps": 12964.668747792974, "speed/FLOPS": 203627938005860.4, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.819530963897705, "optim/lr": 0.002998678259849383, "optim/total_tokens": 1503657984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434807, "loss/out": 6.20682430267334, "created_at": "2025-01-14T21:05:29.547533+00:00"} {"global_step": 5737, "acc_step": 0, "speed/wps": 12952.408651318932, "speed/FLOPS": 203435376343597.6, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 20.369253158569336, "optim/lr": 0.002998674661360982, "optim/total_tokens": 1503920128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358017, "loss/out": 6.047916412353516, "created_at": "2025-01-14T21:05:34.612081+00:00"} {"global_step": 5738, "acc_step": 0, "speed/wps": 12956.158885345656, "speed/FLOPS": 203494278922348.88, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0009, "optim/grad_norm": 5.467800140380859, "optim/lr": 0.0029986710579828916, "optim/total_tokens": 1504182272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 6.213376998901367, "created_at": "2025-01-14T21:05:39.679104+00:00"} {"global_step": 5739, "acc_step": 0, "speed/wps": 12958.81477707089, "speed/FLOPS": 203535993351471.34, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.37866735458374, "optim/lr": 0.002998667449715123, "optim/total_tokens": 1504444416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 6.1471638679504395, "created_at": "2025-01-14T21:05:44.745162+00:00"} {"global_step": 5740, "acc_step": 0, "speed/wps": 12956.76217562226, "speed/FLOPS": 203503754425150.66, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.864302635192871, "optim/lr": 0.0029986638365576885, "optim/total_tokens": 1504706560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 6.184347629547119, "created_at": "2025-01-14T21:05:49.812516+00:00"} {"global_step": 5741, "acc_step": 0, "speed/wps": 12956.870763389275, "speed/FLOPS": 203505459945246.84, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.837017059326172, "optim/lr": 0.0029986602185105996, "optim/total_tokens": 1504968704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437717, "loss/out": 6.181654930114746, "created_at": "2025-01-14T21:05:54.874142+00:00"} {"global_step": 5742, "acc_step": 0, "speed/wps": 12964.592071864732, "speed/FLOPS": 203626733705043.62, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.2444047927856445, "optim/lr": 0.002998656595573868, "optim/total_tokens": 1505230848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314122, "loss/out": 6.202460289001465, "created_at": "2025-01-14T21:05:59.937022+00:00"} {"global_step": 5743, "acc_step": 0, "speed/wps": 12957.890574255813, "speed/FLOPS": 203521477476272.8, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.372014045715332, "optim/lr": 0.002998652967747506, "optim/total_tokens": 1505492992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432127, "loss/out": 6.13791036605835, "created_at": "2025-01-14T21:06:04.998221+00:00"} {"global_step": 5744, "acc_step": 0, "speed/wps": 12962.149925704769, "speed/FLOPS": 203588376443740.9, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 33.937583923339844, "optim/lr": 0.0029986493350315246, "optim/total_tokens": 1505755136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 6.442915439605713, "created_at": "2025-01-14T21:06:10.062964+00:00"} {"global_step": 5745, "acc_step": 0, "speed/wps": 12961.723454143417, "speed/FLOPS": 203581678121836.53, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.172179222106934, "optim/lr": 0.0029986456974259368, "optim/total_tokens": 1506017280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436916, "loss/out": 6.49575138092041, "created_at": "2025-01-14T21:06:15.125197+00:00"} {"global_step": 5746, "acc_step": 0, "speed/wps": 12951.428879461058, "speed/FLOPS": 203419987680222.06, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0008, "optim/grad_norm": 10.45584487915039, "optim/lr": 0.0029986420549307532, "optim/total_tokens": 1506279424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475987, "loss/out": 6.379697322845459, "created_at": "2025-01-14T21:06:20.189210+00:00"} {"global_step": 5747, "acc_step": 0, "speed/wps": 12955.569480889792, "speed/FLOPS": 203485021515442.94, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.654520034790039, "optim/lr": 0.0029986384075459866, "optim/total_tokens": 1506541568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 6.401004791259766, "created_at": "2025-01-14T21:06:25.258425+00:00"} {"global_step": 5748, "acc_step": 0, "speed/wps": 12951.020958845427, "speed/FLOPS": 203413580726411.9, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.581376075744629, "optim/lr": 0.002998634755271649, "optim/total_tokens": 1506803712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322311, "loss/out": 6.337541580200195, "created_at": "2025-01-14T21:06:30.327282+00:00"} {"global_step": 5749, "acc_step": 0, "speed/wps": 12949.708980859175, "speed/FLOPS": 203392974309291.16, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.094635963439941, "optim/lr": 0.0029986310981077514, "optim/total_tokens": 1507065856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450471, "loss/out": 6.234846115112305, "created_at": "2025-01-14T21:06:35.391979+00:00"} {"global_step": 5750, "acc_step": 0, "speed/wps": 12952.649305075312, "speed/FLOPS": 203439156141534.47, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.2666587829589844, "optim/lr": 0.0029986274360543066, "optim/total_tokens": 1507328000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 6.1856513023376465, "created_at": "2025-01-14T21:06:40.456589+00:00"} {"global_step": 5751, "acc_step": 0, "speed/wps": 12959.65804280676, "speed/FLOPS": 203549237998621.72, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 12.855215072631836, "optim/lr": 0.0029986237691113256, "optim/total_tokens": 1507590144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451278, "loss/out": 6.267238140106201, "created_at": "2025-01-14T21:06:45.518065+00:00"} {"global_step": 5752, "acc_step": 0, "speed/wps": 12955.760646883007, "speed/FLOPS": 203488024040055.97, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.1788177490234375, "optim/lr": 0.002998620097278822, "optim/total_tokens": 1507852288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 6.237640380859375, "created_at": "2025-01-14T21:06:50.582323+00:00"} {"global_step": 5753, "acc_step": 0, "speed/wps": 12956.185471388113, "speed/FLOPS": 203494696492678.72, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.558230400085449, "optim/lr": 0.0029986164205568063, "optim/total_tokens": 1508114432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441679, "loss/out": 6.253554344177246, "created_at": "2025-01-14T21:06:55.645214+00:00"} {"global_step": 5754, "acc_step": 0, "speed/wps": 12962.05499503341, "speed/FLOPS": 203586885427098.78, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.073564529418945, "optim/lr": 0.0029986127389452907, "optim/total_tokens": 1508376576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 339038, "loss/out": 6.111812114715576, "created_at": "2025-01-14T21:07:00.706139+00:00"} {"global_step": 5755, "acc_step": 0, "speed/wps": 12957.802205609358, "speed/FLOPS": 203520089525249.0, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.025913238525391, "optim/lr": 0.002998609052444288, "optim/total_tokens": 1508638720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436151, "loss/out": 6.201674461364746, "created_at": "2025-01-14T21:07:05.771192+00:00"} {"global_step": 5756, "acc_step": 0, "speed/wps": 12950.842110016318, "speed/FLOPS": 203410771659786.7, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0009, "optim/grad_norm": 4.825014591217041, "optim/lr": 0.002998605361053809, "optim/total_tokens": 1508900864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 6.107743740081787, "created_at": "2025-01-14T21:07:10.842610+00:00"} {"global_step": 5757, "acc_step": 0, "speed/wps": 12952.682130425776, "speed/FLOPS": 203439671708770.25, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.308493614196777, "optim/lr": 0.002998601664773867, "optim/total_tokens": 1509163008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466683, "loss/out": 6.098193168640137, "created_at": "2025-01-14T21:07:15.912526+00:00"} {"global_step": 5758, "acc_step": 0, "speed/wps": 12951.827180635366, "speed/FLOPS": 203426243547487.72, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.1022419929504395, "optim/lr": 0.0029985979636044735, "optim/total_tokens": 1509425152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322007, "loss/out": 6.19175386428833, "created_at": "2025-01-14T21:07:20.977558+00:00"} {"global_step": 5759, "acc_step": 0, "speed/wps": 12950.706752135478, "speed/FLOPS": 203408645678266.9, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.4048075675964355, "optim/lr": 0.00299859425754564, "optim/total_tokens": 1509687296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350526, "loss/out": 6.149072170257568, "created_at": "2025-01-14T21:07:26.047372+00:00"} {"global_step": 5760, "acc_step": 0, "speed/wps": 12953.820525336952, "speed/FLOPS": 203457551765168.8, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.891665935516357, "optim/lr": 0.00299859054659738, "optim/total_tokens": 1509949440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452325, "loss/out": 6.026988983154297, "created_at": "2025-01-14T21:07:31.114030+00:00"} {"global_step": 5761, "acc_step": 0, "speed/wps": 12956.975805344253, "speed/FLOPS": 203507109773492.56, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.142483711242676, "optim/lr": 0.002998586830759704, "optim/total_tokens": 1510211584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457380, "loss/out": 6.109255790710449, "created_at": "2025-01-14T21:07:36.180230+00:00"} {"global_step": 5762, "acc_step": 0, "speed/wps": 12948.111288396762, "speed/FLOPS": 203367880353709.44, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0009, "optim/grad_norm": 5.26671028137207, "optim/lr": 0.002998583110032625, "optim/total_tokens": 1510473728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429217, "loss/out": 6.168001174926758, "created_at": "2025-01-14T21:07:41.270631+00:00"} {"global_step": 5763, "acc_step": 0, "speed/wps": 12947.642043289861, "speed/FLOPS": 203360510214495.22, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.9432432651519775, "optim/lr": 0.0029985793844161555, "optim/total_tokens": 1510735872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317059, "loss/out": 6.1593546867370605, "created_at": "2025-01-14T21:07:46.342054+00:00"} {"global_step": 5764, "acc_step": 0, "speed/wps": 12952.781980785954, "speed/FLOPS": 203441239995885.62, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.851170539855957, "optim/lr": 0.002998575653910307, "optim/total_tokens": 1510998016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436096, "loss/out": 6.098716735839844, "created_at": "2025-01-14T21:07:51.406463+00:00"} {"global_step": 5765, "acc_step": 0, "speed/wps": 12955.719824703261, "speed/FLOPS": 203487382871627.5, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.9351584911346436, "optim/lr": 0.002998571918515092, "optim/total_tokens": 1511260160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 6.167489051818848, "created_at": "2025-01-14T21:07:56.472317+00:00"} {"global_step": 5766, "acc_step": 0, "speed/wps": 12949.109683222126, "speed/FLOPS": 203383561516380.03, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.888761281967163, "optim/lr": 0.002998568178230522, "optim/total_tokens": 1511522304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440919, "loss/out": 6.057867050170898, "created_at": "2025-01-14T21:08:01.541853+00:00"} {"global_step": 5767, "acc_step": 0, "speed/wps": 12956.235816992219, "speed/FLOPS": 203495487239573.3, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.15805435180664, "optim/lr": 0.0029985644330566105, "optim/total_tokens": 1511784448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 6.088820934295654, "created_at": "2025-01-14T21:08:06.609299+00:00"} {"global_step": 5768, "acc_step": 0, "speed/wps": 12950.653706156063, "speed/FLOPS": 203407812518267.38, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 42.585914611816406, "optim/lr": 0.002998560682993368, "optim/total_tokens": 1512046592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358144, "loss/out": 6.026308536529541, "created_at": "2025-01-14T21:08:11.679341+00:00"} {"global_step": 5769, "acc_step": 0, "speed/wps": 12948.35838941285, "speed/FLOPS": 203371761414719.47, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.619398593902588, "optim/lr": 0.0029985569280408086, "optim/total_tokens": 1512308736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465633, "loss/out": 6.146432876586914, "created_at": "2025-01-14T21:08:16.745907+00:00"} {"global_step": 5770, "acc_step": 0, "speed/wps": 12953.023366092373, "speed/FLOPS": 203445031283821.1, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.433786392211914, "optim/lr": 0.0029985531681989436, "optim/total_tokens": 1512570880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448077, "loss/out": 6.142195701599121, "created_at": "2025-01-14T21:08:21.814330+00:00"} {"global_step": 5771, "acc_step": 0, "speed/wps": 12956.435757668722, "speed/FLOPS": 203498627582645.06, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.054411888122559, "optim/lr": 0.0029985494034677853, "optim/total_tokens": 1512833024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 6.066657066345215, "created_at": "2025-01-14T21:08:26.878488+00:00"} {"global_step": 5772, "acc_step": 0, "speed/wps": 12954.547464434903, "speed/FLOPS": 203468969342621.12, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.6113312244415283, "optim/lr": 0.0029985456338473457, "optim/total_tokens": 1513095168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424678, "loss/out": 6.087242603302002, "created_at": "2025-01-14T21:08:31.945619+00:00"} {"global_step": 5773, "acc_step": 0, "speed/wps": 12960.69443189158, "speed/FLOPS": 203565515913346.66, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.701829433441162, "optim/lr": 0.0029985418593376375, "optim/total_tokens": 1513357312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291719, "loss/out": 5.960172176361084, "created_at": "2025-01-14T21:08:37.005739+00:00"} {"global_step": 5774, "acc_step": 0, "speed/wps": 12955.366005886834, "speed/FLOPS": 203481825660917.5, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.1985392570495605, "optim/lr": 0.002998538079938673, "optim/total_tokens": 1513619456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303928, "loss/out": 6.10566520690918, "created_at": "2025-01-14T21:08:42.068807+00:00"} {"global_step": 5775, "acc_step": 0, "speed/wps": 12945.545043282982, "speed/FLOPS": 203327573947807.2, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.4429426193237305, "optim/lr": 0.002998534295650465, "optim/total_tokens": 1513881600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415820, "loss/out": 6.109462738037109, "created_at": "2025-01-14T21:08:47.140000+00:00"} {"global_step": 5776, "acc_step": 0, "speed/wps": 12948.265273634124, "speed/FLOPS": 203370298903460.1, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.598239898681641, "optim/lr": 0.0029985305064730242, "optim/total_tokens": 1514143744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460578, "loss/out": 6.1237874031066895, "created_at": "2025-01-14T21:08:52.215875+00:00"} {"global_step": 5777, "acc_step": 0, "speed/wps": 12953.896174900747, "speed/FLOPS": 203458739945521.6, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.866119384765625, "optim/lr": 0.002998526712406365, "optim/total_tokens": 1514405888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459528, "loss/out": 6.0608415603637695, "created_at": "2025-01-14T21:08:57.278712+00:00"} {"global_step": 5778, "acc_step": 0, "speed/wps": 12959.808713048076, "speed/FLOPS": 203551604481804.22, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0012, "optim/grad_norm": 8.40056324005127, "optim/lr": 0.0029985229134504984, "optim/total_tokens": 1514668032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 6.151706695556641, "created_at": "2025-01-14T21:09:02.342183+00:00"} {"global_step": 5779, "acc_step": 0, "speed/wps": 12962.499234507011, "speed/FLOPS": 203593862818480.72, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.842912673950195, "optim/lr": 0.0029985191096054374, "optim/total_tokens": 1514930176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324713, "loss/out": 6.076956748962402, "created_at": "2025-01-14T21:09:07.406870+00:00"} {"global_step": 5780, "acc_step": 0, "speed/wps": 12948.328533225247, "speed/FLOPS": 203371292482266.47, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.515040397644043, "optim/lr": 0.002998515300871194, "optim/total_tokens": 1515192320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452325, "loss/out": 5.941301345825195, "created_at": "2025-01-14T21:09:12.475106+00:00"} {"global_step": 5781, "acc_step": 0, "speed/wps": 12946.04475598262, "speed/FLOPS": 203335422622432.22, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.2275466918945312, "optim/lr": 0.0029985114872477813, "optim/total_tokens": 1515454464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 6.157610893249512, "created_at": "2025-01-14T21:09:17.545620+00:00"} {"global_step": 5782, "acc_step": 0, "speed/wps": 12963.414840841293, "speed/FLOPS": 203608243674135.56, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 17.818471908569336, "optim/lr": 0.002998507668735211, "optim/total_tokens": 1515716608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451278, "loss/out": 6.022632598876953, "created_at": "2025-01-14T21:09:22.606280+00:00"} {"global_step": 5783, "acc_step": 0, "speed/wps": 12956.993185658868, "speed/FLOPS": 203507382755216.62, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.480737209320068, "optim/lr": 0.002998503845333496, "optim/total_tokens": 1515978752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 6.081276893615723, "created_at": "2025-01-14T21:09:27.673042+00:00"} {"global_step": 5784, "acc_step": 0, "speed/wps": 12957.237163913449, "speed/FLOPS": 203511214768966.1, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.295729637145996, "optim/lr": 0.0029985000170426484, "optim/total_tokens": 1516240896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310904, "loss/out": 5.929466724395752, "created_at": "2025-01-14T21:09:32.734338+00:00"} {"global_step": 5785, "acc_step": 0, "speed/wps": 12954.307739363734, "speed/FLOPS": 203465204130961.66, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 17.69988441467285, "optim/lr": 0.0029984961838626814, "optim/total_tokens": 1516503040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 399017, "loss/out": 6.11125373840332, "created_at": "2025-01-14T21:09:37.797001+00:00"} {"global_step": 5786, "acc_step": 0, "speed/wps": 12951.758454073375, "speed/FLOPS": 203425164102391.47, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.5375373363494873, "optim/lr": 0.0029984923457936066, "optim/total_tokens": 1516765184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435852, "loss/out": 6.119072437286377, "created_at": "2025-01-14T21:09:42.865846+00:00"} {"global_step": 5787, "acc_step": 0, "speed/wps": 12948.710162423582, "speed/FLOPS": 203377286493239.78, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.486080169677734, "optim/lr": 0.002998488502835437, "optim/total_tokens": 1517027328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480426, "loss/out": 6.1074628829956055, "created_at": "2025-01-14T21:09:47.930869+00:00"} {"global_step": 5788, "acc_step": 0, "speed/wps": 12952.894531249089, "speed/FLOPS": 203443007755569.7, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.016432523727417, "optim/lr": 0.0029984846549881857, "optim/total_tokens": 1517289472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451559, "loss/out": 6.062798023223877, "created_at": "2025-01-14T21:09:52.996625+00:00"} {"global_step": 5789, "acc_step": 0, "speed/wps": 12955.632709304988, "speed/FLOPS": 203486014604588.53, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.23102855682373, "optim/lr": 0.0029984808022518643, "optim/total_tokens": 1517551616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457095, "loss/out": 6.250726222991943, "created_at": "2025-01-14T21:09:58.061220+00:00"} {"global_step": 5790, "acc_step": 0, "speed/wps": 12962.275019324477, "speed/FLOPS": 203590341210935.06, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.57124137878418, "optim/lr": 0.002998476944626486, "optim/total_tokens": 1517813760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456045, "loss/out": 6.222557067871094, "created_at": "2025-01-14T21:10:03.121416+00:00"} {"global_step": 5791, "acc_step": 0, "speed/wps": 12962.463095276476, "speed/FLOPS": 203593295202204.66, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.417675971984863, "optim/lr": 0.002998473082112062, "optim/total_tokens": 1518075904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305035, "loss/out": 6.235545635223389, "created_at": "2025-01-14T21:10:08.184594+00:00"} {"global_step": 5792, "acc_step": 0, "speed/wps": 12962.674615908913, "speed/FLOPS": 203596617424396.88, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.357324600219727, "optim/lr": 0.0029984692147086074, "optim/total_tokens": 1518338048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 409401, "loss/out": 6.087643146514893, "created_at": "2025-01-14T21:10:13.246087+00:00"} {"global_step": 5793, "acc_step": 0, "speed/wps": 12947.65234448463, "speed/FLOPS": 203360672008914.38, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.5212316513061523, "optim/lr": 0.002998465342416133, "optim/total_tokens": 1518600192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 6.204376697540283, "created_at": "2025-01-14T21:10:18.311820+00:00"} {"global_step": 5794, "acc_step": 0, "speed/wps": 12961.38684984789, "speed/FLOPS": 203576391288832.5, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.4901697635650635, "optim/lr": 0.002998461465234652, "optim/total_tokens": 1518862336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432941, "loss/out": 6.107993125915527, "created_at": "2025-01-14T21:10:23.372340+00:00"} {"global_step": 5795, "acc_step": 0, "speed/wps": 12963.19475397318, "speed/FLOPS": 203604786907440.78, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.117882251739502, "optim/lr": 0.0029984575831641766, "optim/total_tokens": 1519124480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467490, "loss/out": 6.07951545715332, "created_at": "2025-01-14T21:10:28.438031+00:00"} {"global_step": 5796, "acc_step": 0, "speed/wps": 12950.995390322785, "speed/FLOPS": 203413179137628.97, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.4928104877471924, "optim/lr": 0.00299845369620472, "optim/total_tokens": 1519386624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323123, "loss/out": 6.000771522521973, "created_at": "2025-01-14T21:10:33.502332+00:00"} {"global_step": 5797, "acc_step": 0, "speed/wps": 12954.368871144725, "speed/FLOPS": 203466164289585.97, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.9564919471740723, "optim/lr": 0.0029984498043562945, "optim/total_tokens": 1519648768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354159, "loss/out": 5.967270851135254, "created_at": "2025-01-14T21:10:38.569411+00:00"} {"global_step": 5798, "acc_step": 0, "speed/wps": 12955.907223861352, "speed/FLOPS": 203490326232918.75, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.777256488800049, "optim/lr": 0.002998445907618913, "optim/total_tokens": 1519910912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 6.0474534034729, "created_at": "2025-01-14T21:10:43.632587+00:00"} {"global_step": 5799, "acc_step": 0, "speed/wps": 12957.844763665413, "speed/FLOPS": 203520757958000.22, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0008, "optim/grad_norm": 10.078312873840332, "optim/lr": 0.0029984420059925884, "optim/total_tokens": 1520173056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 5.971451759338379, "created_at": "2025-01-14T21:10:48.696700+00:00"} {"global_step": 5800, "acc_step": 0, "speed/wps": 12960.721497269911, "speed/FLOPS": 203565941012305.1, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.841930389404297, "optim/lr": 0.002998438099477333, "optim/total_tokens": 1520435200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 6.111438751220703, "created_at": "2025-01-14T21:10:53.760569+00:00"} {"global_step": 5801, "acc_step": 0, "speed/wps": 12956.83054495541, "speed/FLOPS": 203504828259477.1, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.4392805099487305, "optim/lr": 0.0029984341880731603, "optim/total_tokens": 1520697344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449696, "loss/out": 6.126657485961914, "created_at": "2025-01-14T21:10:58.826891+00:00"} {"global_step": 5802, "acc_step": 0, "speed/wps": 12955.725802271761, "speed/FLOPS": 203487476757554.84, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.288520097732544, "optim/lr": 0.0029984302717800814, "optim/total_tokens": 1520959488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 6.047196388244629, "created_at": "2025-01-14T21:11:03.891168+00:00"} {"global_step": 5803, "acc_step": 0, "speed/wps": 12957.964584372046, "speed/FLOPS": 203522639906848.8, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.052390098571777, "optim/lr": 0.002998426350598111, "optim/total_tokens": 1521221632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429977, "loss/out": 6.052665710449219, "created_at": "2025-01-14T21:11:08.954828+00:00"} {"global_step": 5804, "acc_step": 0, "speed/wps": 12957.65946063885, "speed/FLOPS": 203517847519335.03, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.108902454376221, "optim/lr": 0.002998422424527261, "optim/total_tokens": 1521483776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 340643, "loss/out": 6.122960090637207, "created_at": "2025-01-14T21:11:14.019692+00:00"} {"global_step": 5805, "acc_step": 0, "speed/wps": 12957.992485429379, "speed/FLOPS": 203523078131294.28, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.814965724945068, "optim/lr": 0.0029984184935675446, "optim/total_tokens": 1521745920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440919, "loss/out": 6.174903869628906, "created_at": "2025-01-14T21:11:19.080862+00:00"} {"global_step": 5806, "acc_step": 0, "speed/wps": 12954.366591272363, "speed/FLOPS": 203466128481057.66, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7564986944198608, "optim/lr": 0.002998414557718974, "optim/total_tokens": 1522008064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 6.07882022857666, "created_at": "2025-01-14T21:11:24.147899+00:00"} {"global_step": 5807, "acc_step": 0, "speed/wps": 12958.040437446003, "speed/FLOPS": 203523831283608.34, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.091912269592285, "optim/lr": 0.002998410616981562, "optim/total_tokens": 1522270208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454182, "loss/out": 6.019843101501465, "created_at": "2025-01-14T21:11:29.212565+00:00"} {"global_step": 5808, "acc_step": 0, "speed/wps": 12959.140670536986, "speed/FLOPS": 203541111956180.94, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9589766263961792, "optim/lr": 0.0029984066713553224, "optim/total_tokens": 1522532352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436916, "loss/out": 6.138704299926758, "created_at": "2025-01-14T21:11:34.276593+00:00"} {"global_step": 5809, "acc_step": 0, "speed/wps": 12957.271350996261, "speed/FLOPS": 203511751724079.62, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.4410150051116943, "optim/lr": 0.0029984027208402675, "optim/total_tokens": 1522794496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327927, "loss/out": 6.065859794616699, "created_at": "2025-01-14T21:11:39.344343+00:00"} {"global_step": 5810, "acc_step": 0, "speed/wps": 12954.189192953349, "speed/FLOPS": 203463342196686.7, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 10.205718994140625, "optim/lr": 0.0029983987654364098, "optim/total_tokens": 1523056640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439054, "loss/out": 6.0875020027160645, "created_at": "2025-01-14T21:11:44.412278+00:00"} {"global_step": 5811, "acc_step": 0, "speed/wps": 12959.439467750724, "speed/FLOPS": 203545804977013.1, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 9.648015975952148, "optim/lr": 0.0029983948051437626, "optim/total_tokens": 1523318784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447026, "loss/out": 5.978852272033691, "created_at": "2025-01-14T21:11:49.478480+00:00"} {"global_step": 5812, "acc_step": 0, "speed/wps": 12954.456076864968, "speed/FLOPS": 203467533975255.62, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.2055270671844482, "optim/lr": 0.0029983908399623393, "optim/total_tokens": 1523580928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 6.233035087585449, "created_at": "2025-01-14T21:11:54.545207+00:00"} {"global_step": 5813, "acc_step": 0, "speed/wps": 12959.904613293695, "speed/FLOPS": 203553110726942.4, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0013, "optim/grad_norm": 9.883472442626953, "optim/lr": 0.002998386869892152, "optim/total_tokens": 1523843072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450746, "loss/out": 6.073611259460449, "created_at": "2025-01-14T21:11:59.607346+00:00"} {"global_step": 5814, "acc_step": 0, "speed/wps": 12959.738231007806, "speed/FLOPS": 203550497464509.3, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.157683849334717, "optim/lr": 0.0029983828949332143, "optim/total_tokens": 1524105216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448606, "loss/out": 6.154789447784424, "created_at": "2025-01-14T21:12:04.669106+00:00"} {"global_step": 5815, "acc_step": 0, "speed/wps": 12961.44648470616, "speed/FLOPS": 203577327936228.53, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.2084243297576904, "optim/lr": 0.002998378915085539, "optim/total_tokens": 1524367360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 286083, "loss/out": 6.162595748901367, "created_at": "2025-01-14T21:12:09.732176+00:00"} {"global_step": 5816, "acc_step": 0, "speed/wps": 12955.445752194626, "speed/FLOPS": 203483078186264.3, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.749258041381836, "optim/lr": 0.0029983749303491383, "optim/total_tokens": 1524629504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455247, "loss/out": 5.945088863372803, "created_at": "2025-01-14T21:12:14.796846+00:00"} {"global_step": 5817, "acc_step": 0, "speed/wps": 12960.359881353255, "speed/FLOPS": 203560261337421.7, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.484287261962891, "optim/lr": 0.0029983709407240262, "optim/total_tokens": 1524891648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 5.964615821838379, "created_at": "2025-01-14T21:12:19.863553+00:00"} {"global_step": 5818, "acc_step": 0, "speed/wps": 12958.904837220216, "speed/FLOPS": 203537407869872.12, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.134994983673096, "optim/lr": 0.0029983669462102154, "optim/total_tokens": 1525153792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447269, "loss/out": 6.068530559539795, "created_at": "2025-01-14T21:12:24.930724+00:00"} {"global_step": 5819, "acc_step": 0, "speed/wps": 12959.422472285756, "speed/FLOPS": 203545538039881.56, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.09110426902771, "optim/lr": 0.0029983629468077187, "optim/total_tokens": 1525415936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443022, "loss/out": 5.956892013549805, "created_at": "2025-01-14T21:12:29.992277+00:00"} {"global_step": 5820, "acc_step": 0, "speed/wps": 12955.421788127982, "speed/FLOPS": 203482701797667.6, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.5818769931793213, "optim/lr": 0.0029983589425165496, "optim/total_tokens": 1525678080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294022, "loss/out": 5.996443748474121, "created_at": "2025-01-14T21:12:35.056221+00:00"} {"global_step": 5821, "acc_step": 0, "speed/wps": 12961.820777056953, "speed/FLOPS": 203583206711929.3, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.569845199584961, "optim/lr": 0.0029983549333367212, "optim/total_tokens": 1525940224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 404614, "loss/out": 6.009415626525879, "created_at": "2025-01-14T21:12:40.115972+00:00"} {"global_step": 5822, "acc_step": 0, "speed/wps": 12962.305123717239, "speed/FLOPS": 203590814041791.16, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.1655702590942383, "optim/lr": 0.0029983509192682465, "optim/total_tokens": 1526202368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 6.028148651123047, "created_at": "2025-01-14T21:12:45.175428+00:00"} {"global_step": 5823, "acc_step": 0, "speed/wps": 12958.878674352936, "speed/FLOPS": 203536996946089.97, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.851564407348633, "optim/lr": 0.0029983469003111375, "optim/total_tokens": 1526464512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 6.092496395111084, "created_at": "2025-01-14T21:12:50.236192+00:00"} {"global_step": 5824, "acc_step": 0, "speed/wps": 12958.568016575482, "speed/FLOPS": 203532117638806.5, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.271368026733398, "optim/lr": 0.0029983428764654083, "optim/total_tokens": 1526726656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415046, "loss/out": 6.039294242858887, "created_at": "2025-01-14T21:12:55.298555+00:00"} {"global_step": 5825, "acc_step": 0, "speed/wps": 12957.983169951396, "speed/FLOPS": 203522931818911.78, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.9239392280578613, "optim/lr": 0.0029983388477310725, "optim/total_tokens": 1526988800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484902, "loss/out": 6.055367946624756, "created_at": "2025-01-14T21:13:00.364735+00:00"} {"global_step": 5826, "acc_step": 0, "speed/wps": 12964.28424986968, "speed/FLOPS": 203621898937619.56, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.317581176757812, "optim/lr": 0.002998334814108143, "optim/total_tokens": 1527250944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450990, "loss/out": 6.175993919372559, "created_at": "2025-01-14T21:13:05.426769+00:00"} {"global_step": 5827, "acc_step": 0, "speed/wps": 12957.165707790753, "speed/FLOPS": 203510092452368.16, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.710976600646973, "optim/lr": 0.0029983307755966316, "optim/total_tokens": 1527513088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441971, "loss/out": 5.973531723022461, "created_at": "2025-01-14T21:13:10.488444+00:00"} {"global_step": 5828, "acc_step": 0, "speed/wps": 12954.133061233733, "speed/FLOPS": 203462460570897.06, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.089921951293945, "optim/lr": 0.0029983267321965533, "optim/total_tokens": 1527775232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304748, "loss/out": 6.089879035949707, "created_at": "2025-01-14T21:13:15.551649+00:00"} {"global_step": 5829, "acc_step": 0, "speed/wps": 12955.968373124739, "speed/FLOPS": 203491286666128.16, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.758056640625, "optim/lr": 0.0029983226839079198, "optim/total_tokens": 1528037376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 398670, "loss/out": 6.196427345275879, "created_at": "2025-01-14T21:13:20.619019+00:00"} {"global_step": 5830, "acc_step": 0, "speed/wps": 12958.699224224962, "speed/FLOPS": 203534178435239.78, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.925446510314941, "optim/lr": 0.0029983186307307457, "optim/total_tokens": 1528299520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 6.118638515472412, "created_at": "2025-01-14T21:13:25.683433+00:00"} {"global_step": 5831, "acc_step": 0, "speed/wps": 12958.843015892848, "speed/FLOPS": 203536436880974.4, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.7217187881469727, "optim/lr": 0.0029983145726650436, "optim/total_tokens": 1528561664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427570, "loss/out": 5.986257553100586, "created_at": "2025-01-14T21:13:30.749790+00:00"} {"global_step": 5832, "acc_step": 0, "speed/wps": 12958.536132625435, "speed/FLOPS": 203531616857557.88, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.996073007583618, "optim/lr": 0.002998310509710826, "optim/total_tokens": 1528823808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 6.011847972869873, "created_at": "2025-01-14T21:13:35.816356+00:00"} {"global_step": 5833, "acc_step": 0, "speed/wps": 12961.206583017016, "speed/FLOPS": 203573559950541.06, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 15.129539489746094, "optim/lr": 0.0029983064418681077, "optim/total_tokens": 1529085952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327253, "loss/out": 6.074552536010742, "created_at": "2025-01-14T21:13:40.879330+00:00"} {"global_step": 5834, "acc_step": 0, "speed/wps": 12964.436126415807, "speed/FLOPS": 203624284367476.75, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.52267599105835, "optim/lr": 0.0029983023691369007, "optim/total_tokens": 1529348096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352687, "loss/out": 6.087474822998047, "created_at": "2025-01-14T21:13:45.939422+00:00"} {"global_step": 5835, "acc_step": 0, "speed/wps": 12962.372161888668, "speed/FLOPS": 203591866968393.56, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.448755264282227, "optim/lr": 0.0029982982915172187, "optim/total_tokens": 1529610240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443545, "loss/out": 6.0394086837768555, "created_at": "2025-01-14T21:13:51.004923+00:00"} {"global_step": 5836, "acc_step": 0, "speed/wps": 12955.062108800805, "speed/FLOPS": 203477052539583.06, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.850107192993164, "optim/lr": 0.002998294209009075, "optim/total_tokens": 1529872384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 5.981204509735107, "created_at": "2025-01-14T21:13:56.076297+00:00"} {"global_step": 5837, "acc_step": 0, "speed/wps": 12961.823085816877, "speed/FLOPS": 203583242974176.5, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 109.93497467041016, "optim/lr": 0.002998290121612483, "optim/total_tokens": 1530134528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 6.010125637054443, "created_at": "2025-01-14T21:14:01.140123+00:00"} {"global_step": 5838, "acc_step": 0, "speed/wps": 12958.951507283893, "speed/FLOPS": 203538140887353.56, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.001, "optim/grad_norm": 3.593080520629883, "optim/lr": 0.002998286029327456, "optim/total_tokens": 1530396672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434230, "loss/out": 6.1066999435424805, "created_at": "2025-01-14T21:14:06.201004+00:00"} {"global_step": 5839, "acc_step": 0, "speed/wps": 12966.23727182041, "speed/FLOPS": 203652573831089.2, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.462236166000366, "optim/lr": 0.0029982819321540074, "optim/total_tokens": 1530658816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311191, "loss/out": 5.997921943664551, "created_at": "2025-01-14T21:14:11.259482+00:00"} {"global_step": 5840, "acc_step": 0, "speed/wps": 12959.961668634218, "speed/FLOPS": 203554006859467.56, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.860337972640991, "optim/lr": 0.00299827783009215, "optim/total_tokens": 1530920960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295655, "loss/out": 5.986821174621582, "created_at": "2025-01-14T21:14:16.322884+00:00"} {"global_step": 5841, "acc_step": 0, "speed/wps": 12959.47209840468, "speed/FLOPS": 203546317486272.7, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0009, "optim/grad_norm": 9.128280639648438, "optim/lr": 0.002998273723141898, "optim/total_tokens": 1531183104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368957, "loss/out": 5.966342926025391, "created_at": "2025-01-14T21:14:21.389182+00:00"} {"global_step": 5842, "acc_step": 0, "speed/wps": 12958.06874925063, "speed/FLOPS": 203524275959404.5, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 63.588932037353516, "optim/lr": 0.0029982696113032645, "optim/total_tokens": 1531445248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 6.130102157592773, "created_at": "2025-01-14T21:14:26.450531+00:00"} {"global_step": 5843, "acc_step": 0, "speed/wps": 12965.598024369468, "speed/FLOPS": 203642533571454.1, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.79438304901123, "optim/lr": 0.002998265494576263, "optim/total_tokens": 1531707392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458410, "loss/out": 6.169461250305176, "created_at": "2025-01-14T21:14:31.508759+00:00"} {"global_step": 5844, "acc_step": 0, "speed/wps": 12960.814073749556, "speed/FLOPS": 203567395053131.34, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.035770416259766, "optim/lr": 0.002998261372960907, "optim/total_tokens": 1531969536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445933, "loss/out": 6.11458158493042, "created_at": "2025-01-14T21:14:36.573014+00:00"} {"global_step": 5845, "acc_step": 0, "speed/wps": 12955.596703762023, "speed/FLOPS": 203485449088059.47, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.998108386993408, "optim/lr": 0.002998257246457209, "optim/total_tokens": 1532231680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 6.046998977661133, "created_at": "2025-01-14T21:14:41.641925+00:00"} {"global_step": 5846, "acc_step": 0, "speed/wps": 12962.532267396546, "speed/FLOPS": 203594381645402.8, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.481900691986084, "optim/lr": 0.0029982531150651837, "optim/total_tokens": 1532493824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329272, "loss/out": 5.955691337585449, "created_at": "2025-01-14T21:14:46.705400+00:00"} {"global_step": 5847, "acc_step": 0, "speed/wps": 12962.602803709493, "speed/FLOPS": 203595489515124.7, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.977297782897949, "optim/lr": 0.0029982489787848446, "optim/total_tokens": 1532755968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379212, "loss/out": 6.055044174194336, "created_at": "2025-01-14T21:14:51.765012+00:00"} {"global_step": 5848, "acc_step": 0, "speed/wps": 12958.78158175683, "speed/FLOPS": 203535471973448.9, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.565166711807251, "optim/lr": 0.002998244837616204, "optim/total_tokens": 1533018112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452615, "loss/out": 6.0627007484436035, "created_at": "2025-01-14T21:14:56.825715+00:00"} {"global_step": 5849, "acc_step": 0, "speed/wps": 12955.631526211426, "speed/FLOPS": 203485996022478.4, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.2900285720825195, "optim/lr": 0.0029982406915592765, "optim/total_tokens": 1533280256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457665, "loss/out": 5.94227409362793, "created_at": "2025-01-14T21:15:01.891110+00:00"} {"global_step": 5850, "acc_step": 0, "speed/wps": 12953.555641980935, "speed/FLOPS": 203453391408074.66, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0008, "optim/grad_norm": 18.441776275634766, "optim/lr": 0.002998236540614075, "optim/total_tokens": 1533542400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 5.9199628829956055, "created_at": "2025-01-14T21:15:06.959279+00:00"} {"global_step": 5851, "acc_step": 0, "speed/wps": 12960.038492687112, "speed/FLOPS": 203555213486785.6, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.663928508758545, "optim/lr": 0.002998232384780613, "optim/total_tokens": 1533804544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431320, "loss/out": 6.131770133972168, "created_at": "2025-01-14T21:15:12.024726+00:00"} {"global_step": 5852, "acc_step": 0, "speed/wps": 12950.456019802348, "speed/FLOPS": 203404707582432.75, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.001, "optim/grad_norm": 104.35984802246094, "optim/lr": 0.002998228224058905, "optim/total_tokens": 1534066688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349477, "loss/out": 6.0311079025268555, "created_at": "2025-01-14T21:15:17.097083+00:00"} {"global_step": 5853, "acc_step": 0, "speed/wps": 12958.241691367864, "speed/FLOPS": 203526992253003.06, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.383725643157959, "optim/lr": 0.0029982240584489634, "optim/total_tokens": 1534328832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439054, "loss/out": 6.057157516479492, "created_at": "2025-01-14T21:15:22.159301+00:00"} {"global_step": 5854, "acc_step": 0, "speed/wps": 12955.841507235258, "speed/FLOPS": 203489294063001.84, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0009, "optim/grad_norm": 2.866806983947754, "optim/lr": 0.0029982198879508026, "optim/total_tokens": 1534590976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451797, "loss/out": 6.071717262268066, "created_at": "2025-01-14T21:15:27.221881+00:00"} {"global_step": 5855, "acc_step": 0, "speed/wps": 12958.627141064238, "speed/FLOPS": 203533046270148.0, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 51.93937683105469, "optim/lr": 0.002998215712564436, "optim/total_tokens": 1534853120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 5.990741729736328, "created_at": "2025-01-14T21:15:32.285153+00:00"} {"global_step": 5856, "acc_step": 0, "speed/wps": 12957.81398563984, "speed/FLOPS": 203520274546815.16, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.875727653503418, "optim/lr": 0.0029982115322898764, "optim/total_tokens": 1535115264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427855, "loss/out": 6.0700483322143555, "created_at": "2025-01-14T21:15:37.349638+00:00"} {"global_step": 5857, "acc_step": 0, "speed/wps": 12947.315160396645, "speed/FLOPS": 203355376069473.1, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0016, "optim/grad_norm": 3.3540873527526855, "optim/lr": 0.0029982073471271392, "optim/total_tokens": 1535377408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450990, "loss/out": 6.139509677886963, "created_at": "2025-01-14T21:15:42.419434+00:00"} {"global_step": 5858, "acc_step": 0, "speed/wps": 12954.207715239812, "speed/FLOPS": 203463633114647.9, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.426168441772461, "optim/lr": 0.0029982031570762366, "optim/total_tokens": 1535639552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302630, "loss/out": 6.057074069976807, "created_at": "2025-01-14T21:15:47.488521+00:00"} {"global_step": 5859, "acc_step": 0, "speed/wps": 12961.458557853473, "speed/FLOPS": 203577517561597.3, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.548751354217529, "optim/lr": 0.002998198962137183, "optim/total_tokens": 1535901696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 6.195340156555176, "created_at": "2025-01-14T21:15:52.549314+00:00"} {"global_step": 5860, "acc_step": 0, "speed/wps": 12954.871807707308, "speed/FLOPS": 203474063599408.6, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.993270397186279, "optim/lr": 0.0029981947623099916, "optim/total_tokens": 1536163840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 5.991351127624512, "created_at": "2025-01-14T21:15:57.613745+00:00"} {"global_step": 5861, "acc_step": 0, "speed/wps": 12953.14150966981, "speed/FLOPS": 203446886891051.03, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.401755332946777, "optim/lr": 0.002998190557594676, "optim/total_tokens": 1536425984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431076, "loss/out": 5.995139122009277, "created_at": "2025-01-14T21:16:02.679968+00:00"} {"global_step": 5862, "acc_step": 0, "speed/wps": 12951.374616645193, "speed/FLOPS": 203419135408133.97, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.010179042816162, "optim/lr": 0.0029981863479912503, "optim/total_tokens": 1536688128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 6.064064979553223, "created_at": "2025-01-14T21:16:07.746066+00:00"} {"global_step": 5863, "acc_step": 0, "speed/wps": 12955.901756594294, "speed/FLOPS": 203490240361976.8, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.41105842590332, "optim/lr": 0.002998182133499728, "optim/total_tokens": 1536950272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425201, "loss/out": 6.03298282623291, "created_at": "2025-01-14T21:16:12.815424+00:00"} {"global_step": 5864, "acc_step": 0, "speed/wps": 12959.848124132313, "speed/FLOPS": 203552223487038.66, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.472042083740234, "optim/lr": 0.0029981779141201237, "optim/total_tokens": 1537212416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357486, "loss/out": 6.099160194396973, "created_at": "2025-01-14T21:16:17.877328+00:00"} {"global_step": 5865, "acc_step": 0, "speed/wps": 12945.330147104192, "speed/FLOPS": 203324198708021.0, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.020528793334961, "optim/lr": 0.00299817368985245, "optim/total_tokens": 1537474560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 6.079888343811035, "created_at": "2025-01-14T21:16:22.943754+00:00"} {"global_step": 5866, "acc_step": 0, "speed/wps": 12957.522526250525, "speed/FLOPS": 203515696776598.8, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.502669811248779, "optim/lr": 0.002998169460696721, "optim/total_tokens": 1537736704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472789, "loss/out": 6.139931678771973, "created_at": "2025-01-14T21:16:28.008075+00:00"} {"global_step": 5867, "acc_step": 0, "speed/wps": 12958.519058394411, "speed/FLOPS": 203531348683297.22, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 195.22915649414062, "optim/lr": 0.0029981652266529513, "optim/total_tokens": 1537998848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 6.135134220123291, "created_at": "2025-01-14T21:16:33.068905+00:00"} {"global_step": 5868, "acc_step": 0, "speed/wps": 12959.122389502636, "speed/FLOPS": 203540824827415.7, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.221794605255127, "optim/lr": 0.0029981609877211537, "optim/total_tokens": 1538260992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462150, "loss/out": 6.114076614379883, "created_at": "2025-01-14T21:16:38.129897+00:00"} {"global_step": 5869, "acc_step": 0, "speed/wps": 12950.17829953522, "speed/FLOPS": 203400345604009.88, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.940917015075684, "optim/lr": 0.0029981567439013417, "optim/total_tokens": 1538523136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440105, "loss/out": 6.014534950256348, "created_at": "2025-01-14T21:16:43.194702+00:00"} {"global_step": 5870, "acc_step": 0, "speed/wps": 12954.810873901695, "speed/FLOPS": 203473106550259.62, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0009, "optim/grad_norm": 10.10062026977539, "optim/lr": 0.0029981524951935307, "optim/total_tokens": 1538785280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460840, "loss/out": 6.036738395690918, "created_at": "2025-01-14T21:16:48.288622+00:00"} {"global_step": 5871, "acc_step": 0, "speed/wps": 12955.454119483924, "speed/FLOPS": 203483209606040.6, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.903599739074707, "optim/lr": 0.002998148241597733, "optim/total_tokens": 1539047424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440390, "loss/out": 6.056207656860352, "created_at": "2025-01-14T21:16:53.355014+00:00"} {"global_step": 5872, "acc_step": 0, "speed/wps": 12958.70679865913, "speed/FLOPS": 203534297402136.78, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.30917501449585, "optim/lr": 0.002998143983113964, "optim/total_tokens": 1539309568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 6.216392517089844, "created_at": "2025-01-14T21:16:58.421211+00:00"} {"global_step": 5873, "acc_step": 0, "speed/wps": 12952.26090187696, "speed/FLOPS": 203433055735582.4, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0011, "optim/grad_norm": 3.465390920639038, "optim/lr": 0.0029981397197422367, "optim/total_tokens": 1539571712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453375, "loss/out": 6.039393424987793, "created_at": "2025-01-14T21:17:03.485774+00:00"} {"global_step": 5874, "acc_step": 0, "speed/wps": 12953.443544998852, "speed/FLOPS": 203451630770933.7, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.818896532058716, "optim/lr": 0.0029981354514825643, "optim/total_tokens": 1539833856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462150, "loss/out": 6.018190383911133, "created_at": "2025-01-14T21:17:08.551487+00:00"} {"global_step": 5875, "acc_step": 0, "speed/wps": 12950.058596566905, "speed/FLOPS": 203398465504403.38, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.788865566253662, "optim/lr": 0.0029981311783349616, "optim/total_tokens": 1540096000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300816, "loss/out": 6.000751495361328, "created_at": "2025-01-14T21:17:13.620687+00:00"} {"global_step": 5876, "acc_step": 0, "speed/wps": 12949.374053547137, "speed/FLOPS": 203387713815617.22, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 130.75726318359375, "optim/lr": 0.0029981269002994433, "optim/total_tokens": 1540358144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434003, "loss/out": 6.136586666107178, "created_at": "2025-01-14T21:17:18.686678+00:00"} {"global_step": 5877, "acc_step": 0, "speed/wps": 12958.12840846216, "speed/FLOPS": 203525212989301.94, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.698034286499023, "optim/lr": 0.002998122617376022, "optim/total_tokens": 1540620288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 6.125812530517578, "created_at": "2025-01-14T21:17:23.748792+00:00"} {"global_step": 5878, "acc_step": 0, "speed/wps": 12958.743148429874, "speed/FLOPS": 203534868325236.12, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.780034065246582, "optim/lr": 0.0029981183295647116, "optim/total_tokens": 1540882432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435582, "loss/out": 6.040634632110596, "created_at": "2025-01-14T21:17:28.814115+00:00"} {"global_step": 5879, "acc_step": 0, "speed/wps": 12949.075563022518, "speed/FLOPS": 203383025611759.12, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.791258811950684, "optim/lr": 0.002998114036865528, "optim/total_tokens": 1541144576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 5.94255256652832, "created_at": "2025-01-14T21:17:33.883047+00:00"} {"global_step": 5880, "acc_step": 0, "speed/wps": 12956.999266990671, "speed/FLOPS": 203507478270889.25, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 17.380508422851562, "optim/lr": 0.0029981097392784827, "optim/total_tokens": 1541406720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 6.150160789489746, "created_at": "2025-01-14T21:17:38.952538+00:00"} {"global_step": 5881, "acc_step": 0, "speed/wps": 12955.917616929299, "speed/FLOPS": 203490489470332.44, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.001, "optim/grad_norm": 3.2350831031799316, "optim/lr": 0.0029981054368035916, "optim/total_tokens": 1541668864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326058, "loss/out": 6.050606727600098, "created_at": "2025-01-14T21:17:44.020398+00:00"} {"global_step": 5882, "acc_step": 0, "speed/wps": 12953.391294461677, "speed/FLOPS": 203450810104447.56, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.9403647184371948, "optim/lr": 0.0029981011294408673, "optim/total_tokens": 1541931008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 403804, "loss/out": 6.01773738861084, "created_at": "2025-01-14T21:17:49.086788+00:00"} {"global_step": 5883, "acc_step": 0, "speed/wps": 12952.115897376067, "speed/FLOPS": 203430778240638.8, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0007, "optim/grad_norm": 17.277359008789062, "optim/lr": 0.0029980968171903254, "optim/total_tokens": 1542193152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 5.923169136047363, "created_at": "2025-01-14T21:17:54.153857+00:00"} {"global_step": 5884, "acc_step": 0, "speed/wps": 12947.694118808582, "speed/FLOPS": 203361328132075.97, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 20.618762969970703, "optim/lr": 0.0029980925000519785, "optim/total_tokens": 1542455296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445933, "loss/out": 6.105639934539795, "created_at": "2025-01-14T21:17:59.219236+00:00"} {"global_step": 5885, "acc_step": 0, "speed/wps": 12947.801299115481, "speed/FLOPS": 203363011546077.0, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.394985198974609, "optim/lr": 0.0029980881780258418, "optim/total_tokens": 1542717440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447269, "loss/out": 6.1679229736328125, "created_at": "2025-01-14T21:18:04.285224+00:00"} {"global_step": 5886, "acc_step": 0, "speed/wps": 12950.883852488301, "speed/FLOPS": 203411427282669.38, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.398842811584473, "optim/lr": 0.002998083851111929, "optim/total_tokens": 1542979584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289551, "loss/out": 5.958740711212158, "created_at": "2025-01-14T21:18:09.358438+00:00"} {"global_step": 5887, "acc_step": 0, "speed/wps": 12954.11946358982, "speed/FLOPS": 203462247001214.28, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 15.391687393188477, "optim/lr": 0.002998079519310254, "optim/total_tokens": 1543241728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446747, "loss/out": 6.069908142089844, "created_at": "2025-01-14T21:18:14.422222+00:00"} {"global_step": 5888, "acc_step": 0, "speed/wps": 12946.193159159237, "speed/FLOPS": 203337753498248.3, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 44.270782470703125, "optim/lr": 0.002998075182620831, "optim/total_tokens": 1543503872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 6.208806991577148, "created_at": "2025-01-14T21:18:19.489316+00:00"} {"global_step": 5889, "acc_step": 0, "speed/wps": 12952.076888515438, "speed/FLOPS": 203430165552878.84, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.8660669326782227, "optim/lr": 0.002998070841043674, "optim/total_tokens": 1543766016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457380, "loss/out": 6.175436019897461, "created_at": "2025-01-14T21:18:24.562335+00:00"} {"global_step": 5890, "acc_step": 0, "speed/wps": 12945.291376768537, "speed/FLOPS": 203323589766623.5, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 35.96688461303711, "optim/lr": 0.0029980664945787982, "optim/total_tokens": 1544028160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320196, "loss/out": 6.097257614135742, "created_at": "2025-01-14T21:18:29.629931+00:00"} {"global_step": 5891, "acc_step": 0, "speed/wps": 12951.639794601246, "speed/FLOPS": 203423300392326.47, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.605183601379395, "optim/lr": 0.0029980621432262166, "optim/total_tokens": 1544290304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362577, "loss/out": 6.2929463386535645, "created_at": "2025-01-14T21:18:34.697962+00:00"} {"global_step": 5892, "acc_step": 0, "speed/wps": 12955.567737275043, "speed/FLOPS": 203484994129577.3, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0014, "optim/grad_norm": 8.905879974365234, "optim/lr": 0.002998057786985943, "optim/total_tokens": 1544552448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445976, "loss/out": 6.273503303527832, "created_at": "2025-01-14T21:18:39.765013+00:00"} {"global_step": 5893, "acc_step": 0, "speed/wps": 12957.659613343865, "speed/FLOPS": 203517849917777.12, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.724081039428711, "optim/lr": 0.002998053425857994, "optim/total_tokens": 1544814592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 6.198958873748779, "created_at": "2025-01-14T21:18:44.827460+00:00"} {"global_step": 5894, "acc_step": 0, "speed/wps": 12960.310188760077, "speed/FLOPS": 203559480846961.03, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 11.469170570373535, "optim/lr": 0.0029980490598423806, "optim/total_tokens": 1545076736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457902, "loss/out": 6.193099021911621, "created_at": "2025-01-14T21:18:49.894137+00:00"} {"global_step": 5895, "acc_step": 0, "speed/wps": 12962.354748038548, "speed/FLOPS": 203591593459948.28, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.686781644821167, "optim/lr": 0.0029980446889391194, "optim/total_tokens": 1545338880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468777, "loss/out": 6.174304962158203, "created_at": "2025-01-14T21:18:54.955567+00:00"} {"global_step": 5896, "acc_step": 0, "speed/wps": 12954.941413994698, "speed/FLOPS": 203475156861801.12, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.012653350830078, "optim/lr": 0.0029980403131482244, "optim/total_tokens": 1545601024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 6.171769142150879, "created_at": "2025-01-14T21:19:00.019157+00:00"} {"global_step": 5897, "acc_step": 0, "speed/wps": 12955.931215963059, "speed/FLOPS": 203490703061844.66, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 15.461962699890137, "optim/lr": 0.002998035932469709, "optim/total_tokens": 1545863168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301523, "loss/out": 6.073953628540039, "created_at": "2025-01-14T21:19:05.085455+00:00"} {"global_step": 5898, "acc_step": 0, "speed/wps": 12959.610801997842, "speed/FLOPS": 203548496016802.0, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.0386550426483154, "optim/lr": 0.002998031546903588, "optim/total_tokens": 1546125312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454466, "loss/out": 6.106707572937012, "created_at": "2025-01-14T21:19:10.150380+00:00"} {"global_step": 5899, "acc_step": 0, "speed/wps": 12955.360539076542, "speed/FLOPS": 203481739797149.72, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0009, "optim/grad_norm": 8.980440139770508, "optim/lr": 0.0029980271564498755, "optim/total_tokens": 1546387456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 6.01092529296875, "created_at": "2025-01-14T21:19:15.212769+00:00"} {"global_step": 5900, "acc_step": 0, "speed/wps": 12954.430423232188, "speed/FLOPS": 203467131049701.03, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.5510382652282715, "optim/lr": 0.0029980227611085854, "optim/total_tokens": 1546649600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440921, "loss/out": 6.256122589111328, "created_at": "2025-01-14T21:19:20.276260+00:00"} {"global_step": 5901, "acc_step": 0, "speed/wps": 12955.947973275108, "speed/FLOPS": 203490966258457.34, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.788717269897461, "optim/lr": 0.002998018360879733, "optim/total_tokens": 1546911744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470641, "loss/out": 6.159994125366211, "created_at": "2025-01-14T21:19:25.338412+00:00"} {"global_step": 5902, "acc_step": 0, "speed/wps": 12957.419112176827, "speed/FLOPS": 203514072516465.56, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 63.05831527709961, "optim/lr": 0.002998013955763333, "optim/total_tokens": 1547173888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455507, "loss/out": 6.162743091583252, "created_at": "2025-01-14T21:19:30.399796+00:00"} {"global_step": 5903, "acc_step": 0, "speed/wps": 12951.078934718273, "speed/FLOPS": 203414491317162.16, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.240612983703613, "optim/lr": 0.002998009545759398, "optim/total_tokens": 1547436032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307966, "loss/out": 6.191597938537598, "created_at": "2025-01-14T21:19:35.464334+00:00"} {"global_step": 5904, "acc_step": 0, "speed/wps": 12956.610166302913, "speed/FLOPS": 203501366909907.62, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.4056553840637207, "optim/lr": 0.002998005130867944, "optim/total_tokens": 1547698176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424340, "loss/out": 6.168301105499268, "created_at": "2025-01-14T21:19:40.530437+00:00"} {"global_step": 5905, "acc_step": 0, "speed/wps": 12951.474575569324, "speed/FLOPS": 203420705400395.4, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 22.07476806640625, "optim/lr": 0.0029980007110889844, "optim/total_tokens": 1547960320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 5.944414138793945, "created_at": "2025-01-14T21:19:45.595413+00:00"} {"global_step": 5906, "acc_step": 0, "speed/wps": 12954.557209020651, "speed/FLOPS": 203469122394730.75, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.47433853149414, "optim/lr": 0.0029979962864225343, "optim/total_tokens": 1548222464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463248, "loss/out": 6.126466274261475, "created_at": "2025-01-14T21:19:50.661652+00:00"} {"global_step": 5907, "acc_step": 0, "speed/wps": 12954.174462443527, "speed/FLOPS": 203463110833788.12, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.527406215667725, "optim/lr": 0.0029979918568686075, "optim/total_tokens": 1548484608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447791, "loss/out": 6.2322893142700195, "created_at": "2025-01-14T21:19:55.733163+00:00"} {"global_step": 5908, "acc_step": 0, "speed/wps": 12958.67586858215, "speed/FLOPS": 203533811602775.4, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.741390705108643, "optim/lr": 0.002997987422427219, "optim/total_tokens": 1548746752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 6.089388847351074, "created_at": "2025-01-14T21:20:00.799650+00:00"} {"global_step": 5909, "acc_step": 0, "speed/wps": 12958.383933729669, "speed/FLOPS": 203529226364756.7, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.357558727264404, "optim/lr": 0.0029979829830983832, "optim/total_tokens": 1549008896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323655, "loss/out": 6.069591522216797, "created_at": "2025-01-14T21:20:05.866525+00:00"} {"global_step": 5910, "acc_step": 0, "speed/wps": 12956.229640957468, "speed/FLOPS": 203495390236460.78, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.90824556350708, "optim/lr": 0.002997978538882114, "optim/total_tokens": 1549271040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360416, "loss/out": 5.996759414672852, "created_at": "2025-01-14T21:20:10.928834+00:00"} {"global_step": 5911, "acc_step": 0, "speed/wps": 12951.167507256057, "speed/FLOPS": 203415882470579.47, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0008, "optim/grad_norm": 10.54887866973877, "optim/lr": 0.002997974089778427, "optim/total_tokens": 1549533184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441442, "loss/out": 6.075428009033203, "created_at": "2025-01-14T21:20:16.019778+00:00"} {"global_step": 5912, "acc_step": 0, "speed/wps": 12957.720070496882, "speed/FLOPS": 203518799480442.1, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 15.337160110473633, "optim/lr": 0.0029979696357873356, "optim/total_tokens": 1549795328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474937, "loss/out": 6.179745197296143, "created_at": "2025-01-14T21:20:21.081017+00:00"} {"global_step": 5913, "acc_step": 0, "speed/wps": 12946.387795220764, "speed/FLOPS": 203340810525052.56, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 10.077406883239746, "optim/lr": 0.002997965176908855, "optim/total_tokens": 1550057472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432365, "loss/out": 6.420657634735107, "created_at": "2025-01-14T21:20:26.146723+00:00"} {"global_step": 5914, "acc_step": 0, "speed/wps": 12954.866272149455, "speed/FLOPS": 203473976655865.88, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.131970405578613, "optim/lr": 0.002997960713143, "optim/total_tokens": 1550319616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298942, "loss/out": 6.124704837799072, "created_at": "2025-01-14T21:20:31.213647+00:00"} {"global_step": 5915, "acc_step": 0, "speed/wps": 12952.067738882093, "speed/FLOPS": 203430021845314.25, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.938869595527649, "optim/lr": 0.0029979562444897843, "optim/total_tokens": 1550581760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423627, "loss/out": 6.14274787902832, "created_at": "2025-01-14T21:20:36.280340+00:00"} {"global_step": 5916, "acc_step": 0, "speed/wps": 12958.511162025217, "speed/FLOPS": 203531224659968.28, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.335111141204834, "optim/lr": 0.002997951770949223, "optim/total_tokens": 1550843904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 6.0952348709106445, "created_at": "2025-01-14T21:20:41.346989+00:00"} {"global_step": 5917, "acc_step": 0, "speed/wps": 12959.229250754484, "speed/FLOPS": 203542503230218.72, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.958301544189453, "optim/lr": 0.0029979472925213304, "optim/total_tokens": 1551106048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 6.030432224273682, "created_at": "2025-01-14T21:20:46.413097+00:00"} {"global_step": 5918, "acc_step": 0, "speed/wps": 12956.865323937658, "speed/FLOPS": 203505374511184.62, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.20432186126709, "optim/lr": 0.0029979428092061213, "optim/total_tokens": 1551368192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423627, "loss/out": 6.1616530418396, "created_at": "2025-01-14T21:20:51.479660+00:00"} {"global_step": 5919, "acc_step": 0, "speed/wps": 12959.51392994052, "speed/FLOPS": 203546974508025.62, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.5337326526641846, "optim/lr": 0.002997938321003611, "optim/total_tokens": 1551630336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329951, "loss/out": 6.059474945068359, "created_at": "2025-01-14T21:20:56.541551+00:00"} {"global_step": 5920, "acc_step": 0, "speed/wps": 12950.902422528487, "speed/FLOPS": 203411718950668.38, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 15.796985626220703, "optim/lr": 0.002997933827913813, "optim/total_tokens": 1551892480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356429, "loss/out": 6.062258720397949, "created_at": "2025-01-14T21:21:01.608127+00:00"} {"global_step": 5921, "acc_step": 0, "speed/wps": 12952.284272546452, "speed/FLOPS": 203433422804061.44, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.0599467754364014, "optim/lr": 0.0029979293299367423, "optim/total_tokens": 1552154624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432431, "loss/out": 6.1421051025390625, "created_at": "2025-01-14T21:21:06.672072+00:00"} {"global_step": 5922, "acc_step": 0, "speed/wps": 12953.876248962204, "speed/FLOPS": 203458426981275.66, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 19.249746322631836, "optim/lr": 0.0029979248270724135, "optim/total_tokens": 1552416768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 6.0805768966674805, "created_at": "2025-01-14T21:21:11.740852+00:00"} {"global_step": 5923, "acc_step": 0, "speed/wps": 12956.04811015047, "speed/FLOPS": 203492539045687.4, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.485456466674805, "optim/lr": 0.0029979203193208417, "optim/total_tokens": 1552678912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435852, "loss/out": 6.144537925720215, "created_at": "2025-01-14T21:21:16.808280+00:00"} {"global_step": 5924, "acc_step": 0, "speed/wps": 12957.426613536702, "speed/FLOPS": 203514190335630.34, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.3306403160095215, "optim/lr": 0.0029979158066820413, "optim/total_tokens": 1552941056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 6.063068866729736, "created_at": "2025-01-14T21:21:21.869674+00:00"} {"global_step": 5925, "acc_step": 0, "speed/wps": 12959.09248604197, "speed/FLOPS": 203540355152474.25, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 94.70063781738281, "optim/lr": 0.0029979112891560275, "optim/total_tokens": 1553203200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433993, "loss/out": 6.040953636169434, "created_at": "2025-01-14T21:21:26.934217+00:00"} {"global_step": 5926, "acc_step": 0, "speed/wps": 12959.11823689889, "speed/FLOPS": 203540759605067.6, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.689505577087402, "optim/lr": 0.002997906766742815, "optim/total_tokens": 1553465344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352115, "loss/out": 6.117292404174805, "created_at": "2025-01-14T21:21:31.994939+00:00"} {"global_step": 5927, "acc_step": 0, "speed/wps": 12943.67293675664, "speed/FLOPS": 203298169942267.03, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3118555545806885, "optim/lr": 0.0029979022394424175, "optim/total_tokens": 1553727488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456567, "loss/out": 6.155953884124756, "created_at": "2025-01-14T21:21:37.065798+00:00"} {"global_step": 5928, "acc_step": 0, "speed/wps": 12956.299932166621, "speed/FLOPS": 203496494256490.78, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.478015661239624, "optim/lr": 0.0029978977072548513, "optim/total_tokens": 1553989632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449696, "loss/out": 6.05579137802124, "created_at": "2025-01-14T21:21:42.141763+00:00"} {"global_step": 5929, "acc_step": 0, "speed/wps": 12961.379740208056, "speed/FLOPS": 203576279622169.25, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.0313940048217773, "optim/lr": 0.0029978931701801293, "optim/total_tokens": 1554251776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460578, "loss/out": 6.034397125244141, "created_at": "2025-01-14T21:21:47.202490+00:00"} {"global_step": 5930, "acc_step": 0, "speed/wps": 12956.548164090755, "speed/FLOPS": 203500393079965.7, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 11.07395076751709, "optim/lr": 0.002997888628218268, "optim/total_tokens": 1554513920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316000, "loss/out": 5.938982963562012, "created_at": "2025-01-14T21:21:52.265648+00:00"} {"global_step": 5931, "acc_step": 0, "speed/wps": 12956.077031076897, "speed/FLOPS": 203492993288578.47, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.8118064403533936, "optim/lr": 0.0029978840813692817, "optim/total_tokens": 1554776064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376202, "loss/out": 6.155966758728027, "created_at": "2025-01-14T21:21:57.328577+00:00"} {"global_step": 5932, "acc_step": 0, "speed/wps": 12957.125134424608, "speed/FLOPS": 203509455191899.2, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.777721405029297, "optim/lr": 0.002997879529633185, "optim/total_tokens": 1555038208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439871, "loss/out": 6.039168357849121, "created_at": "2025-01-14T21:22:02.390559+00:00"} {"global_step": 5933, "acc_step": 0, "speed/wps": 12952.244847851729, "speed/FLOPS": 203432803585055.62, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0012, "optim/grad_norm": 2.5467982292175293, "optim/lr": 0.002997874973009993, "optim/total_tokens": 1555300352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443016, "loss/out": 5.9926276206970215, "created_at": "2025-01-14T21:22:07.461857+00:00"} {"global_step": 5934, "acc_step": 0, "speed/wps": 12962.396314042622, "speed/FLOPS": 203592246311159.94, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.8410284519195557, "optim/lr": 0.0029978704114997203, "optim/total_tokens": 1555562496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 6.037797451019287, "created_at": "2025-01-14T21:22:12.522667+00:00"} {"global_step": 5935, "acc_step": 0, "speed/wps": 12959.596068702287, "speed/FLOPS": 203548264610149.66, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 11.666746139526367, "optim/lr": 0.0029978658451023823, "optim/total_tokens": 1555824640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 6.061505317687988, "created_at": "2025-01-14T21:22:17.588734+00:00"} {"global_step": 5936, "acc_step": 0, "speed/wps": 12956.100253309813, "speed/FLOPS": 203493358025657.1, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.27450704574585, "optim/lr": 0.0029978612738179927, "optim/total_tokens": 1556086784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 6.049206733703613, "created_at": "2025-01-14T21:22:22.655520+00:00"} {"global_step": 5937, "acc_step": 0, "speed/wps": 12964.501447776614, "speed/FLOPS": 203625310329209.2, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.24357008934021, "optim/lr": 0.0029978566976465683, "optim/total_tokens": 1556348928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327927, "loss/out": 6.075274467468262, "created_at": "2025-01-14T21:22:27.719730+00:00"} {"global_step": 5938, "acc_step": 0, "speed/wps": 12961.053168336406, "speed/FLOPS": 203571150362168.22, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.983400344848633, "optim/lr": 0.0029978521165881225, "optim/total_tokens": 1556611072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 384894, "loss/out": 6.109990119934082, "created_at": "2025-01-14T21:22:32.782217+00:00"} {"global_step": 5939, "acc_step": 0, "speed/wps": 12950.431356801657, "speed/FLOPS": 203404320216116.38, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.181717872619629, "optim/lr": 0.0029978475306426705, "optim/total_tokens": 1556873216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471217, "loss/out": 6.067861557006836, "created_at": "2025-01-14T21:22:37.850511+00:00"} {"global_step": 5940, "acc_step": 0, "speed/wps": 12956.660088176695, "speed/FLOPS": 203502151001535.22, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.938394546508789, "optim/lr": 0.002997842939810228, "optim/total_tokens": 1557135360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427376, "loss/out": 6.016664505004883, "created_at": "2025-01-14T21:22:42.913072+00:00"} {"global_step": 5941, "acc_step": 0, "speed/wps": 12960.905450784274, "speed/FLOPS": 203568830255026.78, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0008, "optim/grad_norm": 8.501975059509277, "optim/lr": 0.0029978383440908093, "optim/total_tokens": 1557397504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 6.10959529876709, "created_at": "2025-01-14T21:22:47.973414+00:00"} {"global_step": 5942, "acc_step": 0, "speed/wps": 12959.785453651355, "speed/FLOPS": 203551239161016.97, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.908509731292725, "optim/lr": 0.0029978337434844295, "optim/total_tokens": 1557659648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444641, "loss/out": 6.084190368652344, "created_at": "2025-01-14T21:22:53.034120+00:00"} {"global_step": 5943, "acc_step": 0, "speed/wps": 12953.756732771522, "speed/FLOPS": 203456549815268.56, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.397911071777344, "optim/lr": 0.0029978291379911034, "optim/total_tokens": 1557921792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298880, "loss/out": 6.055010795593262, "created_at": "2025-01-14T21:22:58.096854+00:00"} {"global_step": 5944, "acc_step": 0, "speed/wps": 12959.979978292113, "speed/FLOPS": 203554294437804.9, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0012, "optim/grad_norm": 5.500054359436035, "optim/lr": 0.0029978245276108466, "optim/total_tokens": 1558183936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 373517, "loss/out": 6.04400634765625, "created_at": "2025-01-14T21:23:03.161655+00:00"} {"global_step": 5945, "acc_step": 0, "speed/wps": 12960.412450697797, "speed/FLOPS": 203561087011215.56, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.707173347473145, "optim/lr": 0.002997819912343674, "optim/total_tokens": 1558446080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442208, "loss/out": 6.139960765838623, "created_at": "2025-01-14T21:23:08.226707+00:00"} {"global_step": 5946, "acc_step": 0, "speed/wps": 12961.51878545604, "speed/FLOPS": 203578463518857.03, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.023231506347656, "optim/lr": 0.0029978152921896007, "optim/total_tokens": 1558708224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 6.105584144592285, "created_at": "2025-01-14T21:23:13.286589+00:00"} {"global_step": 5947, "acc_step": 0, "speed/wps": 12964.067040133168, "speed/FLOPS": 203618487360230.2, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 15.565781593322754, "optim/lr": 0.002997810667148641, "optim/total_tokens": 1558970368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 5.886563301086426, "created_at": "2025-01-14T21:23:18.349688+00:00"} {"global_step": 5948, "acc_step": 0, "speed/wps": 12961.060587962887, "speed/FLOPS": 203571266897597.72, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 11.057506561279297, "optim/lr": 0.002997806037220811, "optim/total_tokens": 1559232512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426544, "loss/out": 6.049981117248535, "created_at": "2025-01-14T21:23:23.410031+00:00"} {"global_step": 5949, "acc_step": 0, "speed/wps": 12957.360032113615, "speed/FLOPS": 203513144582888.28, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.518580436706543, "optim/lr": 0.0029978014024061254, "optim/total_tokens": 1559494656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 6.075933933258057, "created_at": "2025-01-14T21:23:28.487804+00:00"} {"global_step": 5950, "acc_step": 0, "speed/wps": 12967.191535901922, "speed/FLOPS": 203667561859785.6, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.5919828414917, "optim/lr": 0.002997796762704599, "optim/total_tokens": 1559756800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293593, "loss/out": 6.022094249725342, "created_at": "2025-01-14T21:23:33.546226+00:00"} {"global_step": 5951, "acc_step": 0, "speed/wps": 12959.476318062392, "speed/FLOPS": 203546383761795.47, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.197009801864624, "optim/lr": 0.0029977921181162477, "optim/total_tokens": 1560018944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448077, "loss/out": 5.972322463989258, "created_at": "2025-01-14T21:23:38.613085+00:00"} {"global_step": 5952, "acc_step": 0, "speed/wps": 12963.10620072952, "speed/FLOPS": 203603396057064.22, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.756506443023682, "optim/lr": 0.0029977874686410856, "optim/total_tokens": 1560281088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 5.986248970031738, "created_at": "2025-01-14T21:23:43.676561+00:00"} {"global_step": 5953, "acc_step": 0, "speed/wps": 12967.803587324766, "speed/FLOPS": 203677174968428.16, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.945896863937378, "optim/lr": 0.002997782814279129, "optim/total_tokens": 1560543232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422006, "loss/out": 6.05011510848999, "created_at": "2025-01-14T21:23:48.738514+00:00"} {"global_step": 5954, "acc_step": 0, "speed/wps": 12964.26629348054, "speed/FLOPS": 203621616907853.47, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.8209261894226074, "optim/lr": 0.0029977781550303917, "optim/total_tokens": 1560805376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 6.0097150802612305, "created_at": "2025-01-14T21:23:53.797201+00:00"} {"global_step": 5955, "acc_step": 0, "speed/wps": 12964.27662114494, "speed/FLOPS": 203621779118014.56, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.42024564743042, "optim/lr": 0.0029977734908948903, "optim/total_tokens": 1561067520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425730, "loss/out": 5.959352016448975, "created_at": "2025-01-14T21:23:58.857119+00:00"} {"global_step": 5956, "acc_step": 0, "speed/wps": 12959.11366665129, "speed/FLOPS": 203540687823048.88, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.683784484863281, "optim/lr": 0.0029977688218726394, "optim/total_tokens": 1561329664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310371, "loss/out": 5.835643768310547, "created_at": "2025-01-14T21:24:03.918888+00:00"} {"global_step": 5957, "acc_step": 0, "speed/wps": 12963.03453180388, "speed/FLOPS": 203602270398103.6, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.55053186416626, "optim/lr": 0.0029977641479636548, "optim/total_tokens": 1561591808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427047, "loss/out": 5.895153045654297, "created_at": "2025-01-14T21:24:08.978543+00:00"} {"global_step": 5958, "acc_step": 0, "speed/wps": 12960.785033797112, "speed/FLOPS": 203566938940773.12, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0011, "optim/grad_norm": 4.749170303344727, "optim/lr": 0.0029977594691679506, "optim/total_tokens": 1561853952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450746, "loss/out": 5.969779014587402, "created_at": "2025-01-14T21:24:14.042649+00:00"} {"global_step": 5959, "acc_step": 0, "speed/wps": 12956.948531346732, "speed/FLOPS": 203506681397883.1, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.046011447906494, "optim/lr": 0.002997754785485543, "optim/total_tokens": 1562116096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 5.983953475952148, "created_at": "2025-01-14T21:24:19.104721+00:00"} {"global_step": 5960, "acc_step": 0, "speed/wps": 12958.371971305727, "speed/FLOPS": 203529038478450.6, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.4036226272583, "optim/lr": 0.002997750096916447, "optim/total_tokens": 1562378240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469353, "loss/out": 5.979601860046387, "created_at": "2025-01-14T21:24:24.173087+00:00"} {"global_step": 5961, "acc_step": 0, "speed/wps": 12956.874795266447, "speed/FLOPS": 203505523271418.22, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.3915278911590576, "optim/lr": 0.0029977454034606776, "optim/total_tokens": 1562640384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 6.019521713256836, "created_at": "2025-01-14T21:24:29.237146+00:00"} {"global_step": 5962, "acc_step": 0, "speed/wps": 12955.426000378122, "speed/FLOPS": 203482767956844.28, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 22.52913475036621, "optim/lr": 0.0029977407051182507, "optim/total_tokens": 1562902528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455517, "loss/out": 5.964807033538818, "created_at": "2025-01-14T21:24:34.301092+00:00"} {"global_step": 5963, "acc_step": 0, "speed/wps": 12963.404798737605, "speed/FLOPS": 203608085949097.84, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 30.401905059814453, "optim/lr": 0.0029977360018891815, "optim/total_tokens": 1563164672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304502, "loss/out": 6.0757904052734375, "created_at": "2025-01-14T21:24:39.360168+00:00"} {"global_step": 5964, "acc_step": 0, "speed/wps": 12958.785309360183, "speed/FLOPS": 203535530520581.8, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.598198890686035, "optim/lr": 0.002997731293773485, "optim/total_tokens": 1563426816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453666, "loss/out": 6.074257850646973, "created_at": "2025-01-14T21:24:44.427195+00:00"} {"global_step": 5965, "acc_step": 0, "speed/wps": 12957.515960077026, "speed/FLOPS": 203515593645821.12, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.23941969871521, "optim/lr": 0.002997726580771177, "optim/total_tokens": 1563688960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 6.054343223571777, "created_at": "2025-01-14T21:24:49.493031+00:00"} {"global_step": 5966, "acc_step": 0, "speed/wps": 12964.485721672885, "speed/FLOPS": 203625063329140.2, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.203585624694824, "optim/lr": 0.002997721862882272, "optim/total_tokens": 1563951104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422528, "loss/out": 6.067404747009277, "created_at": "2025-01-14T21:24:54.552509+00:00"} {"global_step": 5967, "acc_step": 0, "speed/wps": 12964.443275236767, "speed/FLOPS": 203624396649533.44, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.858802795410156, "optim/lr": 0.0029977171401067864, "optim/total_tokens": 1564213248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484902, "loss/out": 5.932602882385254, "created_at": "2025-01-14T21:24:59.616202+00:00"} {"global_step": 5968, "acc_step": 0, "speed/wps": 12961.227469733962, "speed/FLOPS": 203573888005132.03, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.883909225463867, "optim/lr": 0.0029977124124447355, "optim/total_tokens": 1564475392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425492, "loss/out": 6.009096622467041, "created_at": "2025-01-14T21:25:04.676015+00:00"} {"global_step": 5969, "acc_step": 0, "speed/wps": 12959.88592672867, "speed/FLOPS": 203552817228760.66, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.315535068511963, "optim/lr": 0.002997707679896134, "optim/total_tokens": 1564737536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300290, "loss/out": 6.023170471191406, "created_at": "2025-01-14T21:25:09.736516+00:00"} {"global_step": 5970, "acc_step": 0, "speed/wps": 12960.223558232927, "speed/FLOPS": 203558120195489.3, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 140.4553985595703, "optim/lr": 0.002997702942460998, "optim/total_tokens": 1564999680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440105, "loss/out": 6.001842498779297, "created_at": "2025-01-14T21:25:14.796710+00:00"} {"global_step": 5971, "acc_step": 0, "speed/wps": 12957.335416772565, "speed/FLOPS": 203512757965132.06, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 16.688955307006836, "optim/lr": 0.0029976982001393426, "optim/total_tokens": 1565261824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471454, "loss/out": 6.209450721740723, "created_at": "2025-01-14T21:25:19.858161+00:00"} {"global_step": 5972, "acc_step": 0, "speed/wps": 12965.769187038046, "speed/FLOPS": 203645221916366.25, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.456343650817871, "optim/lr": 0.0029976934529311835, "optim/total_tokens": 1565523968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 6.144535541534424, "created_at": "2025-01-14T21:25:24.918211+00:00"} {"global_step": 5973, "acc_step": 0, "speed/wps": 12960.977432770158, "speed/FLOPS": 203569960831028.72, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 17.12607765197754, "optim/lr": 0.002997688700836536, "optim/total_tokens": 1565786112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436916, "loss/out": 6.152690887451172, "created_at": "2025-01-14T21:25:29.981552+00:00"} {"global_step": 5974, "acc_step": 0, "speed/wps": 12960.816850037892, "speed/FLOPS": 203567438658554.7, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 37.815704345703125, "optim/lr": 0.0029976839438554163, "optim/total_tokens": 1566048256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428932, "loss/out": 6.05343770980835, "created_at": "2025-01-14T21:25:35.044628+00:00"} {"global_step": 5975, "acc_step": 0, "speed/wps": 12960.754296706844, "speed/FLOPS": 203566456172533.28, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.3603291511535645, "optim/lr": 0.002997679181987838, "optim/total_tokens": 1566310400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 6.241924285888672, "created_at": "2025-01-14T21:25:40.108638+00:00"} {"global_step": 5976, "acc_step": 0, "speed/wps": 12959.503585995932, "speed/FLOPS": 203546812042161.78, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.350868225097656, "optim/lr": 0.002997674415233819, "optim/total_tokens": 1566572544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313070, "loss/out": 5.95261287689209, "created_at": "2025-01-14T21:25:45.176513+00:00"} {"global_step": 5977, "acc_step": 0, "speed/wps": 12956.013101785087, "speed/FLOPS": 203491989191202.53, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0014, "optim/grad_norm": 4.5416669845581055, "optim/lr": 0.002997669643593374, "optim/total_tokens": 1566834688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442200, "loss/out": 6.142220497131348, "created_at": "2025-01-14T21:25:50.243352+00:00"} {"global_step": 5978, "acc_step": 0, "speed/wps": 12959.05248769421, "speed/FLOPS": 203539726923459.22, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.083125114440918, "optim/lr": 0.002997664867066518, "optim/total_tokens": 1567096832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 5.979865550994873, "created_at": "2025-01-14T21:25:55.309535+00:00"} {"global_step": 5979, "acc_step": 0, "speed/wps": 12956.450801060362, "speed/FLOPS": 203498863859782.56, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.574141025543213, "optim/lr": 0.0029976600856532667, "optim/total_tokens": 1567358976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421239, "loss/out": 6.053720474243164, "created_at": "2025-01-14T21:26:00.371947+00:00"} {"global_step": 5980, "acc_step": 0, "speed/wps": 12962.060801764095, "speed/FLOPS": 203586976629783.5, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.300820350646973, "optim/lr": 0.0029976552993536366, "optim/total_tokens": 1567621120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 5.877560615539551, "created_at": "2025-01-14T21:26:05.431496+00:00"} {"global_step": 5981, "acc_step": 0, "speed/wps": 12963.117560561019, "speed/FLOPS": 203603574478827.94, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.420095920562744, "optim/lr": 0.0029976505081676422, "optim/total_tokens": 1567883264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 5.893455982208252, "created_at": "2025-01-14T21:26:10.497034+00:00"} {"global_step": 5982, "acc_step": 0, "speed/wps": 12955.498794503936, "speed/FLOPS": 203483911288619.78, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0011, "optim/grad_norm": 5.149385452270508, "optim/lr": 0.0029976457120953, "optim/total_tokens": 1568145408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314202, "loss/out": 6.061650276184082, "created_at": "2025-01-14T21:26:15.560061+00:00"} {"global_step": 5983, "acc_step": 0, "speed/wps": 12959.152589035459, "speed/FLOPS": 203541299152577.2, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.12469482421875, "optim/lr": 0.002997640911136625, "optim/total_tokens": 1568407552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453666, "loss/out": 5.971517562866211, "created_at": "2025-01-14T21:26:20.628647+00:00"} {"global_step": 5984, "acc_step": 0, "speed/wps": 12962.482109238323, "speed/FLOPS": 203593593842603.38, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.1304707527160645, "optim/lr": 0.0029976361052916336, "optim/total_tokens": 1568669696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 5.988752841949463, "created_at": "2025-01-14T21:26:25.691454+00:00"} {"global_step": 5985, "acc_step": 0, "speed/wps": 12965.222032785203, "speed/FLOPS": 203636628106956.62, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.624804496765137, "optim/lr": 0.0029976312945603403, "optim/total_tokens": 1568931840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429977, "loss/out": 5.953054904937744, "created_at": "2025-01-14T21:26:30.749693+00:00"} {"global_step": 5986, "acc_step": 0, "speed/wps": 12963.858077302046, "speed/FLOPS": 203615205311822.4, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.854905366897583, "optim/lr": 0.002997626478942762, "optim/total_tokens": 1569193984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 6.041391372680664, "created_at": "2025-01-14T21:26:35.809946+00:00"} {"global_step": 5987, "acc_step": 0, "speed/wps": 12955.487607830926, "speed/FLOPS": 203483735586548.06, "speed/curr_iter_time": 1.2779, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.173724174499512, "optim/lr": 0.002997621658438914, "optim/total_tokens": 1569456128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460287, "loss/out": 5.946743965148926, "created_at": "2025-01-14T21:26:40.879820+00:00"} {"global_step": 5988, "acc_step": 0, "speed/wps": 12964.8188370806, "speed/FLOPS": 203630295364366.88, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 25.347820281982422, "optim/lr": 0.002997616833048811, "optim/total_tokens": 1569718272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330862, "loss/out": 5.957906723022461, "created_at": "2025-01-14T21:26:45.948216+00:00"} {"global_step": 5989, "acc_step": 0, "speed/wps": 12964.75120988335, "speed/FLOPS": 203629233186304.56, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.515265464782715, "optim/lr": 0.0029976120027724704, "optim/total_tokens": 1569980416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 388798, "loss/out": 6.217062473297119, "created_at": "2025-01-14T21:26:51.010860+00:00"} {"global_step": 5990, "acc_step": 0, "speed/wps": 12958.7332973164, "speed/FLOPS": 203534713599962.16, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 17.7646484375, "optim/lr": 0.0029976071676099075, "optim/total_tokens": 1570242560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466155, "loss/out": 6.123904705047607, "created_at": "2025-01-14T21:26:56.076493+00:00"} {"global_step": 5991, "acc_step": 0, "speed/wps": 12961.798687494793, "speed/FLOPS": 203582859765001.34, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.850666999816895, "optim/lr": 0.0029976023275611373, "optim/total_tokens": 1570504704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 6.04671573638916, "created_at": "2025-01-14T21:27:01.142523+00:00"} {"global_step": 5992, "acc_step": 0, "speed/wps": 12962.164005809322, "speed/FLOPS": 203588597591131.0, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.505405902862549, "optim/lr": 0.0029975974826261767, "optim/total_tokens": 1570766848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482522, "loss/out": 6.141524791717529, "created_at": "2025-01-14T21:27:06.203258+00:00"} {"global_step": 5993, "acc_step": 0, "speed/wps": 12961.390631472792, "speed/FLOPS": 203576450684448.1, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0009, "optim/grad_norm": 14.889713287353516, "optim/lr": 0.0029975926328050405, "optim/total_tokens": 1571028992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 6.041759490966797, "created_at": "2025-01-14T21:27:11.267360+00:00"} {"global_step": 5994, "acc_step": 0, "speed/wps": 12966.042948720255, "speed/FLOPS": 203649521719773.3, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.897191286087036, "optim/lr": 0.002997587778097745, "optim/total_tokens": 1571291136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306620, "loss/out": 6.0409111976623535, "created_at": "2025-01-14T21:27:16.329238+00:00"} {"global_step": 5995, "acc_step": 0, "speed/wps": 12965.504839643087, "speed/FLOPS": 203641069977277.03, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.473407030105591, "optim/lr": 0.002997582918504306, "optim/total_tokens": 1571553280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 409401, "loss/out": 6.002876281738281, "created_at": "2025-01-14T21:27:21.396143+00:00"} {"global_step": 5996, "acc_step": 0, "speed/wps": 12956.742081025242, "speed/FLOPS": 203503438811891.62, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.274312496185303, "optim/lr": 0.0029975780540247394, "optim/total_tokens": 1571815424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 5.9051408767700195, "created_at": "2025-01-14T21:27:26.457853+00:00"} {"global_step": 5997, "acc_step": 0, "speed/wps": 12962.088056698241, "speed/FLOPS": 203587404705976.47, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 32.91992950439453, "optim/lr": 0.002997573184659061, "optim/total_tokens": 1572077568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440108, "loss/out": 6.052762985229492, "created_at": "2025-01-14T21:27:31.518441+00:00"} {"global_step": 5998, "acc_step": 0, "speed/wps": 12965.27540361459, "speed/FLOPS": 203637466369171.38, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.184021234512329, "optim/lr": 0.002997568310407287, "optim/total_tokens": 1572339712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463248, "loss/out": 6.069476127624512, "created_at": "2025-01-14T21:27:36.582114+00:00"} {"global_step": 5999, "acc_step": 0, "speed/wps": 12961.330455195917, "speed/FLOPS": 203575505533328.9, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 9.269231796264648, "optim/lr": 0.0029975634312694325, "optim/total_tokens": 1572601856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 5.956277370452881, "created_at": "2025-01-14T21:27:41.643725+00:00"} {"global_step": 6000, "acc_step": 0, "speed/wps": 12958.464516609989, "speed/FLOPS": 203530492029624.6, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.177849292755127, "optim/lr": 0.002997558547245514, "optim/total_tokens": 1572864000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438536, "loss/out": 6.037744998931885, "created_at": "2025-01-14T21:27:46.708450+00:00"} {"global_step": 6001, "acc_step": 0, "speed/wps": 12307.20572559855, "speed/FLOPS": 193301577793429.8, "speed/curr_iter_time": 1.2703, "speed/data_load_time": 0.2724, "optim/grad_norm": 4.357061862945557, "optim/lr": 0.0029975536583355476, "optim/total_tokens": 1573126144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449420, "loss/out": 5.9806437492370605, "created_at": "2025-01-14T21:27:52.037072+00:00"} {"global_step": 6002, "acc_step": 0, "speed/wps": 12978.528703348178, "speed/FLOPS": 203845627653429.66, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.614447593688965, "optim/lr": 0.002997548764539549, "optim/total_tokens": 1573388288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 6.07302713394165, "created_at": "2025-01-14T21:27:57.092471+00:00"} {"global_step": 6003, "acc_step": 0, "speed/wps": 12978.831741854552, "speed/FLOPS": 203850387289590.8, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 31.381851196289062, "optim/lr": 0.002997543865857534, "optim/total_tokens": 1573650432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458430, "loss/out": 6.0286102294921875, "created_at": "2025-01-14T21:28:02.145385+00:00"} {"global_step": 6004, "acc_step": 0, "speed/wps": 12987.317818511789, "speed/FLOPS": 203983672784584.03, "speed/curr_iter_time": 1.2719, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.350686073303223, "optim/lr": 0.002997538962289519, "optim/total_tokens": 1573912576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446503, "loss/out": 6.169290542602539, "created_at": "2025-01-14T21:28:07.195602+00:00"} {"global_step": 6005, "acc_step": 0, "speed/wps": 12981.579044449369, "speed/FLOPS": 203893537452031.78, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 256.5239562988281, "optim/lr": 0.00299753405383552, "optim/total_tokens": 1574174720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333550, "loss/out": 5.869708061218262, "created_at": "2025-01-14T21:28:12.252835+00:00"} {"global_step": 6006, "acc_step": 0, "speed/wps": 12983.523652762682, "speed/FLOPS": 203924080197764.22, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 19.49834442138672, "optim/lr": 0.002997529140495552, "optim/total_tokens": 1574436864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290611, "loss/out": 6.064368724822998, "created_at": "2025-01-14T21:28:17.308684+00:00"} {"global_step": 6007, "acc_step": 0, "speed/wps": 12978.93371936734, "speed/FLOPS": 203851988986560.94, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.4344699382781982, "optim/lr": 0.0029975242222696325, "optim/total_tokens": 1574699008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 381078, "loss/out": 6.216358661651611, "created_at": "2025-01-14T21:28:22.362050+00:00"} {"global_step": 6008, "acc_step": 0, "speed/wps": 12990.010823806144, "speed/FLOPS": 204025970133308.97, "speed/curr_iter_time": 1.2712, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.556314468383789, "optim/lr": 0.0029975192991577764, "optim/total_tokens": 1574961152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429961, "loss/out": 6.064268112182617, "created_at": "2025-01-14T21:28:27.411142+00:00"} {"global_step": 6009, "acc_step": 0, "speed/wps": 12976.477110317519, "speed/FLOPS": 203813404565698.9, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.146493673324585, "optim/lr": 0.0029975143711600005, "optim/total_tokens": 1575223296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428688, "loss/out": 5.947120666503906, "created_at": "2025-01-14T21:28:32.470784+00:00"} {"global_step": 6010, "acc_step": 0, "speed/wps": 12982.997335701735, "speed/FLOPS": 203915813665086.62, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.39982271194458, "optim/lr": 0.0029975094382763204, "optim/total_tokens": 1575485440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479944, "loss/out": 6.088636875152588, "created_at": "2025-01-14T21:28:37.523355+00:00"} {"global_step": 6011, "acc_step": 0, "speed/wps": 12980.850306529324, "speed/FLOPS": 203882091621606.66, "speed/curr_iter_time": 1.2722, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.959193706512451, "optim/lr": 0.0029975045005067526, "optim/total_tokens": 1575747584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 6.164298057556152, "created_at": "2025-01-14T21:28:42.576120+00:00"} {"global_step": 6012, "acc_step": 0, "speed/wps": 12984.468596153438, "speed/FLOPS": 203938921832204.34, "speed/curr_iter_time": 1.2721, "speed/data_load_time": 0.0012, "optim/grad_norm": 4.875426292419434, "optim/lr": 0.0029974995578513123, "optim/total_tokens": 1576009728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455517, "loss/out": 6.0347723960876465, "created_at": "2025-01-14T21:28:47.635661+00:00"} {"global_step": 6013, "acc_step": 0, "speed/wps": 12975.43658072144, "speed/FLOPS": 203797061618555.0, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.664456844329834, "optim/lr": 0.0029974946103100172, "optim/total_tokens": 1576271872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440637, "loss/out": 5.970221996307373, "created_at": "2025-01-14T21:28:52.700572+00:00"} {"global_step": 6014, "acc_step": 0, "speed/wps": 12982.60718938483, "speed/FLOPS": 203909685881062.4, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.074515342712402, "optim/lr": 0.002997489657882882, "optim/total_tokens": 1576534016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455769, "loss/out": 5.973798751831055, "created_at": "2025-01-14T21:28:57.753647+00:00"} {"global_step": 6015, "acc_step": 0, "speed/wps": 12978.500275737373, "speed/FLOPS": 203845181158738.34, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.243593215942383, "optim/lr": 0.002997484700569924, "optim/total_tokens": 1576796160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 6.261491775512695, "created_at": "2025-01-14T21:29:02.807266+00:00"} {"global_step": 6016, "acc_step": 0, "speed/wps": 12972.616493333591, "speed/FLOPS": 203752768270922.66, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9762256145477295, "optim/lr": 0.0029974797383711584, "optim/total_tokens": 1577058304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 6.013336658477783, "created_at": "2025-01-14T21:29:07.865593+00:00"} {"global_step": 6017, "acc_step": 0, "speed/wps": 12986.135191542326, "speed/FLOPS": 203965098002928.88, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.3645546436309814, "optim/lr": 0.002997474771286602, "optim/total_tokens": 1577320448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434816, "loss/out": 5.953588485717773, "created_at": "2025-01-14T21:29:12.917264+00:00"} {"global_step": 6018, "acc_step": 0, "speed/wps": 12978.112146942167, "speed/FLOPS": 203839085062664.78, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.762679100036621, "optim/lr": 0.0029974697993162703, "optim/total_tokens": 1577582592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304748, "loss/out": 6.016340732574463, "created_at": "2025-01-14T21:29:17.975628+00:00"} {"global_step": 6019, "acc_step": 0, "speed/wps": 12975.341912310229, "speed/FLOPS": 203795574721068.88, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.785619735717773, "optim/lr": 0.0029974648224601807, "optim/total_tokens": 1577844736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362816, "loss/out": 6.227728843688965, "created_at": "2025-01-14T21:29:23.030453+00:00"} {"global_step": 6020, "acc_step": 0, "speed/wps": 12974.965735783064, "speed/FLOPS": 203789666351789.06, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.3126728534698486, "optim/lr": 0.002997459840718348, "optim/total_tokens": 1578106880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423841, "loss/out": 6.00469970703125, "created_at": "2025-01-14T21:29:28.090047+00:00"} {"global_step": 6021, "acc_step": 0, "speed/wps": 12981.149775240561, "speed/FLOPS": 203886795189229.75, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.775372505187988, "optim/lr": 0.0029974548540907895, "optim/total_tokens": 1578369024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449942, "loss/out": 5.988385200500488, "created_at": "2025-01-14T21:29:33.142555+00:00"} {"global_step": 6022, "acc_step": 0, "speed/wps": 12979.648181418808, "speed/FLOPS": 203863210594853.0, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.310632228851318, "optim/lr": 0.002997449862577522, "optim/total_tokens": 1578631168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461622, "loss/out": 6.1084465980529785, "created_at": "2025-01-14T21:29:38.196368+00:00"} {"global_step": 6023, "acc_step": 0, "speed/wps": 12975.951775939899, "speed/FLOPS": 203805153467413.3, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.025976657867432, "optim/lr": 0.00299744486617856, "optim/total_tokens": 1578893312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 6.019212245941162, "created_at": "2025-01-14T21:29:43.251448+00:00"} {"global_step": 6024, "acc_step": 0, "speed/wps": 12976.554982196936, "speed/FLOPS": 203814627650572.34, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.690514087677002, "optim/lr": 0.0029974398648939214, "optim/total_tokens": 1579155456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446505, "loss/out": 6.144718170166016, "created_at": "2025-01-14T21:29:48.306165+00:00"} {"global_step": 6025, "acc_step": 0, "speed/wps": 12985.159794787442, "speed/FLOPS": 203949778056557.8, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.057888031005859, "optim/lr": 0.002997434858723621, "optim/total_tokens": 1579417600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441679, "loss/out": 6.129758834838867, "created_at": "2025-01-14T21:29:53.361691+00:00"} {"global_step": 6026, "acc_step": 0, "speed/wps": 12978.1938729674, "speed/FLOPS": 203840368682194.4, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.8158938884735107, "optim/lr": 0.0029974298476676763, "optim/total_tokens": 1579679744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446503, "loss/out": 6.012599945068359, "created_at": "2025-01-14T21:29:58.417281+00:00"} {"global_step": 6027, "acc_step": 0, "speed/wps": 12973.28275924951, "speed/FLOPS": 203763232892677.5, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.960416793823242, "optim/lr": 0.0029974248317261036, "optim/total_tokens": 1579941888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437717, "loss/out": 5.930150985717773, "created_at": "2025-01-14T21:30:03.474292+00:00"} {"global_step": 6028, "acc_step": 0, "speed/wps": 12978.995320785429, "speed/FLOPS": 203852956521481.25, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 348.93707275390625, "optim/lr": 0.0029974198108989195, "optim/total_tokens": 1580204032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469875, "loss/out": 5.998858451843262, "created_at": "2025-01-14T21:30:08.527597+00:00"} {"global_step": 6029, "acc_step": 0, "speed/wps": 12978.021434270055, "speed/FLOPS": 203837660295496.97, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 25.181386947631836, "optim/lr": 0.002997414785186139, "optim/total_tokens": 1580466176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 6.179495334625244, "created_at": "2025-01-14T21:30:13.582332+00:00"} {"global_step": 6030, "acc_step": 0, "speed/wps": 12976.226487039978, "speed/FLOPS": 203809468182732.66, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.338335990905762, "optim/lr": 0.00299740975458778, "optim/total_tokens": 1580728320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 6.147292137145996, "created_at": "2025-01-14T21:30:18.640042+00:00"} {"global_step": 6031, "acc_step": 0, "speed/wps": 12971.719242364105, "speed/FLOPS": 203738675711494.84, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.161107063293457, "optim/lr": 0.0029974047191038584, "optim/total_tokens": 1580990464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334369, "loss/out": 6.130835056304932, "created_at": "2025-01-14T21:30:23.696917+00:00"} {"global_step": 6032, "acc_step": 0, "speed/wps": 12978.531276584505, "speed/FLOPS": 203845668069642.2, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 83.57079315185547, "optim/lr": 0.0029973996787343903, "optim/total_tokens": 1581252608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331010, "loss/out": 6.045204162597656, "created_at": "2025-01-14T21:30:28.757252+00:00"} {"global_step": 6033, "acc_step": 0, "speed/wps": 12982.262259219597, "speed/FLOPS": 203904268278833.7, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0008, "optim/grad_norm": 37.12173080444336, "optim/lr": 0.0029973946334793927, "optim/total_tokens": 1581514752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449420, "loss/out": 6.100236415863037, "created_at": "2025-01-14T21:30:33.809236+00:00"} {"global_step": 6034, "acc_step": 0, "speed/wps": 12980.37314968578, "speed/FLOPS": 203874597217696.56, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0011, "optim/grad_norm": 16.426401138305664, "optim/lr": 0.002997389583338881, "optim/total_tokens": 1581776896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443306, "loss/out": 6.20352840423584, "created_at": "2025-01-14T21:30:38.862869+00:00"} {"global_step": 6035, "acc_step": 0, "speed/wps": 12977.013299470602, "speed/FLOPS": 203821826153149.2, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.368398666381836, "optim/lr": 0.002997384528312873, "optim/total_tokens": 1582039040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 6.077764511108398, "created_at": "2025-01-14T21:30:43.918637+00:00"} {"global_step": 6036, "acc_step": 0, "speed/wps": 12973.970810781522, "speed/FLOPS": 203774039687469.62, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 17.229917526245117, "optim/lr": 0.0029973794684013846, "optim/total_tokens": 1582301184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460815, "loss/out": 6.1874799728393555, "created_at": "2025-01-14T21:30:48.977430+00:00"} {"global_step": 6037, "acc_step": 0, "speed/wps": 12983.229425929587, "speed/FLOPS": 203919458961040.44, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.882265567779541, "optim/lr": 0.0029973744036044326, "optim/total_tokens": 1582563328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427351, "loss/out": 6.04826545715332, "created_at": "2025-01-14T21:30:54.033271+00:00"} {"global_step": 6038, "acc_step": 0, "speed/wps": 12981.705567107536, "speed/FLOPS": 203895524664238.66, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.857237815856934, "optim/lr": 0.002997369333922033, "optim/total_tokens": 1582825472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446218, "loss/out": 6.068835258483887, "created_at": "2025-01-14T21:30:59.085932+00:00"} {"global_step": 6039, "acc_step": 0, "speed/wps": 12973.406601656428, "speed/FLOPS": 203765178007855.44, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.408602237701416, "optim/lr": 0.0029973642593542024, "optim/total_tokens": 1583087616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463248, "loss/out": 5.9783735275268555, "created_at": "2025-01-14T21:31:04.141977+00:00"} {"global_step": 6040, "acc_step": 0, "speed/wps": 12971.564875723801, "speed/FLOPS": 203736251171286.88, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0008, "optim/grad_norm": 100.20081329345703, "optim/lr": 0.002997359179900958, "optim/total_tokens": 1583349760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470641, "loss/out": 6.087408065795898, "created_at": "2025-01-14T21:31:09.198773+00:00"} {"global_step": 6041, "acc_step": 0, "speed/wps": 12976.743448839567, "speed/FLOPS": 203817587778175.16, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.060083866119385, "optim/lr": 0.0029973540955623157, "optim/total_tokens": 1583611904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 6.099514484405518, "created_at": "2025-01-14T21:31:14.256727+00:00"} {"global_step": 6042, "acc_step": 0, "speed/wps": 12981.619337640559, "speed/FLOPS": 203894170311968.1, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0006, "optim/grad_norm": 27.41562843322754, "optim/lr": 0.002997349006338292, "optim/total_tokens": 1583874048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448606, "loss/out": 5.97076416015625, "created_at": "2025-01-14T21:31:19.311880+00:00"} {"global_step": 6043, "acc_step": 0, "speed/wps": 12977.85929442177, "speed/FLOPS": 203835113666376.2, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.596329689025879, "optim/lr": 0.002997343912228904, "optim/total_tokens": 1584136192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328845, "loss/out": 6.093201637268066, "created_at": "2025-01-14T21:31:24.372198+00:00"} {"global_step": 6044, "acc_step": 0, "speed/wps": 12974.315331582951, "speed/FLOPS": 203779450860077.25, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.84062385559082, "optim/lr": 0.0029973388132341684, "optim/total_tokens": 1584398336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332206, "loss/out": 6.058238983154297, "created_at": "2025-01-14T21:31:29.426957+00:00"} {"global_step": 6045, "acc_step": 0, "speed/wps": 12970.160498650064, "speed/FLOPS": 203714193499527.53, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.83599853515625, "optim/lr": 0.002997333709354101, "optim/total_tokens": 1584660480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411794, "loss/out": 6.054105758666992, "created_at": "2025-01-14T21:31:34.488025+00:00"} {"global_step": 6046, "acc_step": 0, "speed/wps": 12974.90129219573, "speed/FLOPS": 203788654176695.2, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.093677520751953, "optim/lr": 0.00299732860058872, "optim/total_tokens": 1584922624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442778, "loss/out": 6.047487258911133, "created_at": "2025-01-14T21:31:39.542687+00:00"} {"global_step": 6047, "acc_step": 0, "speed/wps": 12974.56256060412, "speed/FLOPS": 203783333931582.97, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 16.472654342651367, "optim/lr": 0.0029973234869380403, "optim/total_tokens": 1585184768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 6.052145957946777, "created_at": "2025-01-14T21:31:44.598705+00:00"} {"global_step": 6048, "acc_step": 0, "speed/wps": 12974.520454837213, "speed/FLOPS": 203782672602654.3, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 38.32001876831055, "optim/lr": 0.00299731836840208, "optim/total_tokens": 1585446912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 6.107207298278809, "created_at": "2025-01-14T21:31:49.658701+00:00"} {"global_step": 6049, "acc_step": 0, "speed/wps": 12975.958315348824, "speed/FLOPS": 203805256177816.56, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.833677291870117, "optim/lr": 0.0029973132449808548, "optim/total_tokens": 1585709056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436895, "loss/out": 6.151677131652832, "created_at": "2025-01-14T21:31:54.713496+00:00"} {"global_step": 6050, "acc_step": 0, "speed/wps": 12978.504625086734, "speed/FLOPS": 203845249471246.7, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.18485689163208, "optim/lr": 0.0029973081166743817, "optim/total_tokens": 1585971200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 5.937931060791016, "created_at": "2025-01-14T21:31:59.772089+00:00"} {"global_step": 6051, "acc_step": 0, "speed/wps": 12974.445958848155, "speed/FLOPS": 203781502540777.94, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 660.1317138671875, "optim/lr": 0.002997302983482678, "optim/total_tokens": 1586233344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449696, "loss/out": 6.06856632232666, "created_at": "2025-01-14T21:32:04.829891+00:00"} {"global_step": 6052, "acc_step": 0, "speed/wps": 12980.255413899635, "speed/FLOPS": 203872748015389.25, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.214043140411377, "optim/lr": 0.002997297845405759, "optim/total_tokens": 1586495488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445933, "loss/out": 6.105185508728027, "created_at": "2025-01-14T21:32:09.887835+00:00"} {"global_step": 6053, "acc_step": 0, "speed/wps": 12974.728835720967, "speed/FLOPS": 203785945510779.03, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.875241279602051, "optim/lr": 0.0029972927024436435, "optim/total_tokens": 1586757632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 281015, "loss/out": 6.055453777313232, "created_at": "2025-01-14T21:32:14.949256+00:00"} {"global_step": 6054, "acc_step": 0, "speed/wps": 12979.815125602274, "speed/FLOPS": 203865832682659.66, "speed/curr_iter_time": 1.2721, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.693924903869629, "optim/lr": 0.0029972875545963467, "optim/total_tokens": 1587019776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322156, "loss/out": 5.97074031829834, "created_at": "2025-01-14T21:32:20.003614+00:00"} {"global_step": 6055, "acc_step": 0, "speed/wps": 12981.470789318782, "speed/FLOPS": 203891837156447.78, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.162074089050293, "optim/lr": 0.002997282401863886, "optim/total_tokens": 1587281920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334608, "loss/out": 6.0811004638671875, "created_at": "2025-01-14T21:32:25.055679+00:00"} {"global_step": 6056, "acc_step": 0, "speed/wps": 12972.108478674774, "speed/FLOPS": 203744789202620.66, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0013, "optim/grad_norm": 3.734982490539551, "optim/lr": 0.002997277244246278, "optim/total_tokens": 1587544064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 403274, "loss/out": 6.051106929779053, "created_at": "2025-01-14T21:32:30.116146+00:00"} {"global_step": 6057, "acc_step": 0, "speed/wps": 12980.600699563674, "speed/FLOPS": 203878171201215.22, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.465449333190918, "optim/lr": 0.00299727208174354, "optim/total_tokens": 1587806208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429977, "loss/out": 5.994050979614258, "created_at": "2025-01-14T21:32:35.174409+00:00"} {"global_step": 6058, "acc_step": 0, "speed/wps": 12977.207058868129, "speed/FLOPS": 203824869410740.66, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 22.15235710144043, "optim/lr": 0.0029972669143556886, "optim/total_tokens": 1588068352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 6.034041881561279, "created_at": "2025-01-14T21:32:40.233646+00:00"} {"global_step": 6059, "acc_step": 0, "speed/wps": 12983.21856260914, "speed/FLOPS": 203919288337665.03, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0007, "optim/grad_norm": 73.37583923339844, "optim/lr": 0.0029972617420827405, "optim/total_tokens": 1588330496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464535, "loss/out": 6.114109992980957, "created_at": "2025-01-14T21:32:45.285358+00:00"} {"global_step": 6060, "acc_step": 0, "speed/wps": 12973.674301921916, "speed/FLOPS": 203769382608383.62, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 24.62418556213379, "optim/lr": 0.0029972565649247123, "optim/total_tokens": 1588592640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422813, "loss/out": 6.098182201385498, "created_at": "2025-01-14T21:32:50.340893+00:00"} {"global_step": 6061, "acc_step": 0, "speed/wps": 12980.765300057368, "speed/FLOPS": 203880756478153.44, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 118.0068130493164, "optim/lr": 0.0029972513828816216, "optim/total_tokens": 1588854784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451523, "loss/out": 6.214208602905273, "created_at": "2025-01-14T21:32:55.394146+00:00"} {"global_step": 6062, "acc_step": 0, "speed/wps": 12966.74349823136, "speed/FLOPS": 203660524812501.16, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0009, "optim/grad_norm": 6.756002902984619, "optim/lr": 0.0029972461959534846, "optim/total_tokens": 1589116928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 6.208003044128418, "created_at": "2025-01-14T21:33:00.452177+00:00"} {"global_step": 6063, "acc_step": 0, "speed/wps": 12975.022356810367, "speed/FLOPS": 203790555662828.28, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.789602279663086, "optim/lr": 0.0029972410041403186, "optim/total_tokens": 1589379072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454182, "loss/out": 6.103824138641357, "created_at": "2025-01-14T21:33:05.515240+00:00"} {"global_step": 6064, "acc_step": 0, "speed/wps": 12979.47758340841, "speed/FLOPS": 203860531118673.47, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 15.681556701660156, "optim/lr": 0.002997235807442141, "optim/total_tokens": 1589641216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311717, "loss/out": 6.048922061920166, "created_at": "2025-01-14T21:33:10.568760+00:00"} {"global_step": 6065, "acc_step": 0, "speed/wps": 12974.01414727495, "speed/FLOPS": 203774720346649.47, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0008, "optim/grad_norm": 35.022010803222656, "optim/lr": 0.002997230605858968, "optim/total_tokens": 1589903360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309920, "loss/out": 6.148341178894043, "created_at": "2025-01-14T21:33:15.624662+00:00"} {"global_step": 6066, "acc_step": 0, "speed/wps": 12975.759448960891, "speed/FLOPS": 203802132707923.03, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 36.870338439941406, "optim/lr": 0.0029972253993908165, "optim/total_tokens": 1590165504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310904, "loss/out": 6.202777862548828, "created_at": "2025-01-14T21:33:20.678759+00:00"} {"global_step": 6067, "acc_step": 0, "speed/wps": 12969.451354646435, "speed/FLOPS": 203703055418484.53, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 23.78324317932129, "optim/lr": 0.0029972201880377036, "optim/total_tokens": 1590427648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320810, "loss/out": 6.088855266571045, "created_at": "2025-01-14T21:33:25.738417+00:00"} {"global_step": 6068, "acc_step": 0, "speed/wps": 12973.802579329466, "speed/FLOPS": 203771397381340.8, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.176414966583252, "optim/lr": 0.002997214971799647, "optim/total_tokens": 1590689792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365215, "loss/out": 6.160739421844482, "created_at": "2025-01-14T21:33:30.793707+00:00"} {"global_step": 6069, "acc_step": 0, "speed/wps": 12972.973342950485, "speed/FLOPS": 203758373084518.75, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 28.759342193603516, "optim/lr": 0.002997209750676663, "optim/total_tokens": 1590951936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431283, "loss/out": 6.087759494781494, "created_at": "2025-01-14T21:33:35.853307+00:00"} {"global_step": 6070, "acc_step": 0, "speed/wps": 12979.498957591324, "speed/FLOPS": 203860866829587.3, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.980432510375977, "optim/lr": 0.002997204524668769, "optim/total_tokens": 1591214080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428932, "loss/out": 6.112768173217773, "created_at": "2025-01-14T21:33:40.906580+00:00"} {"global_step": 6071, "acc_step": 0, "speed/wps": 12976.438985933868, "speed/FLOPS": 203812805769865.28, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.604260444641113, "optim/lr": 0.0029971992937759826, "optim/total_tokens": 1591476224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 6.0602264404296875, "created_at": "2025-01-14T21:33:45.961715+00:00"} {"global_step": 6072, "acc_step": 0, "speed/wps": 12975.35251841486, "speed/FLOPS": 203795741304516.44, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 15.468212127685547, "optim/lr": 0.002997194057998319, "optim/total_tokens": 1591738368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 6.159650802612305, "created_at": "2025-01-14T21:33:51.016126+00:00"} {"global_step": 6073, "acc_step": 0, "speed/wps": 12979.758850591661, "speed/FLOPS": 203864948806288.03, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.710424423217773, "optim/lr": 0.0029971888173357976, "optim/total_tokens": 1592000512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445933, "loss/out": 6.169072151184082, "created_at": "2025-01-14T21:33:56.070304+00:00"} {"global_step": 6074, "acc_step": 0, "speed/wps": 12978.230748630793, "speed/FLOPS": 203840947865160.44, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.570223093032837, "optim/lr": 0.0029971835717884335, "optim/total_tokens": 1592262656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450471, "loss/out": 6.16337776184082, "created_at": "2025-01-14T21:34:01.123595+00:00"} {"global_step": 6075, "acc_step": 0, "speed/wps": 12975.877978939756, "speed/FLOPS": 203803994384118.34, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3536787033081055, "optim/lr": 0.002997178321356245, "optim/total_tokens": 1592524800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 5.965505599975586, "created_at": "2025-01-14T21:34:06.181903+00:00"} {"global_step": 6076, "acc_step": 0, "speed/wps": 12974.231128315905, "speed/FLOPS": 203778128332060.88, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.6457481384277344, "optim/lr": 0.002997173066039249, "optim/total_tokens": 1592786944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443835, "loss/out": 5.992530822753906, "created_at": "2025-01-14T21:34:11.236787+00:00"} {"global_step": 6077, "acc_step": 0, "speed/wps": 12979.147159179682, "speed/FLOPS": 203855341352110.8, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.608613967895508, "optim/lr": 0.0029971678058374625, "optim/total_tokens": 1593049088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436381, "loss/out": 5.986224174499512, "created_at": "2025-01-14T21:34:16.294404+00:00"} {"global_step": 6078, "acc_step": 0, "speed/wps": 12975.317927217633, "speed/FLOPS": 203795198002230.72, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0009, "optim/grad_norm": 5.55062198638916, "optim/lr": 0.0029971625407509027, "optim/total_tokens": 1593311232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322066, "loss/out": 6.078662395477295, "created_at": "2025-01-14T21:34:21.353197+00:00"} {"global_step": 6079, "acc_step": 0, "speed/wps": 12974.01468787663, "speed/FLOPS": 203774728837541.72, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.397432804107666, "optim/lr": 0.0029971572707795865, "optim/total_tokens": 1593573376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310734, "loss/out": 6.182814598083496, "created_at": "2025-01-14T21:34:26.408841+00:00"} {"global_step": 6080, "acc_step": 0, "speed/wps": 12976.06488016934, "speed/FLOPS": 203806929924758.2, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 30.78592872619629, "optim/lr": 0.002997151995923532, "optim/total_tokens": 1593835520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319377, "loss/out": 6.083091735839844, "created_at": "2025-01-14T21:34:31.463025+00:00"} {"global_step": 6081, "acc_step": 0, "speed/wps": 12980.998428776407, "speed/FLOPS": 203884418085038.38, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 17.32186508178711, "optim/lr": 0.002997146716182755, "optim/total_tokens": 1594097664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298696, "loss/out": 6.240029335021973, "created_at": "2025-01-14T21:34:36.516391+00:00"} {"global_step": 6082, "acc_step": 0, "speed/wps": 12974.511476806361, "speed/FLOPS": 203782531590342.7, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.112712860107422, "optim/lr": 0.0029971414315572748, "optim/total_tokens": 1594359808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390956, "loss/out": 6.216601848602295, "created_at": "2025-01-14T21:34:41.572701+00:00"} {"global_step": 6083, "acc_step": 0, "speed/wps": 12970.852799947479, "speed/FLOPS": 203725067042725.1, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.402817726135254, "optim/lr": 0.0029971361420471064, "optim/total_tokens": 1594621952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444359, "loss/out": 6.1996049880981445, "created_at": "2025-01-14T21:34:46.630439+00:00"} {"global_step": 6084, "acc_step": 0, "speed/wps": 12970.109578602762, "speed/FLOPS": 203713393730213.16, "speed/curr_iter_time": 1.2722, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.585604667663574, "optim/lr": 0.002997130847652268, "optim/total_tokens": 1594884096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 6.379624366760254, "created_at": "2025-01-14T21:34:51.688548+00:00"} {"global_step": 6085, "acc_step": 0, "speed/wps": 12968.949090031012, "speed/FLOPS": 203695166662515.4, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.159074783325195, "optim/lr": 0.002997125548372777, "optim/total_tokens": 1595146240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 6.307222366333008, "created_at": "2025-01-14T21:34:56.751132+00:00"} {"global_step": 6086, "acc_step": 0, "speed/wps": 12965.744761913076, "speed/FLOPS": 203644838286214.97, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.46621322631836, "optim/lr": 0.002997120244208651, "optim/total_tokens": 1595408384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465057, "loss/out": 6.2609148025512695, "created_at": "2025-01-14T21:35:01.810800+00:00"} {"global_step": 6087, "acc_step": 0, "speed/wps": 12973.735312846413, "speed/FLOPS": 203770340868790.66, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 20.04610824584961, "optim/lr": 0.002997114935159907, "optim/total_tokens": 1595670528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 6.332012176513672, "created_at": "2025-01-14T21:35:06.866185+00:00"} {"global_step": 6088, "acc_step": 0, "speed/wps": 12971.9541553786, "speed/FLOPS": 203742365343191.8, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 50.94882583618164, "optim/lr": 0.0029971096212265615, "optim/total_tokens": 1595932672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430791, "loss/out": 6.278560161590576, "created_at": "2025-01-14T21:35:11.921904+00:00"} {"global_step": 6089, "acc_step": 0, "speed/wps": 12973.087679606311, "speed/FLOPS": 203760168898813.53, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 31.28707504272461, "optim/lr": 0.002997104302408633, "optim/total_tokens": 1596194816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321343, "loss/out": 6.257378578186035, "created_at": "2025-01-14T21:35:16.980568+00:00"} {"global_step": 6090, "acc_step": 0, "speed/wps": 12975.929075891352, "speed/FLOPS": 203804796931957.75, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 70.69467163085938, "optim/lr": 0.0029970989787061384, "optim/total_tokens": 1596456960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371881, "loss/out": 6.451692581176758, "created_at": "2025-01-14T21:35:22.036225+00:00"} {"global_step": 6091, "acc_step": 0, "speed/wps": 12979.376442438192, "speed/FLOPS": 203858942560753.9, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.438897132873535, "optim/lr": 0.002997093650119095, "optim/total_tokens": 1596719104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415331, "loss/out": 6.39300537109375, "created_at": "2025-01-14T21:35:27.089336+00:00"} {"global_step": 6092, "acc_step": 0, "speed/wps": 12975.606787797458, "speed/FLOPS": 203799734954572.9, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 21.316802978515625, "optim/lr": 0.0029970883166475205, "optim/total_tokens": 1596981248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439571, "loss/out": 6.30928897857666, "created_at": "2025-01-14T21:35:32.143888+00:00"} {"global_step": 6093, "acc_step": 0, "speed/wps": 12978.401622648817, "speed/FLOPS": 203843631676419.03, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.845926761627197, "optim/lr": 0.0029970829782914317, "optim/total_tokens": 1597243392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353744, "loss/out": 6.324769973754883, "created_at": "2025-01-14T21:35:37.201885+00:00"} {"global_step": 6094, "acc_step": 0, "speed/wps": 12974.85030620936, "speed/FLOPS": 203787853371717.2, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 26.776296615600586, "optim/lr": 0.002997077635050847, "optim/total_tokens": 1597505536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445929, "loss/out": 6.254495143890381, "created_at": "2025-01-14T21:35:42.260793+00:00"} {"global_step": 6095, "acc_step": 0, "speed/wps": 12978.020455320195, "speed/FLOPS": 203837644919744.22, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.478155136108398, "optim/lr": 0.002997072286925782, "optim/total_tokens": 1597767680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440919, "loss/out": 6.201404571533203, "created_at": "2025-01-14T21:35:47.315699+00:00"} {"global_step": 6096, "acc_step": 0, "speed/wps": 12972.46246405824, "speed/FLOPS": 203750349029494.88, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 20.733871459960938, "optim/lr": 0.0029970669339162564, "optim/total_tokens": 1598029824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 6.143918514251709, "created_at": "2025-01-14T21:35:52.371106+00:00"} {"global_step": 6097, "acc_step": 0, "speed/wps": 12980.137984325536, "speed/FLOPS": 203870903622562.88, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.367003440856934, "optim/lr": 0.0029970615760222864, "optim/total_tokens": 1598291968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 6.212416648864746, "created_at": "2025-01-14T21:35:57.423638+00:00"} {"global_step": 6098, "acc_step": 0, "speed/wps": 12974.608087577713, "speed/FLOPS": 203784048995263.88, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.383568286895752, "optim/lr": 0.0029970562132438897, "optim/total_tokens": 1598554112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436610, "loss/out": 6.098904609680176, "created_at": "2025-01-14T21:36:02.487475+00:00"} {"global_step": 6099, "acc_step": 0, "speed/wps": 12970.923015607648, "speed/FLOPS": 203726169876154.47, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.893116474151611, "optim/lr": 0.0029970508455810837, "optim/total_tokens": 1598816256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482522, "loss/out": 6.276191234588623, "created_at": "2025-01-14T21:36:07.543549+00:00"} {"global_step": 6100, "acc_step": 0, "speed/wps": 12977.625098507247, "speed/FLOPS": 203831435297719.7, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.5749945640563965, "optim/lr": 0.0029970454730338866, "optim/total_tokens": 1599078400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450462, "loss/out": 6.067596435546875, "created_at": "2025-01-14T21:36:12.597550+00:00"} {"global_step": 6101, "acc_step": 0, "speed/wps": 12978.72237634373, "speed/FLOPS": 203848669553959.4, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0008, "optim/grad_norm": 18.086658477783203, "optim/lr": 0.0029970400956023146, "optim/total_tokens": 1599340544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427880, "loss/out": 6.066393852233887, "created_at": "2025-01-14T21:36:17.662577+00:00"} {"global_step": 6102, "acc_step": 0, "speed/wps": 12973.021893809026, "speed/FLOPS": 203759135642468.0, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 72.76475524902344, "optim/lr": 0.0029970347132863865, "optim/total_tokens": 1599602688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368718, "loss/out": 6.153247356414795, "created_at": "2025-01-14T21:36:22.722622+00:00"} {"global_step": 6103, "acc_step": 0, "speed/wps": 12969.998536036483, "speed/FLOPS": 203711649654121.25, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 23.81629180908203, "optim/lr": 0.002997029326086119, "optim/total_tokens": 1599864832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449657, "loss/out": 6.225795745849609, "created_at": "2025-01-14T21:36:27.780839+00:00"} {"global_step": 6104, "acc_step": 0, "speed/wps": 12976.817461488923, "speed/FLOPS": 203818750248537.34, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 25.98262596130371, "optim/lr": 0.0029970239340015308, "optim/total_tokens": 1600126976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 6.238893508911133, "created_at": "2025-01-14T21:36:32.834773+00:00"} {"global_step": 6105, "acc_step": 0, "speed/wps": 12968.945335066563, "speed/FLOPS": 203695107685638.88, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.172980308532715, "optim/lr": 0.002997018537032638, "optim/total_tokens": 1600389120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 6.19644832611084, "created_at": "2025-01-14T21:36:37.894265+00:00"} {"global_step": 6106, "acc_step": 0, "speed/wps": 12972.881257559286, "speed/FLOPS": 203756926756911.38, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.001, "optim/grad_norm": 72.14080810546875, "optim/lr": 0.002997013135179459, "optim/total_tokens": 1600651264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 6.1396355628967285, "created_at": "2025-01-14T21:36:42.951018+00:00"} {"global_step": 6107, "acc_step": 0, "speed/wps": 12973.732354017413, "speed/FLOPS": 203770294396315.34, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.061487197875977, "optim/lr": 0.0029970077284420115, "optim/total_tokens": 1600913408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 6.156291484832764, "created_at": "2025-01-14T21:36:48.006181+00:00"} {"global_step": 6108, "acc_step": 0, "speed/wps": 12972.547622493617, "speed/FLOPS": 203751686559742.44, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.518749952316284, "optim/lr": 0.0029970023168203125, "optim/total_tokens": 1601175552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462672, "loss/out": 6.099554061889648, "created_at": "2025-01-14T21:36:53.063246+00:00"} {"global_step": 6109, "acc_step": 0, "speed/wps": 12978.187283433366, "speed/FLOPS": 203840265184507.44, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 29.665006637573242, "optim/lr": 0.0029969969003143812, "optim/total_tokens": 1601437696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444067, "loss/out": 6.178187370300293, "created_at": "2025-01-14T21:36:58.121665+00:00"} {"global_step": 6110, "acc_step": 0, "speed/wps": 12975.760863038675, "speed/FLOPS": 203802154917957.78, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 101.20067596435547, "optim/lr": 0.0029969914789242332, "optim/total_tokens": 1601699840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462435, "loss/out": 6.15585994720459, "created_at": "2025-01-14T21:37:03.176234+00:00"} {"global_step": 6111, "acc_step": 0, "speed/wps": 12967.619521588827, "speed/FLOPS": 203674283963112.4, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.610639572143555, "optim/lr": 0.002996986052649888, "optim/total_tokens": 1601961984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447555, "loss/out": 6.353315830230713, "created_at": "2025-01-14T21:37:08.243226+00:00"} {"global_step": 6112, "acc_step": 0, "speed/wps": 12974.295957558308, "speed/FLOPS": 203779146564397.53, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.705209732055664, "optim/lr": 0.002996980621491362, "optim/total_tokens": 1602224128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312250, "loss/out": 6.10164737701416, "created_at": "2025-01-14T21:37:13.304981+00:00"} {"global_step": 6113, "acc_step": 0, "speed/wps": 12970.887142597754, "speed/FLOPS": 203725606441239.4, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.800414085388184, "optim/lr": 0.002996975185448674, "optim/total_tokens": 1602486272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 393353, "loss/out": 6.097636699676514, "created_at": "2025-01-14T21:37:18.396790+00:00"} {"global_step": 6114, "acc_step": 0, "speed/wps": 12974.31572150403, "speed/FLOPS": 203779456984323.6, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.556343078613281, "optim/lr": 0.0029969697445218404, "optim/total_tokens": 1602748416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429977, "loss/out": 6.074618816375732, "created_at": "2025-01-14T21:37:23.451521+00:00"} {"global_step": 6115, "acc_step": 0, "speed/wps": 12969.895233493608, "speed/FLOPS": 203710027145732.84, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.811442375183105, "optim/lr": 0.0029969642987108806, "optim/total_tokens": 1603010560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 6.165884017944336, "created_at": "2025-01-14T21:37:28.513809+00:00"} {"global_step": 6116, "acc_step": 0, "speed/wps": 12970.000596701451, "speed/FLOPS": 203711682019696.25, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 20.160987854003906, "optim/lr": 0.0029969588480158113, "optim/total_tokens": 1603272704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444357, "loss/out": 5.945852279663086, "created_at": "2025-01-14T21:37:33.570320+00:00"} {"global_step": 6117, "acc_step": 0, "speed/wps": 12977.339474702081, "speed/FLOPS": 203826949183371.3, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 21.216699600219727, "optim/lr": 0.00299695339243665, "optim/total_tokens": 1603534848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426252, "loss/out": 6.096752166748047, "created_at": "2025-01-14T21:37:38.628368+00:00"} {"global_step": 6118, "acc_step": 0, "speed/wps": 12972.892945441214, "speed/FLOPS": 203757110331157.88, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 11.356749534606934, "optim/lr": 0.002996947931973415, "optim/total_tokens": 1603796992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316449, "loss/out": 6.166693687438965, "created_at": "2025-01-14T21:37:43.687742+00:00"} {"global_step": 6119, "acc_step": 0, "speed/wps": 12972.643204357943, "speed/FLOPS": 203753187804264.9, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.011080741882324, "optim/lr": 0.002996942466626125, "optim/total_tokens": 1604059136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311028, "loss/out": 6.114720821380615, "created_at": "2025-01-14T21:37:48.745916+00:00"} {"global_step": 6120, "acc_step": 0, "speed/wps": 12974.811478467678, "speed/FLOPS": 203787243528679.16, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.652433395385742, "optim/lr": 0.0029969369963947956, "optim/total_tokens": 1604321280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296475, "loss/out": 6.169079780578613, "created_at": "2025-01-14T21:37:53.800516+00:00"} {"global_step": 6121, "acc_step": 0, "speed/wps": 12976.216118792, "speed/FLOPS": 203809305335150.7, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.164183616638184, "optim/lr": 0.002996931521279447, "optim/total_tokens": 1604583424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359885, "loss/out": 6.142571926116943, "created_at": "2025-01-14T21:37:58.856498+00:00"} {"global_step": 6122, "acc_step": 0, "speed/wps": 12979.548874052072, "speed/FLOPS": 203861650836195.78, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.377573013305664, "optim/lr": 0.0029969260412800957, "optim/total_tokens": 1604845568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 407007, "loss/out": 6.049811363220215, "created_at": "2025-01-14T21:38:03.909271+00:00"} {"global_step": 6123, "acc_step": 0, "speed/wps": 12978.002977858785, "speed/FLOPS": 203837370412196.22, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 27.00505256652832, "optim/lr": 0.00299692055639676, "optim/total_tokens": 1605107712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445170, "loss/out": 6.086804389953613, "created_at": "2025-01-14T21:38:08.969615+00:00"} {"global_step": 6124, "acc_step": 0, "speed/wps": 12974.294129958525, "speed/FLOPS": 203779117859431.56, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5394949913024902, "optim/lr": 0.0029969150666294578, "optim/total_tokens": 1605369856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 6.074612617492676, "created_at": "2025-01-14T21:38:14.029786+00:00"} {"global_step": 6125, "acc_step": 0, "speed/wps": 12985.330987898229, "speed/FLOPS": 203952466879606.72, "speed/curr_iter_time": 1.272, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.225274562835693, "optim/lr": 0.0029969095719782073, "optim/total_tokens": 1605632000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461622, "loss/out": 6.209278106689453, "created_at": "2025-01-14T21:38:19.080310+00:00"} {"global_step": 6126, "acc_step": 0, "speed/wps": 12972.48205023197, "speed/FLOPS": 203750656657267.53, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 35.92799758911133, "optim/lr": 0.002996904072443025, "optim/total_tokens": 1605894144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437195, "loss/out": 6.133258819580078, "created_at": "2025-01-14T21:38:24.140102+00:00"} {"global_step": 6127, "acc_step": 0, "speed/wps": 12977.25013218903, "speed/FLOPS": 203825545936434.38, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.0223565101623535, "optim/lr": 0.002996898568023931, "optim/total_tokens": 1606156288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478609, "loss/out": 6.262204170227051, "created_at": "2025-01-14T21:38:29.193773+00:00"} {"global_step": 6128, "acc_step": 0, "speed/wps": 12973.20637364352, "speed/FLOPS": 203762033151770.56, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0012, "optim/grad_norm": 7.175487041473389, "optim/lr": 0.0029968930587209413, "optim/total_tokens": 1606418432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463750, "loss/out": 6.014376640319824, "created_at": "2025-01-14T21:38:34.249955+00:00"} {"global_step": 6129, "acc_step": 0, "speed/wps": 12979.534193545202, "speed/FLOPS": 203861420258662.25, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.374974250793457, "optim/lr": 0.0029968875445340747, "optim/total_tokens": 1606680576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439820, "loss/out": 6.143209457397461, "created_at": "2025-01-14T21:38:39.308725+00:00"} {"global_step": 6130, "acc_step": 0, "speed/wps": 12973.187356907853, "speed/FLOPS": 203761734467805.2, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.120718955993652, "optim/lr": 0.0029968820254633495, "optim/total_tokens": 1606942720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309558, "loss/out": 6.318212509155273, "created_at": "2025-01-14T21:38:44.363993+00:00"} {"global_step": 6131, "acc_step": 0, "speed/wps": 12979.122647716575, "speed/FLOPS": 203854956365900.44, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 37.72724914550781, "optim/lr": 0.0029968765015087836, "optim/total_tokens": 1607204864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 343329, "loss/out": 6.404685974121094, "created_at": "2025-01-14T21:38:49.421341+00:00"} {"global_step": 6132, "acc_step": 0, "speed/wps": 12975.449294836393, "speed/FLOPS": 203797261311201.34, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 32.44541549682617, "optim/lr": 0.002996870972670395, "optim/total_tokens": 1607467008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379658, "loss/out": 6.261993885040283, "created_at": "2025-01-14T21:38:54.476801+00:00"} {"global_step": 6133, "acc_step": 0, "speed/wps": 12975.41918438811, "speed/FLOPS": 203796788385235.0, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.63506555557251, "optim/lr": 0.0029968654389482016, "optim/total_tokens": 1607729152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315181, "loss/out": 6.401524543762207, "created_at": "2025-01-14T21:38:59.539023+00:00"} {"global_step": 6134, "acc_step": 0, "speed/wps": 12978.366198958442, "speed/FLOPS": 203843075298684.44, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.240076065063477, "optim/lr": 0.002996859900342221, "optim/total_tokens": 1607991296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451288, "loss/out": 6.215513229370117, "created_at": "2025-01-14T21:39:04.592411+00:00"} {"global_step": 6135, "acc_step": 0, "speed/wps": 12977.066406052476, "speed/FLOPS": 203822660264993.62, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.739300489425659, "optim/lr": 0.0029968543568524722, "optim/total_tokens": 1608253440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433179, "loss/out": 6.268908500671387, "created_at": "2025-01-14T21:39:09.647786+00:00"} {"global_step": 6136, "acc_step": 0, "speed/wps": 12975.37311100725, "speed/FLOPS": 203796064739477.1, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0009, "optim/grad_norm": 3.2096033096313477, "optim/lr": 0.0029968488084789724, "optim/total_tokens": 1608515584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471738, "loss/out": 6.167974472045898, "created_at": "2025-01-14T21:39:14.735814+00:00"} {"global_step": 6137, "acc_step": 0, "speed/wps": 12979.979047756715, "speed/FLOPS": 203868407305347.8, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.141213417053223, "optim/lr": 0.0029968432552217403, "optim/total_tokens": 1608777728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 6.112979888916016, "created_at": "2025-01-14T21:39:19.793708+00:00"} {"global_step": 6138, "acc_step": 0, "speed/wps": 12980.970225017047, "speed/FLOPS": 203883975106241.6, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.791863441467285, "optim/lr": 0.0029968376970807937, "optim/total_tokens": 1609039872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304748, "loss/out": 6.19523811340332, "created_at": "2025-01-14T21:39:24.860579+00:00"} {"global_step": 6139, "acc_step": 0, "speed/wps": 12978.957801614826, "speed/FLOPS": 203852367231350.2, "speed/curr_iter_time": 1.2721, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.279628276824951, "optim/lr": 0.002996832134056151, "optim/total_tokens": 1609302016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379944, "loss/out": 6.207540512084961, "created_at": "2025-01-14T21:39:29.914865+00:00"} {"global_step": 6140, "acc_step": 0, "speed/wps": 12975.25384379173, "speed/FLOPS": 203794191483967.03, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.5126028060913086, "optim/lr": 0.00299682656614783, "optim/total_tokens": 1609564160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431618, "loss/out": 6.164716720581055, "created_at": "2025-01-14T21:39:34.979342+00:00"} {"global_step": 6141, "acc_step": 0, "speed/wps": 12982.16713911028, "speed/FLOPS": 203902774286808.12, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.903453350067139, "optim/lr": 0.0029968209933558493, "optim/total_tokens": 1609826304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 6.1202239990234375, "created_at": "2025-01-14T21:39:40.033441+00:00"} {"global_step": 6142, "acc_step": 0, "speed/wps": 12975.306502927897, "speed/FLOPS": 203795018568061.78, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.735154151916504, "optim/lr": 0.002996815415680227, "optim/total_tokens": 1610088448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458734, "loss/out": 6.253553867340088, "created_at": "2025-01-14T21:39:45.094895+00:00"} {"global_step": 6143, "acc_step": 0, "speed/wps": 12979.003459982061, "speed/FLOPS": 203853084358748.78, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.7440505027770996, "optim/lr": 0.002996809833120981, "optim/total_tokens": 1610350592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468255, "loss/out": 6.193138122558594, "created_at": "2025-01-14T21:39:50.151705+00:00"} {"global_step": 6144, "acc_step": 0, "speed/wps": 12977.188422888858, "speed/FLOPS": 203824576707077.7, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.992160320281982, "optim/lr": 0.00299680424567813, "optim/total_tokens": 1610612736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 6.051885604858398, "created_at": "2025-01-14T21:39:55.206589+00:00"} {"global_step": 6145, "acc_step": 0, "speed/wps": 12980.931355849405, "speed/FLOPS": 203883364612551.97, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.230082035064697, "optim/lr": 0.002996798653351691, "optim/total_tokens": 1610874880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 6.130734443664551, "created_at": "2025-01-14T21:40:00.258920+00:00"} {"global_step": 6146, "acc_step": 0, "speed/wps": 12985.017253920132, "speed/FLOPS": 203947539256365.03, "speed/curr_iter_time": 1.272, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.226585388183594, "optim/lr": 0.002996793056141684, "optim/total_tokens": 1611137024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 6.063049793243408, "created_at": "2025-01-14T21:40:05.309947+00:00"} {"global_step": 6147, "acc_step": 0, "speed/wps": 12977.438104463852, "speed/FLOPS": 203828498299311.4, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.001, "optim/grad_norm": 2.736722707748413, "optim/lr": 0.0029967874540481262, "optim/total_tokens": 1611399168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496385, "loss/out": 6.186540603637695, "created_at": "2025-01-14T21:40:10.366970+00:00"} {"global_step": 6148, "acc_step": 0, "speed/wps": 12977.63934149413, "speed/FLOPS": 203831659003400.6, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.128276824951172, "optim/lr": 0.002996781847071036, "optim/total_tokens": 1611661312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 6.054220199584961, "created_at": "2025-01-14T21:40:15.421194+00:00"} {"global_step": 6149, "acc_step": 0, "speed/wps": 12983.232923275871, "speed/FLOPS": 203919513891669.7, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.4698991775512695, "optim/lr": 0.0029967762352104315, "optim/total_tokens": 1611923456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423627, "loss/out": 6.2395172119140625, "created_at": "2025-01-14T21:40:20.472571+00:00"} {"global_step": 6150, "acc_step": 0, "speed/wps": 12979.187803550201, "speed/FLOPS": 203855979727801.03, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.614684581756592, "optim/lr": 0.0029967706184663317, "optim/total_tokens": 1612185600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316975, "loss/out": 6.1896586418151855, "created_at": "2025-01-14T21:40:25.531323+00:00"} {"global_step": 6151, "acc_step": 0, "speed/wps": 12988.194821389132, "speed/FLOPS": 203997447319900.97, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.477599620819092, "optim/lr": 0.002996764996838754, "optim/total_tokens": 1612447744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 343848, "loss/out": 6.094367027282715, "created_at": "2025-01-14T21:40:30.580738+00:00"} {"global_step": 6152, "acc_step": 0, "speed/wps": 12983.258973774085, "speed/FLOPS": 203919923050541.28, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.4989051818847656, "optim/lr": 0.002996759370327718, "optim/total_tokens": 1612709888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 405382, "loss/out": 6.1934428215026855, "created_at": "2025-01-14T21:40:35.634858+00:00"} {"global_step": 6153, "acc_step": 0, "speed/wps": 12987.97443168634, "speed/FLOPS": 203993985796771.44, "speed/curr_iter_time": 1.2721, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.1770405769348145, "optim/lr": 0.0029967537389332414, "optim/total_tokens": 1612972032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441971, "loss/out": 5.9860734939575195, "created_at": "2025-01-14T21:40:40.694820+00:00"} {"global_step": 6154, "acc_step": 0, "speed/wps": 12986.03826044878, "speed/FLOPS": 203963575566907.47, "speed/curr_iter_time": 1.2723, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.888414144515991, "optim/lr": 0.002996748102655342, "optim/total_tokens": 1613234176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471976, "loss/out": 6.058721542358398, "created_at": "2025-01-14T21:40:45.749227+00:00"} {"global_step": 6155, "acc_step": 0, "speed/wps": 12977.036962968874, "speed/FLOPS": 203822197820906.6, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.103189468383789, "optim/lr": 0.002996742461494039, "optim/total_tokens": 1613496320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460556, "loss/out": 6.072052001953125, "created_at": "2025-01-14T21:40:50.805507+00:00"} {"global_step": 6156, "acc_step": 0, "speed/wps": 12981.03712593214, "speed/FLOPS": 203885025877044.62, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.005383491516113, "optim/lr": 0.0029967368154493502, "optim/total_tokens": 1613758464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 410019, "loss/out": 6.158651828765869, "created_at": "2025-01-14T21:40:55.862583+00:00"} {"global_step": 6157, "acc_step": 0, "speed/wps": 12980.779988138336, "speed/FLOPS": 203880987174648.6, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0008, "optim/grad_norm": 11.873878479003906, "optim/lr": 0.0029967311645212945, "optim/total_tokens": 1614020608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 6.127445220947266, "created_at": "2025-01-14T21:41:00.919663+00:00"} {"global_step": 6158, "acc_step": 0, "speed/wps": 12983.381095112518, "speed/FLOPS": 203921841133973.66, "speed/curr_iter_time": 1.2719, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.118778705596924, "optim/lr": 0.0029967255087098897, "optim/total_tokens": 1614282752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453660, "loss/out": 6.16641092300415, "created_at": "2025-01-14T21:41:05.970999+00:00"} {"global_step": 6159, "acc_step": 0, "speed/wps": 12985.371817177705, "speed/FLOPS": 203953108159546.22, "speed/curr_iter_time": 1.2718, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.769044399261475, "optim/lr": 0.0029967198480151547, "optim/total_tokens": 1614544896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 6.130092620849609, "created_at": "2025-01-14T21:41:11.021547+00:00"} {"global_step": 6160, "acc_step": 0, "speed/wps": 12979.387730254444, "speed/FLOPS": 203859119851419.25, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.8331665992736816, "optim/lr": 0.002996714182437109, "optim/total_tokens": 1614807040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 6.099715232849121, "created_at": "2025-01-14T21:41:16.074415+00:00"} {"global_step": 6161, "acc_step": 0, "speed/wps": 12984.71639916328, "speed/FLOPS": 203942813918990.94, "speed/curr_iter_time": 1.2723, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.39497447013855, "optim/lr": 0.002996708511975769, "optim/total_tokens": 1615069184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 6.095970153808594, "created_at": "2025-01-14T21:41:21.130304+00:00"} {"global_step": 6162, "acc_step": 0, "speed/wps": 12986.032099105216, "speed/FLOPS": 203963478794540.22, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.3671517372131348, "optim/lr": 0.0029967028366311545, "optim/total_tokens": 1615331328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452330, "loss/out": 5.974403381347656, "created_at": "2025-01-14T21:41:26.180818+00:00"} {"global_step": 6163, "acc_step": 0, "speed/wps": 12981.75658543104, "speed/FLOPS": 203896325977115.8, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0007, "optim/grad_norm": 24.568815231323242, "optim/lr": 0.002996697156403284, "optim/total_tokens": 1615593472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320810, "loss/out": 6.076220512390137, "created_at": "2025-01-14T21:41:31.236046+00:00"} {"global_step": 6164, "acc_step": 0, "speed/wps": 12981.748390093344, "speed/FLOPS": 203896197258075.72, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.1216654777526855, "optim/lr": 0.002996691471292176, "optim/total_tokens": 1615855616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311028, "loss/out": 6.283452987670898, "created_at": "2025-01-14T21:41:36.295308+00:00"} {"global_step": 6165, "acc_step": 0, "speed/wps": 12975.644885757516, "speed/FLOPS": 203800333335387.7, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.853054046630859, "optim/lr": 0.0029966857812978483, "optim/total_tokens": 1616117760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447791, "loss/out": 6.034232139587402, "created_at": "2025-01-14T21:41:41.353537+00:00"} {"global_step": 6166, "acc_step": 0, "speed/wps": 12984.937040116407, "speed/FLOPS": 203946279388352.22, "speed/curr_iter_time": 1.2712, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.520346641540527, "optim/lr": 0.0029966800864203204, "optim/total_tokens": 1616379904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450227, "loss/out": 6.170623779296875, "created_at": "2025-01-14T21:41:46.428513+00:00"} {"global_step": 6167, "acc_step": 0, "speed/wps": 12983.586898001817, "speed/FLOPS": 203925073551152.9, "speed/curr_iter_time": 1.2723, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.154084205627441, "optim/lr": 0.00299667438665961, "optim/total_tokens": 1616642048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 6.140930652618408, "created_at": "2025-01-14T21:41:51.480844+00:00"} {"global_step": 6168, "acc_step": 0, "speed/wps": 12980.47719386797, "speed/FLOPS": 203876231374549.56, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.782052993774414, "optim/lr": 0.0029966686820157365, "optim/total_tokens": 1616904192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 6.09372615814209, "created_at": "2025-01-14T21:41:56.533487+00:00"} {"global_step": 6169, "acc_step": 0, "speed/wps": 12978.871354900866, "speed/FLOPS": 203851009466917.22, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0014, "optim/grad_norm": 44.9553337097168, "optim/lr": 0.0029966629724887182, "optim/total_tokens": 1617166336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 6.239388465881348, "created_at": "2025-01-14T21:42:01.591325+00:00"} {"global_step": 6170, "acc_step": 0, "speed/wps": 12986.623533261458, "speed/FLOPS": 203972768080679.7, "speed/curr_iter_time": 1.2715, "speed/data_load_time": 0.0007, "optim/grad_norm": 30.90731430053711, "optim/lr": 0.002996657258078574, "optim/total_tokens": 1617428480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447839, "loss/out": 6.210570812225342, "created_at": "2025-01-14T21:42:06.646091+00:00"} {"global_step": 6171, "acc_step": 0, "speed/wps": 12985.971389454207, "speed/FLOPS": 203962525266047.3, "speed/curr_iter_time": 1.2723, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.380681037902832, "optim/lr": 0.0029966515387853214, "optim/total_tokens": 1617690624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 6.0840163230896, "created_at": "2025-01-14T21:42:11.702699+00:00"} {"global_step": 6172, "acc_step": 0, "speed/wps": 12984.318983549842, "speed/FLOPS": 203936571960675.6, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0008, "optim/grad_norm": 11.422895431518555, "optim/lr": 0.0029966458146089806, "optim/total_tokens": 1617952768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311717, "loss/out": 6.2615203857421875, "created_at": "2025-01-14T21:42:16.754316+00:00"} {"global_step": 6173, "acc_step": 0, "speed/wps": 12986.75450915015, "speed/FLOPS": 203974825236991.34, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.44729471206665, "optim/lr": 0.00299664008554957, "optim/total_tokens": 1618214912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444644, "loss/out": 6.472910404205322, "created_at": "2025-01-14T21:42:21.805543+00:00"} {"global_step": 6174, "acc_step": 0, "speed/wps": 12983.212070975389, "speed/FLOPS": 203919186377636.53, "speed/curr_iter_time": 1.2715, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.186310291290283, "optim/lr": 0.0029966343516071073, "optim/total_tokens": 1618477056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325381, "loss/out": 6.269583702087402, "created_at": "2025-01-14T21:42:26.858189+00:00"} {"global_step": 6175, "acc_step": 0, "speed/wps": 12988.075498140977, "speed/FLOPS": 203995573184321.38, "speed/curr_iter_time": 1.2715, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.4018681049346924, "optim/lr": 0.0029966286127816122, "optim/total_tokens": 1618739200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318319, "loss/out": 6.169571399688721, "created_at": "2025-01-14T21:42:31.910953+00:00"} {"global_step": 6176, "acc_step": 0, "speed/wps": 12992.107654395986, "speed/FLOPS": 204058903739069.6, "speed/curr_iter_time": 1.2713, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.6855263710021973, "optim/lr": 0.0029966228690731025, "optim/total_tokens": 1619001344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297534, "loss/out": 6.161762237548828, "created_at": "2025-01-14T21:42:36.960957+00:00"} {"global_step": 6177, "acc_step": 0, "speed/wps": 12990.309398805886, "speed/FLOPS": 204030659663964.97, "speed/curr_iter_time": 1.2722, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.825745582580566, "optim/lr": 0.0029966171204815987, "optim/total_tokens": 1619263488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345608, "loss/out": 6.163435935974121, "created_at": "2025-01-14T21:42:42.010786+00:00"} {"global_step": 6178, "acc_step": 0, "speed/wps": 12979.100190327837, "speed/FLOPS": 203854603641750.62, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.776419639587402, "optim/lr": 0.002996611367007117, "optim/total_tokens": 1619525632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401173, "loss/out": 6.159378528594971, "created_at": "2025-01-14T21:42:47.064906+00:00"} {"global_step": 6179, "acc_step": 0, "speed/wps": 12985.796110387364, "speed/FLOPS": 203959772267443.88, "speed/curr_iter_time": 1.2717, "speed/data_load_time": 0.0009, "optim/grad_norm": 5.461127281188965, "optim/lr": 0.002996605608649678, "optim/total_tokens": 1619787776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449411, "loss/out": 6.136960506439209, "created_at": "2025-01-14T21:42:52.118642+00:00"} {"global_step": 6180, "acc_step": 0, "speed/wps": 12989.543704124562, "speed/FLOPS": 204018633376820.0, "speed/curr_iter_time": 1.2709, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5762991905212402, "optim/lr": 0.0029965998454093003, "optim/total_tokens": 1620049920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441679, "loss/out": 6.07028865814209, "created_at": "2025-01-14T21:42:57.168037+00:00"} {"global_step": 6181, "acc_step": 0, "speed/wps": 12977.77454013432, "speed/FLOPS": 203833782483827.2, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.934822082519531, "optim/lr": 0.002996594077286002, "optim/total_tokens": 1620312064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 6.0315446853637695, "created_at": "2025-01-14T21:43:02.229269+00:00"} {"global_step": 6182, "acc_step": 0, "speed/wps": 12965.569839712523, "speed/FLOPS": 203642090892687.0, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.981549263000488, "optim/lr": 0.0029965883042798024, "optim/total_tokens": 1620574208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433993, "loss/out": 6.080204963684082, "created_at": "2025-01-14T21:43:07.291191+00:00"} {"global_step": 6183, "acc_step": 0, "speed/wps": 12978.965399750025, "speed/FLOPS": 203852486570504.4, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.79193115234375, "optim/lr": 0.00299658252639072, "optim/total_tokens": 1620836352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 6.078099250793457, "created_at": "2025-01-14T21:43:12.349176+00:00"} {"global_step": 6184, "acc_step": 0, "speed/wps": 12968.916645871444, "speed/FLOPS": 203694657082406.47, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.669025421142578, "optim/lr": 0.002996576743618774, "optim/total_tokens": 1621098496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443306, "loss/out": 6.019630432128906, "created_at": "2025-01-14T21:43:17.406124+00:00"} {"global_step": 6185, "acc_step": 0, "speed/wps": 12973.762724535756, "speed/FLOPS": 203770771407039.4, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.041361093521118, "optim/lr": 0.0029965709559639834, "optim/total_tokens": 1621360640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431605, "loss/out": 6.086139678955078, "created_at": "2025-01-14T21:43:22.465292+00:00"} {"global_step": 6186, "acc_step": 0, "speed/wps": 12972.894181913687, "speed/FLOPS": 203757129751657.16, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1379427909851074, "optim/lr": 0.002996565163426367, "optim/total_tokens": 1621622784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 263853, "loss/out": 6.002058029174805, "created_at": "2025-01-14T21:43:27.529062+00:00"} {"global_step": 6187, "acc_step": 0, "speed/wps": 12971.18313911962, "speed/FLOPS": 203730255473357.16, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.972646713256836, "optim/lr": 0.002996559366005943, "optim/total_tokens": 1621884928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295714, "loss/out": 6.077452182769775, "created_at": "2025-01-14T21:43:32.586542+00:00"} {"global_step": 6188, "acc_step": 0, "speed/wps": 12971.198831264579, "speed/FLOPS": 203730501940057.0, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0009, "optim/grad_norm": 4.15703821182251, "optim/lr": 0.0029965535637027315, "optim/total_tokens": 1622147072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 389608, "loss/out": 6.285181999206543, "created_at": "2025-01-14T21:43:37.644653+00:00"} {"global_step": 6189, "acc_step": 0, "speed/wps": 12973.386795035938, "speed/FLOPS": 203764866917663.5, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0007, "optim/grad_norm": 32.81759262084961, "optim/lr": 0.00299654775651675, "optim/total_tokens": 1622409216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433708, "loss/out": 6.160680770874023, "created_at": "2025-01-14T21:43:42.699747+00:00"} {"global_step": 6190, "acc_step": 0, "speed/wps": 12971.108055014853, "speed/FLOPS": 203729076174315.4, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0009, "optim/grad_norm": 5.545276165008545, "optim/lr": 0.0029965419444480195, "optim/total_tokens": 1622671360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 6.43986701965332, "created_at": "2025-01-14T21:43:47.761417+00:00"} {"global_step": 6191, "acc_step": 0, "speed/wps": 12970.517421154864, "speed/FLOPS": 203719799457928.3, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.422321319580078, "optim/lr": 0.002996536127496556, "optim/total_tokens": 1622933504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429961, "loss/out": 6.192131996154785, "created_at": "2025-01-14T21:43:52.828716+00:00"} {"global_step": 6192, "acc_step": 0, "speed/wps": 12970.08008579859, "speed/FLOPS": 203712930505196.28, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.7627668380737305, "optim/lr": 0.0029965303056623815, "optim/total_tokens": 1623195648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479707, "loss/out": 5.981708526611328, "created_at": "2025-01-14T21:43:57.888083+00:00"} {"global_step": 6193, "acc_step": 0, "speed/wps": 12977.286047583033, "speed/FLOPS": 203826110037050.1, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.061781883239746, "optim/lr": 0.0029965244789455134, "optim/total_tokens": 1623457792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438531, "loss/out": 6.062521457672119, "created_at": "2025-01-14T21:44:02.943487+00:00"} {"global_step": 6194, "acc_step": 0, "speed/wps": 12973.352599398404, "speed/FLOPS": 203764329828186.28, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0009, "optim/grad_norm": 10.072757720947266, "optim/lr": 0.0029965186473459708, "optim/total_tokens": 1623719936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287670, "loss/out": 6.111443519592285, "created_at": "2025-01-14T21:44:08.000276+00:00"} {"global_step": 6195, "acc_step": 0, "speed/wps": 12970.57714484331, "speed/FLOPS": 203720737500524.3, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.626454472541809, "optim/lr": 0.002996512810863773, "optim/total_tokens": 1623982080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432352, "loss/out": 6.034791946411133, "created_at": "2025-01-14T21:44:13.061877+00:00"} {"global_step": 6196, "acc_step": 0, "speed/wps": 12965.675361996624, "speed/FLOPS": 203643748265161.62, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.40724778175354, "optim/lr": 0.002996506969498939, "optim/total_tokens": 1624244224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 6.072691917419434, "created_at": "2025-01-14T21:44:18.125049+00:00"} {"global_step": 6197, "acc_step": 0, "speed/wps": 12969.152458902845, "speed/FLOPS": 203698360850105.66, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.6548439264297485, "optim/lr": 0.002996501123251488, "optim/total_tokens": 1624506368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 6.2065653800964355, "created_at": "2025-01-14T21:44:23.189913+00:00"} {"global_step": 6198, "acc_step": 0, "speed/wps": 12966.762419620432, "speed/FLOPS": 203660821998917.25, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.959568977355957, "optim/lr": 0.0029964952721214387, "optim/total_tokens": 1624768512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 6.169280052185059, "created_at": "2025-01-14T21:44:28.252661+00:00"} {"global_step": 6199, "acc_step": 0, "speed/wps": 12966.96911054067, "speed/FLOPS": 203664068363842.66, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0008, "optim/grad_norm": 9.949462890625, "optim/lr": 0.002996489416108811, "optim/total_tokens": 1625030656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 6.138646125793457, "created_at": "2025-01-14T21:44:33.310186+00:00"} {"global_step": 6200, "acc_step": 0, "speed/wps": 12969.189201935344, "speed/FLOPS": 203698937949921.28, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.472738265991211, "optim/lr": 0.0029964835552136223, "optim/total_tokens": 1625292800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447311, "loss/out": 6.146787643432617, "created_at": "2025-01-14T21:44:38.367005+00:00"} {"global_step": 6201, "acc_step": 0, "speed/wps": 12967.55574598024, "speed/FLOPS": 203673282279542.7, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7362940311431885, "optim/lr": 0.0029964776894358934, "optim/total_tokens": 1625554944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432127, "loss/out": 6.083520412445068, "created_at": "2025-01-14T21:44:43.426827+00:00"} {"global_step": 6202, "acc_step": 0, "speed/wps": 12976.614787822247, "speed/FLOPS": 203815566980099.4, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.91122817993164, "optim/lr": 0.002996471818775643, "optim/total_tokens": 1625817088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 283388, "loss/out": 6.139719009399414, "created_at": "2025-01-14T21:44:48.481714+00:00"} {"global_step": 6203, "acc_step": 0, "speed/wps": 12975.082842116162, "speed/FLOPS": 203791505667671.3, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0008, "optim/grad_norm": 17.364038467407227, "optim/lr": 0.002996465943232889, "optim/total_tokens": 1626079232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368187, "loss/out": 6.290201187133789, "created_at": "2025-01-14T21:44:53.562907+00:00"} {"global_step": 6204, "acc_step": 0, "speed/wps": 12978.946758720267, "speed/FLOPS": 203852193787516.62, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.703066349029541, "optim/lr": 0.002996460062807653, "optim/total_tokens": 1626341376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 6.318163871765137, "created_at": "2025-01-14T21:44:58.620743+00:00"} {"global_step": 6205, "acc_step": 0, "speed/wps": 12976.769506805183, "speed/FLOPS": 203817997054332.56, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.567302942276001, "optim/lr": 0.0029964541774999526, "optim/total_tokens": 1626603520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431842, "loss/out": 6.085058689117432, "created_at": "2025-01-14T21:45:03.677244+00:00"} {"global_step": 6206, "acc_step": 0, "speed/wps": 12978.462836026694, "speed/FLOPS": 203844593116634.6, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.404224872589111, "optim/lr": 0.0029964482873098067, "optim/total_tokens": 1626865664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 6.129837989807129, "created_at": "2025-01-14T21:45:08.730671+00:00"} {"global_step": 6207, "acc_step": 0, "speed/wps": 12969.567720831332, "speed/FLOPS": 203704883109322.25, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.612360954284668, "optim/lr": 0.0029964423922372355, "optim/total_tokens": 1627127808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 6.1436004638671875, "created_at": "2025-01-14T21:45:13.787762+00:00"} {"global_step": 6208, "acc_step": 0, "speed/wps": 12967.31142916819, "speed/FLOPS": 203669444948282.97, "speed/curr_iter_time": 1.2716, "speed/data_load_time": 0.0009, "optim/grad_norm": 3.9738872051239014, "optim/lr": 0.002996436492282258, "optim/total_tokens": 1627389952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329272, "loss/out": 6.067346572875977, "created_at": "2025-01-14T21:45:18.846000+00:00"} {"global_step": 6209, "acc_step": 0, "speed/wps": 12977.506898000709, "speed/FLOPS": 203829578796339.97, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.7784621715545654, "optim/lr": 0.0029964305874448927, "optim/total_tokens": 1627652096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297534, "loss/out": 6.087841987609863, "created_at": "2025-01-14T21:45:23.900535+00:00"} {"global_step": 6210, "acc_step": 0, "speed/wps": 12979.266684710623, "speed/FLOPS": 203857218664819.53, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.450496792793274, "optim/lr": 0.00299642467772516, "optim/total_tokens": 1627914240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368187, "loss/out": 5.999028205871582, "created_at": "2025-01-14T21:45:28.955230+00:00"} {"global_step": 6211, "acc_step": 0, "speed/wps": 12979.743994836406, "speed/FLOPS": 203864715476237.8, "speed/curr_iter_time": 1.2717, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.951274871826172, "optim/lr": 0.002996418763123078, "optim/total_tokens": 1628176384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454481, "loss/out": 6.039255142211914, "created_at": "2025-01-14T21:45:34.008569+00:00"} {"global_step": 6212, "acc_step": 0, "speed/wps": 12977.8895574109, "speed/FLOPS": 203835588988206.16, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8469083309173584, "optim/lr": 0.002996412843638667, "optim/total_tokens": 1628438528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445170, "loss/out": 6.183903694152832, "created_at": "2025-01-14T21:45:39.061875+00:00"} {"global_step": 6213, "acc_step": 0, "speed/wps": 12979.843774450723, "speed/FLOPS": 203866282652192.2, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.8262059688568115, "optim/lr": 0.0029964069192719464, "optim/total_tokens": 1628700672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427020, "loss/out": 5.941205978393555, "created_at": "2025-01-14T21:45:44.115137+00:00"} {"global_step": 6214, "acc_step": 0, "speed/wps": 12979.828552190414, "speed/FLOPS": 203866043565676.47, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 17.624692916870117, "optim/lr": 0.0029964009900229345, "optim/total_tokens": 1628962816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448890, "loss/out": 6.068142414093018, "created_at": "2025-01-14T21:45:49.169559+00:00"} {"global_step": 6215, "acc_step": 0, "speed/wps": 12978.828117612811, "speed/FLOPS": 203850330365894.06, "speed/curr_iter_time": 1.2719, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5278511047363281, "optim/lr": 0.002996395055891651, "optim/total_tokens": 1629224960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447791, "loss/out": 6.158777713775635, "created_at": "2025-01-14T21:45:54.223855+00:00"} {"global_step": 6216, "acc_step": 0, "speed/wps": 12978.977117748496, "speed/FLOPS": 203852670617772.62, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.757416248321533, "optim/lr": 0.0029963891168781156, "optim/total_tokens": 1629487104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 6.052227973937988, "created_at": "2025-01-14T21:45:59.282091+00:00"} {"global_step": 6217, "acc_step": 0, "speed/wps": 12982.044267020981, "speed/FLOPS": 203900844411801.78, "speed/curr_iter_time": 1.2719, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.5884740352630615, "optim/lr": 0.0029963831729823477, "optim/total_tokens": 1629749248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440399, "loss/out": 6.090612411499023, "created_at": "2025-01-14T21:46:04.337994+00:00"} {"global_step": 6218, "acc_step": 0, "speed/wps": 12975.195880955962, "speed/FLOPS": 203793281097982.03, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.5919768810272217, "optim/lr": 0.002996377224204367, "optim/total_tokens": 1630011392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444644, "loss/out": 6.015283584594727, "created_at": "2025-01-14T21:46:09.392631+00:00"} {"global_step": 6219, "acc_step": 0, "speed/wps": 12978.083414782574, "speed/FLOPS": 203838633784616.25, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.668346405029297, "optim/lr": 0.002996371270544192, "optim/total_tokens": 1630273536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 6.119199752807617, "created_at": "2025-01-14T21:46:14.451720+00:00"} {"global_step": 6220, "acc_step": 0, "speed/wps": 12973.931447708715, "speed/FLOPS": 203773421436320.66, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.0541064739227295, "optim/lr": 0.0029963653120018423, "optim/total_tokens": 1630535680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438240, "loss/out": 5.992660999298096, "created_at": "2025-01-14T21:46:19.506658+00:00"} {"global_step": 6221, "acc_step": 0, "speed/wps": 12977.088760676688, "speed/FLOPS": 203823011375085.34, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.24427318572998, "optim/lr": 0.0029963593485773377, "optim/total_tokens": 1630797824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327688, "loss/out": 6.022269248962402, "created_at": "2025-01-14T21:46:24.565953+00:00"} {"global_step": 6222, "acc_step": 0, "speed/wps": 12980.677878647228, "speed/FLOPS": 203879383404777.6, "speed/curr_iter_time": 1.2722, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.0990447998046875, "optim/lr": 0.002996353380270698, "optim/total_tokens": 1631059968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338880, "loss/out": 6.157681941986084, "created_at": "2025-01-14T21:46:29.624057+00:00"} {"global_step": 6223, "acc_step": 0, "speed/wps": 12976.025377571123, "speed/FLOPS": 203806309482170.9, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 53.209136962890625, "optim/lr": 0.0029963474070819414, "optim/total_tokens": 1631322112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445933, "loss/out": 6.008678436279297, "created_at": "2025-01-14T21:46:34.680634+00:00"} {"global_step": 6224, "acc_step": 0, "speed/wps": 12977.884198429352, "speed/FLOPS": 203835504818036.94, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.85412073135376, "optim/lr": 0.002996341429011089, "optim/total_tokens": 1631584256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445695, "loss/out": 6.06135368347168, "created_at": "2025-01-14T21:46:39.739283+00:00"} {"global_step": 6225, "acc_step": 0, "speed/wps": 12981.226364572522, "speed/FLOPS": 203887998129932.97, "speed/curr_iter_time": 1.2721, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.691247940063477, "optim/lr": 0.0029963354460581593, "optim/total_tokens": 1631846400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473241, "loss/out": 6.2259979248046875, "created_at": "2025-01-14T21:46:44.798784+00:00"} {"global_step": 6226, "acc_step": 0, "speed/wps": 12973.610753771225, "speed/FLOPS": 203768384497352.72, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.70958423614502, "optim/lr": 0.002996329458223172, "optim/total_tokens": 1632108544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 6.000018119812012, "created_at": "2025-01-14T21:46:49.854995+00:00"} {"global_step": 6227, "acc_step": 0, "speed/wps": 12984.059885369972, "speed/FLOPS": 203932502467722.53, "speed/curr_iter_time": 1.2718, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9907078742980957, "optim/lr": 0.002996323465506146, "optim/total_tokens": 1632370688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460840, "loss/out": 6.012442588806152, "created_at": "2025-01-14T21:46:54.906953+00:00"} {"global_step": 6228, "acc_step": 0, "speed/wps": 12971.879406688291, "speed/FLOPS": 203741191312295.03, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0008, "optim/grad_norm": 37.30521011352539, "optim/lr": 0.0029963174679071025, "optim/total_tokens": 1632632832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417192, "loss/out": 6.043086528778076, "created_at": "2025-01-14T21:46:59.966419+00:00"} {"global_step": 6229, "acc_step": 0, "speed/wps": 12980.568015101831, "speed/FLOPS": 203877657846828.8, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.198162317276001, "optim/lr": 0.00299631146542606, "optim/total_tokens": 1632894976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 6.104248046875, "created_at": "2025-01-14T21:47:05.023752+00:00"} {"global_step": 6230, "acc_step": 0, "speed/wps": 12973.762181563123, "speed/FLOPS": 203770762878908.06, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.2803494930267334, "optim/lr": 0.0029963054580630383, "optim/total_tokens": 1633157120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442208, "loss/out": 6.112060546875, "created_at": "2025-01-14T21:47:10.078899+00:00"} {"global_step": 6231, "acc_step": 0, "speed/wps": 12966.916339848798, "speed/FLOPS": 203663239527612.28, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0009, "optim/grad_norm": 6.243349552154541, "optim/lr": 0.0029962994458180565, "optim/total_tokens": 1633419264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332731, "loss/out": 6.1020684242248535, "created_at": "2025-01-14T21:47:15.163636+00:00"} {"global_step": 6232, "acc_step": 0, "speed/wps": 12976.318648643059, "speed/FLOPS": 203810915707352.34, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.23129653930664, "optim/lr": 0.002996293428691135, "optim/total_tokens": 1633681408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445933, "loss/out": 6.1050848960876465, "created_at": "2025-01-14T21:47:20.218311+00:00"} {"global_step": 6233, "acc_step": 0, "speed/wps": 12977.026098042892, "speed/FLOPS": 203822027172314.1, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.6022210121154785, "optim/lr": 0.0029962874066822927, "optim/total_tokens": 1633943552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490524, "loss/out": 6.004274845123291, "created_at": "2025-01-14T21:47:25.271993+00:00"} {"global_step": 6234, "acc_step": 0, "speed/wps": 12974.73791692195, "speed/FLOPS": 203786088143519.3, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.075249195098877, "optim/lr": 0.0029962813797915493, "optim/total_tokens": 1634205696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 6.0382399559021, "created_at": "2025-01-14T21:47:30.327051+00:00"} {"global_step": 6235, "acc_step": 0, "speed/wps": 12976.863679140908, "speed/FLOPS": 203819476160271.47, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.499084234237671, "optim/lr": 0.002996275348018925, "optim/total_tokens": 1634467840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420954, "loss/out": 6.015758514404297, "created_at": "2025-01-14T21:47:35.382331+00:00"} {"global_step": 6236, "acc_step": 0, "speed/wps": 12975.044855115502, "speed/FLOPS": 203790909029626.34, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.762844562530518, "optim/lr": 0.0029962693113644396, "optim/total_tokens": 1634729984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469590, "loss/out": 5.989292144775391, "created_at": "2025-01-14T21:47:40.442658+00:00"} {"global_step": 6237, "acc_step": 0, "speed/wps": 12977.08448167832, "speed/FLOPS": 203822944167536.0, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.745729684829712, "optim/lr": 0.002996263269828112, "optim/total_tokens": 1634992128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460271, "loss/out": 6.003922462463379, "created_at": "2025-01-14T21:47:45.496678+00:00"} {"global_step": 6238, "acc_step": 0, "speed/wps": 12961.809252370025, "speed/FLOPS": 203583025700884.53, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 16.165817260742188, "optim/lr": 0.002996257223409963, "optim/total_tokens": 1635254272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445933, "loss/out": 5.97960090637207, "created_at": "2025-01-14T21:47:50.563173+00:00"} {"global_step": 6239, "acc_step": 0, "speed/wps": 12978.750967705204, "speed/FLOPS": 203849118620580.0, "speed/curr_iter_time": 1.2723, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.500544548034668, "optim/lr": 0.0029962511721100104, "optim/total_tokens": 1635516416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292060, "loss/out": 5.950362682342529, "created_at": "2025-01-14T21:47:55.623768+00:00"} {"global_step": 6240, "acc_step": 0, "speed/wps": 12970.203936396096, "speed/FLOPS": 203714875749018.66, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.142741680145264, "optim/lr": 0.002996245115928276, "optim/total_tokens": 1635778560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311065, "loss/out": 6.047708511352539, "created_at": "2025-01-14T21:48:00.681362+00:00"} {"global_step": 6241, "acc_step": 0, "speed/wps": 12975.096589072713, "speed/FLOPS": 203791721582514.25, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0011, "optim/grad_norm": 10.818022727966309, "optim/lr": 0.0029962390548647785, "optim/total_tokens": 1636040704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368718, "loss/out": 6.021733283996582, "created_at": "2025-01-14T21:48:05.740885+00:00"} {"global_step": 6242, "acc_step": 0, "speed/wps": 12962.689719161119, "speed/FLOPS": 203596854641726.8, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.725473880767822, "optim/lr": 0.0029962329889195384, "optim/total_tokens": 1636302848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427880, "loss/out": 6.062224388122559, "created_at": "2025-01-14T21:48:10.806559+00:00"} {"global_step": 6243, "acc_step": 0, "speed/wps": 12976.236400660682, "speed/FLOPS": 203809623889768.56, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 25.220949172973633, "optim/lr": 0.002996226918092575, "optim/total_tokens": 1636564992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 5.999964714050293, "created_at": "2025-01-14T21:48:15.861786+00:00"} {"global_step": 6244, "acc_step": 0, "speed/wps": 12975.528855519718, "speed/FLOPS": 203798510921060.8, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0006, "optim/grad_norm": 27.062925338745117, "optim/lr": 0.0029962208423839074, "optim/total_tokens": 1636827136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446505, "loss/out": 6.041210174560547, "created_at": "2025-01-14T21:48:20.920222+00:00"} {"global_step": 6245, "acc_step": 0, "speed/wps": 12969.938824847992, "speed/FLOPS": 203710711807854.25, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.906180381774902, "optim/lr": 0.002996214761793557, "optim/total_tokens": 1637089280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 6.210594654083252, "created_at": "2025-01-14T21:48:25.979297+00:00"} {"global_step": 6246, "acc_step": 0, "speed/wps": 12973.3388848386, "speed/FLOPS": 203764114422178.78, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 46.66609573364258, "optim/lr": 0.002996208676321543, "optim/total_tokens": 1637351424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 6.105508804321289, "created_at": "2025-01-14T21:48:31.034795+00:00"} {"global_step": 6247, "acc_step": 0, "speed/wps": 12968.016331422117, "speed/FLOPS": 203680516406818.06, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.990225791931152, "optim/lr": 0.002996202585967884, "optim/total_tokens": 1637613568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445695, "loss/out": 6.165958881378174, "created_at": "2025-01-14T21:48:36.092366+00:00"} {"global_step": 6248, "acc_step": 0, "speed/wps": 12975.683830942335, "speed/FLOPS": 203800945023031.56, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 41.82106018066406, "optim/lr": 0.0029961964907326013, "optim/total_tokens": 1637875712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295655, "loss/out": 6.081255912780762, "created_at": "2025-01-14T21:48:41.149118+00:00"} {"global_step": 6249, "acc_step": 0, "speed/wps": 12975.654178793835, "speed/FLOPS": 203800479295293.03, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 54.69305419921875, "optim/lr": 0.002996190390615715, "optim/total_tokens": 1638137856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365501, "loss/out": 6.103522300720215, "created_at": "2025-01-14T21:48:46.203938+00:00"} {"global_step": 6250, "acc_step": 0, "speed/wps": 12971.005865708736, "speed/FLOPS": 203727471150840.03, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 2322.22900390625, "optim/lr": 0.0029961842856172437, "optim/total_tokens": 1638400000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449657, "loss/out": 6.082671642303467, "created_at": "2025-01-14T21:48:51.265338+00:00"} {"global_step": 6251, "acc_step": 0, "speed/wps": 12971.63407806601, "speed/FLOPS": 203737338089165.25, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 45.90630340576172, "optim/lr": 0.0029961781757372086, "optim/total_tokens": 1638662144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468255, "loss/out": 6.130939483642578, "created_at": "2025-01-14T21:48:56.325335+00:00"} {"global_step": 6252, "acc_step": 0, "speed/wps": 12972.98865197754, "speed/FLOPS": 203758613533825.44, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.4469313621521, "optim/lr": 0.0029961720609756287, "optim/total_tokens": 1638924288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476746, "loss/out": 6.167808532714844, "created_at": "2025-01-14T21:49:01.380773+00:00"} {"global_step": 6253, "acc_step": 0, "speed/wps": 12967.186704278916, "speed/FLOPS": 203667485972506.94, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.230378150939941, "optim/lr": 0.0029961659413325244, "optim/total_tokens": 1639186432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444881, "loss/out": 6.231935501098633, "created_at": "2025-01-14T21:49:06.443422+00:00"} {"global_step": 6254, "acc_step": 0, "speed/wps": 12975.379881909928, "speed/FLOPS": 203796171085807.88, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 15.352641105651855, "optim/lr": 0.0029961598168079156, "optim/total_tokens": 1639448576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 6.026767730712891, "created_at": "2025-01-14T21:49:11.503232+00:00"} {"global_step": 6255, "acc_step": 0, "speed/wps": 12965.813163962188, "speed/FLOPS": 203645912634390.53, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 166.68893432617188, "optim/lr": 0.002996153687401822, "optim/total_tokens": 1639710720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425201, "loss/out": 6.20479679107666, "created_at": "2025-01-14T21:49:16.585533+00:00"} {"global_step": 6256, "acc_step": 0, "speed/wps": 12965.025578313638, "speed/FLOPS": 203633542519523.97, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.829636573791504, "optim/lr": 0.0029961475531142642, "optim/total_tokens": 1639972864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333702, "loss/out": 6.147656440734863, "created_at": "2025-01-14T21:49:21.649157+00:00"} {"global_step": 6257, "acc_step": 0, "speed/wps": 12975.174185997117, "speed/FLOPS": 203792940348841.38, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 26.282804489135742, "optim/lr": 0.002996141413945262, "optim/total_tokens": 1640235008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359599, "loss/out": 6.191046237945557, "created_at": "2025-01-14T21:49:26.708036+00:00"} {"global_step": 6258, "acc_step": 0, "speed/wps": 12976.912874152427, "speed/FLOPS": 203820248835528.4, "speed/curr_iter_time": 1.272, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.330073833465576, "optim/lr": 0.0029961352698948352, "optim/total_tokens": 1640497152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438531, "loss/out": 6.095076560974121, "created_at": "2025-01-14T21:49:31.766502+00:00"} {"global_step": 6259, "acc_step": 0, "speed/wps": 12972.843419770448, "speed/FLOPS": 203756332462443.1, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.696232795715332, "optim/lr": 0.002996129120963004, "optim/total_tokens": 1640759296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457902, "loss/out": 6.13394832611084, "created_at": "2025-01-14T21:49:36.821789+00:00"} {"global_step": 6260, "acc_step": 0, "speed/wps": 12965.719357396183, "speed/FLOPS": 203644439273367.72, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.001, "optim/grad_norm": 3.962963581085205, "optim/lr": 0.002996122967149788, "optim/total_tokens": 1641021440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 5.972805023193359, "created_at": "2025-01-14T21:49:41.885058+00:00"} {"global_step": 6261, "acc_step": 0, "speed/wps": 12961.366905632858, "speed/FLOPS": 203576078037529.22, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.224967002868652, "optim/lr": 0.0029961168084552087, "optim/total_tokens": 1641283584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449657, "loss/out": 6.100117206573486, "created_at": "2025-01-14T21:49:46.946203+00:00"} {"global_step": 6262, "acc_step": 0, "speed/wps": 12971.502247296095, "speed/FLOPS": 203735267505769.28, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 15.446555137634277, "optim/lr": 0.0029961106448792843, "optim/total_tokens": 1641545728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 5.984097480773926, "created_at": "2025-01-14T21:49:52.007322+00:00"} {"global_step": 6263, "acc_step": 0, "speed/wps": 12969.364905443843, "speed/FLOPS": 203701697614964.2, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 18.43486213684082, "optim/lr": 0.002996104476422036, "optim/total_tokens": 1641807872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440921, "loss/out": 6.184535980224609, "created_at": "2025-01-14T21:49:57.068352+00:00"} {"global_step": 6264, "acc_step": 0, "speed/wps": 12966.809920459127, "speed/FLOPS": 203661568064861.94, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.8973989486694336, "optim/lr": 0.0029960983030834836, "optim/total_tokens": 1642070016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316287, "loss/out": 6.118180274963379, "created_at": "2025-01-14T21:50:02.126042+00:00"} {"global_step": 6265, "acc_step": 0, "speed/wps": 12970.819832226285, "speed/FLOPS": 203724549239361.34, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.216846942901611, "optim/lr": 0.0029960921248636473, "optim/total_tokens": 1642332160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322156, "loss/out": 5.984777927398682, "created_at": "2025-01-14T21:50:07.187779+00:00"} {"global_step": 6266, "acc_step": 0, "speed/wps": 12968.548040563777, "speed/FLOPS": 203688867629532.78, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.4801409244537354, "optim/lr": 0.0029960859417625476, "optim/total_tokens": 1642594304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345194, "loss/out": 6.284543991088867, "created_at": "2025-01-14T21:50:12.254383+00:00"} {"global_step": 6267, "acc_step": 0, "speed/wps": 12964.79924059919, "speed/FLOPS": 203629987574697.94, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.162507057189941, "optim/lr": 0.0029960797537802045, "optim/total_tokens": 1642856448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435582, "loss/out": 5.963872909545898, "created_at": "2025-01-14T21:50:17.312952+00:00"} {"global_step": 6268, "acc_step": 0, "speed/wps": 12966.660838763375, "speed/FLOPS": 203659226531970.44, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 48.98746109008789, "optim/lr": 0.002996073560916638, "optim/total_tokens": 1643118592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468018, "loss/out": 6.077445983886719, "created_at": "2025-01-14T21:50:22.375898+00:00"} {"global_step": 6269, "acc_step": 0, "speed/wps": 12971.490927760107, "speed/FLOPS": 203735089716901.9, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.222421646118164, "optim/lr": 0.0029960673631718683, "optim/total_tokens": 1643380736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458173, "loss/out": 6.284546375274658, "created_at": "2025-01-14T21:50:27.436855+00:00"} {"global_step": 6270, "acc_step": 0, "speed/wps": 12963.890453167014, "speed/FLOPS": 203615713819270.84, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.428269386291504, "optim/lr": 0.0029960611605459155, "optim/total_tokens": 1643642880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438240, "loss/out": 6.168173313140869, "created_at": "2025-01-14T21:50:32.495531+00:00"} {"global_step": 6271, "acc_step": 0, "speed/wps": 12969.75094306116, "speed/FLOPS": 203707760866215.44, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.185873508453369, "optim/lr": 0.0029960549530388004, "optim/total_tokens": 1643905024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 6.141700267791748, "created_at": "2025-01-14T21:50:37.553233+00:00"} {"global_step": 6272, "acc_step": 0, "speed/wps": 12968.804768066735, "speed/FLOPS": 203692899887747.28, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.256172180175781, "optim/lr": 0.002996048740650543, "optim/total_tokens": 1644167168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 6.0915021896362305, "created_at": "2025-01-14T21:50:42.610107+00:00"} {"global_step": 6273, "acc_step": 0, "speed/wps": 12972.520950126749, "speed/FLOPS": 203751267633569.1, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.702693462371826, "optim/lr": 0.002996042523381163, "optim/total_tokens": 1644429312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432656, "loss/out": 6.137118339538574, "created_at": "2025-01-14T21:50:47.666787+00:00"} {"global_step": 6274, "acc_step": 0, "speed/wps": 12966.0438279107, "speed/FLOPS": 203649535528667.4, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.7084708213806152, "optim/lr": 0.002996036301230681, "optim/total_tokens": 1644691456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293306, "loss/out": 6.059698581695557, "created_at": "2025-01-14T21:50:52.726033+00:00"} {"global_step": 6275, "acc_step": 0, "speed/wps": 12963.676258633108, "speed/FLOPS": 203612349599781.7, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.332670211791992, "optim/lr": 0.0029960300741991184, "optim/total_tokens": 1644953600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401173, "loss/out": 6.128957748413086, "created_at": "2025-01-14T21:50:57.786476+00:00"} {"global_step": 6276, "acc_step": 0, "speed/wps": 12965.463342089637, "speed/FLOPS": 203640418201176.5, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.663461685180664, "optim/lr": 0.002996023842286494, "optim/total_tokens": 1645215744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434807, "loss/out": 6.054908752441406, "created_at": "2025-01-14T21:51:02.850753+00:00"} {"global_step": 6277, "acc_step": 0, "speed/wps": 12972.880929906669, "speed/FLOPS": 203756921610676.78, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 34.975624084472656, "optim/lr": 0.0029960176054928287, "optim/total_tokens": 1645477888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 6.035604476928711, "created_at": "2025-01-14T21:51:07.910090+00:00"} {"global_step": 6278, "acc_step": 0, "speed/wps": 12967.228372964035, "speed/FLOPS": 203668140436465.5, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.365376949310303, "optim/lr": 0.002996011363818143, "optim/total_tokens": 1645740032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427351, "loss/out": 6.110798358917236, "created_at": "2025-01-14T21:51:12.971998+00:00"} {"global_step": 6279, "acc_step": 0, "speed/wps": 12965.816840664438, "speed/FLOPS": 203645970382051.7, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0014, "optim/grad_norm": 17.466808319091797, "optim/lr": 0.0029960051172624575, "optim/total_tokens": 1646002176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435329, "loss/out": 6.084654808044434, "created_at": "2025-01-14T21:51:18.036736+00:00"} {"global_step": 6280, "acc_step": 0, "speed/wps": 12972.247790078607, "speed/FLOPS": 203746977279651.78, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 53.50687026977539, "optim/lr": 0.0029959988658257914, "optim/total_tokens": 1646264320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429193, "loss/out": 6.154507637023926, "created_at": "2025-01-14T21:51:23.098489+00:00"} {"global_step": 6281, "acc_step": 0, "speed/wps": 12973.478677121828, "speed/FLOPS": 203766310052081.6, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.596488952636719, "optim/lr": 0.0029959926095081664, "optim/total_tokens": 1646526464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470925, "loss/out": 6.135114669799805, "created_at": "2025-01-14T21:51:28.153536+00:00"} {"global_step": 6282, "acc_step": 0, "speed/wps": 12976.805649273832, "speed/FLOPS": 203818564721467.66, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0012, "optim/grad_norm": 5.861518383026123, "optim/lr": 0.002995986348309602, "optim/total_tokens": 1646788608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 6.178691864013672, "created_at": "2025-01-14T21:51:33.207836+00:00"} {"global_step": 6283, "acc_step": 0, "speed/wps": 12958.37057294682, "speed/FLOPS": 203529016515302.4, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.3520779609680176, "optim/lr": 0.00299598008223012, "optim/total_tokens": 1647050752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435866, "loss/out": 5.98097038269043, "created_at": "2025-01-14T21:51:38.274074+00:00"} {"global_step": 6284, "acc_step": 0, "speed/wps": 12974.472313790713, "speed/FLOPS": 203781916481366.3, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.715178966522217, "optim/lr": 0.0029959738112697388, "optim/total_tokens": 1647312896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298880, "loss/out": 6.113733291625977, "created_at": "2025-01-14T21:51:43.331714+00:00"} {"global_step": 6285, "acc_step": 0, "speed/wps": 12968.836009429171, "speed/FLOPS": 203693390576274.47, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.1783127784729004, "optim/lr": 0.0029959675354284806, "optim/total_tokens": 1647575040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413664, "loss/out": 6.0548906326293945, "created_at": "2025-01-14T21:51:48.388886+00:00"} {"global_step": 6286, "acc_step": 0, "speed/wps": 12962.566752250927, "speed/FLOPS": 203594923277428.0, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.163478136062622, "optim/lr": 0.0029959612547063646, "optim/total_tokens": 1647837184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 6.0387163162231445, "created_at": "2025-01-14T21:51:53.448375+00:00"} {"global_step": 6287, "acc_step": 0, "speed/wps": 12973.901727024748, "speed/FLOPS": 203772954632138.44, "speed/curr_iter_time": 1.2723, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.007836818695068, "optim/lr": 0.0029959549691034124, "optim/total_tokens": 1648099328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470403, "loss/out": 6.065479278564453, "created_at": "2025-01-14T21:51:58.505472+00:00"} {"global_step": 6288, "acc_step": 0, "speed/wps": 12975.023060180272, "speed/FLOPS": 203790566710219.2, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.4060232639312744, "optim/lr": 0.0029959486786196436, "optim/total_tokens": 1648361472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461077, "loss/out": 6.0516815185546875, "created_at": "2025-01-14T21:52:03.563942+00:00"} {"global_step": 6289, "acc_step": 0, "speed/wps": 12969.049578563341, "speed/FLOPS": 203696744973001.34, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.132818222045898, "optim/lr": 0.00299594238325508, "optim/total_tokens": 1648623616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452615, "loss/out": 6.00052547454834, "created_at": "2025-01-14T21:52:08.621225+00:00"} {"global_step": 6290, "acc_step": 0, "speed/wps": 12969.137801904253, "speed/FLOPS": 203698130641801.97, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.325066089630127, "optim/lr": 0.002995936083009741, "optim/total_tokens": 1648885760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 6.143503189086914, "created_at": "2025-01-14T21:52:13.687550+00:00"} {"global_step": 6291, "acc_step": 0, "speed/wps": 12967.471403414509, "speed/FLOPS": 203671957563649.03, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.713128566741943, "optim/lr": 0.002995929777883647, "optim/total_tokens": 1649147904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420140, "loss/out": 6.008446216583252, "created_at": "2025-01-14T21:52:18.749748+00:00"} {"global_step": 6292, "acc_step": 0, "speed/wps": 12973.378892495868, "speed/FLOPS": 203764742797412.47, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0413296222686768, "optim/lr": 0.0029959234678768186, "optim/total_tokens": 1649410048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319909, "loss/out": 6.083910942077637, "created_at": "2025-01-14T21:52:23.805329+00:00"} {"global_step": 6293, "acc_step": 0, "speed/wps": 12970.416221348361, "speed/FLOPS": 203718209975904.1, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.5129475593566895, "optim/lr": 0.0029959171529892778, "optim/total_tokens": 1649672192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336239, "loss/out": 6.1111040115356445, "created_at": "2025-01-14T21:52:28.865925+00:00"} {"global_step": 6294, "acc_step": 0, "speed/wps": 12971.522595730068, "speed/FLOPS": 203735587105886.5, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.462512969970703, "optim/lr": 0.002995910833221044, "optim/total_tokens": 1649934336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357905, "loss/out": 6.169095039367676, "created_at": "2025-01-14T21:52:33.922922+00:00"} {"global_step": 6295, "acc_step": 0, "speed/wps": 12966.121034675662, "speed/FLOPS": 203650748167004.34, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5128753185272217, "optim/lr": 0.0029959045085721377, "optim/total_tokens": 1650196480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426734, "loss/out": 6.1866888999938965, "created_at": "2025-01-14T21:52:38.984322+00:00"} {"global_step": 6296, "acc_step": 0, "speed/wps": 12972.86245702237, "speed/FLOPS": 203756631468644.44, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.094350814819336, "optim/lr": 0.00299589817904258, "optim/total_tokens": 1650458624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443545, "loss/out": 5.974499702453613, "created_at": "2025-01-14T21:52:44.045396+00:00"} {"global_step": 6297, "acc_step": 0, "speed/wps": 12972.37608232415, "speed/FLOPS": 203748992285660.03, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.1796369552612305, "optim/lr": 0.0029958918446323915, "optim/total_tokens": 1650720768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 6.138072967529297, "created_at": "2025-01-14T21:52:49.106518+00:00"} {"global_step": 6298, "acc_step": 0, "speed/wps": 12966.795913887916, "speed/FLOPS": 203661348072414.0, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3704833984375, "optim/lr": 0.002995885505341592, "optim/total_tokens": 1650982912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447554, "loss/out": 6.153703689575195, "created_at": "2025-01-14T21:52:54.167031+00:00"} {"global_step": 6299, "acc_step": 0, "speed/wps": 12971.064766624851, "speed/FLOPS": 203728396270665.97, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.744919776916504, "optim/lr": 0.0029958791611702043, "optim/total_tokens": 1651245056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482002, "loss/out": 6.115081787109375, "created_at": "2025-01-14T21:52:59.227176+00:00"} {"global_step": 6300, "acc_step": 0, "speed/wps": 12967.759400939107, "speed/FLOPS": 203676480960522.9, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.52234411239624, "optim/lr": 0.002995872812118247, "optim/total_tokens": 1651507200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429977, "loss/out": 6.015860080718994, "created_at": "2025-01-14T21:53:04.284690+00:00"} {"global_step": 6301, "acc_step": 0, "speed/wps": 12972.413874340726, "speed/FLOPS": 203749585861212.03, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.001, "optim/grad_norm": 4.202273845672607, "optim/lr": 0.0029958664581857413, "optim/total_tokens": 1651769344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329272, "loss/out": 6.013514995574951, "created_at": "2025-01-14T21:53:09.348460+00:00"} {"global_step": 6302, "acc_step": 0, "speed/wps": 12958.671973994542, "speed/FLOPS": 203533750432925.3, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.48915433883667, "optim/lr": 0.0029958600993727088, "optim/total_tokens": 1652031488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 410747, "loss/out": 6.043933868408203, "created_at": "2025-01-14T21:53:14.410575+00:00"} {"global_step": 6303, "acc_step": 0, "speed/wps": 12961.985011444613, "speed/FLOPS": 203585786238669.22, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.698657989501953, "optim/lr": 0.002995853735679169, "optim/total_tokens": 1652293632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444067, "loss/out": 6.106081008911133, "created_at": "2025-01-14T21:53:19.473552+00:00"} {"global_step": 6304, "acc_step": 0, "speed/wps": 12974.570261245217, "speed/FLOPS": 203783454880733.2, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.286355495452881, "optim/lr": 0.002995847367105144, "optim/total_tokens": 1652555776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478081, "loss/out": 5.975247859954834, "created_at": "2025-01-14T21:53:24.530593+00:00"} {"global_step": 6305, "acc_step": 0, "speed/wps": 12972.856281867445, "speed/FLOPS": 203756534479350.8, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.3533782958984375, "optim/lr": 0.002995840993650654, "optim/total_tokens": 1652817920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468540, "loss/out": 6.068034648895264, "created_at": "2025-01-14T21:53:29.587737+00:00"} {"global_step": 6306, "acc_step": 0, "speed/wps": 12971.512380856055, "speed/FLOPS": 203735426667253.4, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.366638660430908, "optim/lr": 0.002995834615315719, "optim/total_tokens": 1653080064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450990, "loss/out": 6.162988185882568, "created_at": "2025-01-14T21:53:34.647749+00:00"} {"global_step": 6307, "acc_step": 0, "speed/wps": 12965.35949545308, "speed/FLOPS": 203638787147049.12, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0009, "optim/grad_norm": 12.611629486083984, "optim/lr": 0.002995828232100361, "optim/total_tokens": 1653342208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478811, "loss/out": 5.991193771362305, "created_at": "2025-01-14T21:53:39.714754+00:00"} {"global_step": 6308, "acc_step": 0, "speed/wps": 12968.844111987504, "speed/FLOPS": 203693517838087.2, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.91878890991211, "optim/lr": 0.0029958218440046, "optim/total_tokens": 1653604352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432637, "loss/out": 5.9990434646606445, "created_at": "2025-01-14T21:53:44.776664+00:00"} {"global_step": 6309, "acc_step": 0, "speed/wps": 12972.121011700245, "speed/FLOPS": 203744986051007.84, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.278491497039795, "optim/lr": 0.0029958154510284574, "optim/total_tokens": 1653866496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 6.097196578979492, "created_at": "2025-01-14T21:53:49.834735+00:00"} {"global_step": 6310, "acc_step": 0, "speed/wps": 12969.616913129928, "speed/FLOPS": 203705655741969.0, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.8048255443573, "optim/lr": 0.002995809053171954, "optim/total_tokens": 1654128640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459765, "loss/out": 5.969883918762207, "created_at": "2025-01-14T21:53:54.895268+00:00"} {"global_step": 6311, "acc_step": 0, "speed/wps": 12971.297470523188, "speed/FLOPS": 203732051205158.0, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.4752261638641357, "optim/lr": 0.00299580265043511, "optim/total_tokens": 1654390784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308499, "loss/out": 6.039103031158447, "created_at": "2025-01-14T21:53:59.955580+00:00"} {"global_step": 6312, "acc_step": 0, "speed/wps": 12976.11662704659, "speed/FLOPS": 203807742680573.03, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.14625883102417, "optim/lr": 0.002995796242817947, "optim/total_tokens": 1654652928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364683, "loss/out": 5.903134822845459, "created_at": "2025-01-14T21:54:05.011172+00:00"} {"global_step": 6313, "acc_step": 0, "speed/wps": 12967.110280525081, "speed/FLOPS": 203666285632435.72, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0009, "optim/grad_norm": 4.5218377113342285, "optim/lr": 0.0029957898303204855, "optim/total_tokens": 1654915072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433708, "loss/out": 6.010846138000488, "created_at": "2025-01-14T21:54:10.073371+00:00"} {"global_step": 6314, "acc_step": 0, "speed/wps": 12966.429636784147, "speed/FLOPS": 203655595187181.84, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0009, "optim/grad_norm": 5.8582868576049805, "optim/lr": 0.002995783412942747, "optim/total_tokens": 1655177216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 6.115327835083008, "created_at": "2025-01-14T21:54:15.131823+00:00"} {"global_step": 6315, "acc_step": 0, "speed/wps": 12974.013788468548, "speed/FLOPS": 203774714711101.9, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.877936601638794, "optim/lr": 0.002995776990684752, "optim/total_tokens": 1655439360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451797, "loss/out": 5.964485168457031, "created_at": "2025-01-14T21:54:20.188472+00:00"} {"global_step": 6316, "acc_step": 0, "speed/wps": 12970.830842168927, "speed/FLOPS": 203724722165639.78, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.970659017562866, "optim/lr": 0.002995770563546522, "optim/total_tokens": 1655701504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436666, "loss/out": 5.970871925354004, "created_at": "2025-01-14T21:54:25.245744+00:00"} {"global_step": 6317, "acc_step": 0, "speed/wps": 12962.802334417294, "speed/FLOPS": 203598623419074.62, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.556118965148926, "optim/lr": 0.0029957641315280763, "optim/total_tokens": 1655963648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 6.094961166381836, "created_at": "2025-01-14T21:54:30.310627+00:00"} {"global_step": 6318, "acc_step": 0, "speed/wps": 12974.447193224485, "speed/FLOPS": 203781521928354.44, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.95761775970459, "optim/lr": 0.0029957576946294377, "optim/total_tokens": 1656225792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447026, "loss/out": 5.967138767242432, "created_at": "2025-01-14T21:54:35.365493+00:00"} {"global_step": 6319, "acc_step": 0, "speed/wps": 12970.366064316362, "speed/FLOPS": 203717422190793.56, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.2630889415740967, "optim/lr": 0.0029957512528506263, "optim/total_tokens": 1656487936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 6.039385795593262, "created_at": "2025-01-14T21:54:40.423249+00:00"} {"global_step": 6320, "acc_step": 0, "speed/wps": 12972.431109563076, "speed/FLOPS": 203749856564062.75, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 17.129371643066406, "optim/lr": 0.002995744806191664, "optim/total_tokens": 1656750080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 5.921148777008057, "created_at": "2025-01-14T21:54:45.482910+00:00"} {"global_step": 6321, "acc_step": 0, "speed/wps": 12969.82433332542, "speed/FLOPS": 203708913561164.4, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.562906265258789, "optim/lr": 0.0029957383546525703, "optim/total_tokens": 1657012224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444881, "loss/out": 5.9587082862854, "created_at": "2025-01-14T21:54:50.539750+00:00"} {"global_step": 6322, "acc_step": 0, "speed/wps": 12969.025769701042, "speed/FLOPS": 203696371022102.16, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 29.632957458496094, "optim/lr": 0.002995731898233368, "optim/total_tokens": 1657274368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331148, "loss/out": 6.024410247802734, "created_at": "2025-01-14T21:54:55.596652+00:00"} {"global_step": 6323, "acc_step": 0, "speed/wps": 12968.59851097228, "speed/FLOPS": 203689660336651.94, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0012, "optim/grad_norm": 6.087108612060547, "optim/lr": 0.002995725436934077, "optim/total_tokens": 1657536512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296822, "loss/out": 6.06909704208374, "created_at": "2025-01-14T21:55:00.656221+00:00"} {"global_step": 6324, "acc_step": 0, "speed/wps": 12976.569481277387, "speed/FLOPS": 203814855378554.66, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.483516216278076, "optim/lr": 0.002995718970754719, "optim/total_tokens": 1657798656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 399773, "loss/out": 6.063515663146973, "created_at": "2025-01-14T21:55:05.718632+00:00"} {"global_step": 6325, "acc_step": 0, "speed/wps": 12973.333400457874, "speed/FLOPS": 203764028282443.16, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.707810401916504, "optim/lr": 0.0029957124996953144, "optim/total_tokens": 1658060800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446503, "loss/out": 6.0106658935546875, "created_at": "2025-01-14T21:55:10.774704+00:00"} {"global_step": 6326, "acc_step": 0, "speed/wps": 12973.703227443744, "speed/FLOPS": 203769836923451.78, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.008575439453125, "optim/lr": 0.0029957060237558847, "optim/total_tokens": 1658322944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 5.9089250564575195, "created_at": "2025-01-14T21:55:15.832543+00:00"} {"global_step": 6327, "acc_step": 0, "speed/wps": 12967.54633549878, "speed/FLOPS": 203673134475000.0, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.8425815105438232, "optim/lr": 0.0029956995429364507, "optim/total_tokens": 1658585088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435866, "loss/out": 6.033908367156982, "created_at": "2025-01-14T21:55:20.892360+00:00"} {"global_step": 6328, "acc_step": 0, "speed/wps": 12964.531617448456, "speed/FLOPS": 203625784185362.94, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.802809715270996, "optim/lr": 0.0029956930572370352, "optim/total_tokens": 1658847232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472260, "loss/out": 5.877091884613037, "created_at": "2025-01-14T21:55:25.978783+00:00"} {"global_step": 6329, "acc_step": 0, "speed/wps": 12971.300554966916, "speed/FLOPS": 203732099650585.22, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.2089524269104004, "optim/lr": 0.0029956865666576566, "optim/total_tokens": 1659109376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349068, "loss/out": 6.108852863311768, "created_at": "2025-01-14T21:55:31.039752+00:00"} {"global_step": 6330, "acc_step": 0, "speed/wps": 12963.689642300287, "speed/FLOPS": 203612559808665.84, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.761420488357544, "optim/lr": 0.002995680071198338, "optim/total_tokens": 1659371520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451797, "loss/out": 5.95439338684082, "created_at": "2025-01-14T21:55:36.101802+00:00"} {"global_step": 6331, "acc_step": 0, "speed/wps": 12967.891179539567, "speed/FLOPS": 203678550724525.9, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.4510610103607178, "optim/lr": 0.002995673570859101, "optim/total_tokens": 1659633664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 5.923595428466797, "created_at": "2025-01-14T21:55:41.159166+00:00"} {"global_step": 6332, "acc_step": 0, "speed/wps": 12969.866346464116, "speed/FLOPS": 203709573435239.75, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 89.09947967529297, "optim/lr": 0.002995667065639965, "optim/total_tokens": 1659895808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 6.025911808013916, "created_at": "2025-01-14T21:55:46.216801+00:00"} {"global_step": 6333, "acc_step": 0, "speed/wps": 12965.815703490549, "speed/FLOPS": 203645952521173.12, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0009, "optim/grad_norm": 14.65315055847168, "optim/lr": 0.002995660555540952, "optim/total_tokens": 1660157952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 6.030935287475586, "created_at": "2025-01-14T21:55:51.279236+00:00"} {"global_step": 6334, "acc_step": 0, "speed/wps": 12965.797100214115, "speed/FLOPS": 203645660331153.16, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.40731430053711, "optim/lr": 0.002995654040562084, "optim/total_tokens": 1660420096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426252, "loss/out": 6.009326934814453, "created_at": "2025-01-14T21:55:56.341203+00:00"} {"global_step": 6335, "acc_step": 0, "speed/wps": 12969.20401441762, "speed/FLOPS": 203699170600309.94, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.529124736785889, "optim/lr": 0.0029956475207033816, "optim/total_tokens": 1660682240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 6.1873369216918945, "created_at": "2025-01-14T21:56:01.402102+00:00"} {"global_step": 6336, "acc_step": 0, "speed/wps": 12969.94517173538, "speed/FLOPS": 203710811494442.22, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.9207682609558105, "optim/lr": 0.002995640995964866, "optim/total_tokens": 1660944384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459237, "loss/out": 6.06522274017334, "created_at": "2025-01-14T21:56:06.458788+00:00"} {"global_step": 6337, "acc_step": 0, "speed/wps": 12967.310451833404, "speed/FLOPS": 203669429597897.12, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.6751222610473633, "optim/lr": 0.0029956344663465586, "optim/total_tokens": 1661206528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319751, "loss/out": 6.003661155700684, "created_at": "2025-01-14T21:56:11.516461+00:00"} {"global_step": 6338, "acc_step": 0, "speed/wps": 12964.063572211844, "speed/FLOPS": 203618432891760.38, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 57.93326950073242, "optim/lr": 0.00299562793184848, "optim/total_tokens": 1661468672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330329, "loss/out": 6.096017360687256, "created_at": "2025-01-14T21:56:16.578669+00:00"} {"global_step": 6339, "acc_step": 0, "speed/wps": 12970.827592978698, "speed/FLOPS": 203724671132642.38, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.41198444366455, "optim/lr": 0.0029956213924706536, "optim/total_tokens": 1661730816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441439, "loss/out": 6.080071449279785, "created_at": "2025-01-14T21:56:21.634833+00:00"} {"global_step": 6340, "acc_step": 0, "speed/wps": 12972.055823971266, "speed/FLOPS": 203743962188146.9, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.536863327026367, "optim/lr": 0.002995614848213099, "optim/total_tokens": 1661992960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426015, "loss/out": 6.166164875030518, "created_at": "2025-01-14T21:56:26.722037+00:00"} {"global_step": 6341, "acc_step": 0, "speed/wps": 12968.662299228197, "speed/FLOPS": 203690662218865.34, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.1385178565979, "optim/lr": 0.0029956082990758373, "optim/total_tokens": 1662255104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468255, "loss/out": 6.072596073150635, "created_at": "2025-01-14T21:56:31.784070+00:00"} {"global_step": 6342, "acc_step": 0, "speed/wps": 12968.597428281457, "speed/FLOPS": 203689643331504.78, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.001, "optim/grad_norm": 9.908744812011719, "optim/lr": 0.002995601745058891, "optim/total_tokens": 1662517248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453144, "loss/out": 6.068261623382568, "created_at": "2025-01-14T21:56:36.844076+00:00"} {"global_step": 6343, "acc_step": 0, "speed/wps": 12969.59291573234, "speed/FLOPS": 203705278829864.0, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.260128021240234, "optim/lr": 0.0029955951861622807, "optim/total_tokens": 1662779392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332206, "loss/out": 6.0210113525390625, "created_at": "2025-01-14T21:56:41.900696+00:00"} {"global_step": 6344, "acc_step": 0, "speed/wps": 12972.70957703034, "speed/FLOPS": 203754230278290.5, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.8724544048309326, "optim/lr": 0.002995588622386028, "optim/total_tokens": 1663041536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292485, "loss/out": 6.12017297744751, "created_at": "2025-01-14T21:56:46.959937+00:00"} {"global_step": 6345, "acc_step": 0, "speed/wps": 12962.916186201368, "speed/FLOPS": 203600411617792.25, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 9.523233413696289, "optim/lr": 0.002995582053730155, "optim/total_tokens": 1663303680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401352, "loss/out": 5.945474624633789, "created_at": "2025-01-14T21:56:52.058322+00:00"} {"global_step": 6346, "acc_step": 0, "speed/wps": 12974.58958585173, "speed/FLOPS": 203783758400233.28, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.437425374984741, "optim/lr": 0.0029955754801946816, "optim/total_tokens": 1663565824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 5.977014541625977, "created_at": "2025-01-14T21:56:57.114298+00:00"} {"global_step": 6347, "acc_step": 0, "speed/wps": 12969.051684340835, "speed/FLOPS": 203696778047130.53, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.248004913330078, "optim/lr": 0.0029955689017796315, "optim/total_tokens": 1663827968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454718, "loss/out": 6.1277666091918945, "created_at": "2025-01-14T21:57:02.176216+00:00"} {"global_step": 6348, "acc_step": 0, "speed/wps": 12969.84758096028, "speed/FLOPS": 203709278697215.5, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 8.984980583190918, "optim/lr": 0.0029955623184850237, "optim/total_tokens": 1664090112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 6.216676712036133, "created_at": "2025-01-14T21:57:07.232785+00:00"} {"global_step": 6349, "acc_step": 0, "speed/wps": 12970.471203272687, "speed/FLOPS": 203719073542578.72, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.480933666229248, "optim/lr": 0.0029955557303108812, "optim/total_tokens": 1664352256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450990, "loss/out": 6.169867515563965, "created_at": "2025-01-14T21:57:12.290001+00:00"} {"global_step": 6350, "acc_step": 0, "speed/wps": 12971.953822990246, "speed/FLOPS": 203742360122575.97, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 43.500732421875, "optim/lr": 0.0029955491372572252, "optim/total_tokens": 1664614400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325000, "loss/out": 6.076462268829346, "created_at": "2025-01-14T21:57:17.354198+00:00"} {"global_step": 6351, "acc_step": 0, "speed/wps": 12974.427019812862, "speed/FLOPS": 203781205077203.72, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.192316055297852, "optim/lr": 0.0029955425393240766, "optim/total_tokens": 1664876544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358303, "loss/out": 6.1760735511779785, "created_at": "2025-01-14T21:57:22.409546+00:00"} {"global_step": 6352, "acc_step": 0, "speed/wps": 12969.548248634293, "speed/FLOPS": 203704577271710.16, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.116682052612305, "optim/lr": 0.0029955359365114584, "optim/total_tokens": 1665138688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 405960, "loss/out": 6.03295373916626, "created_at": "2025-01-14T21:57:27.469887+00:00"} {"global_step": 6353, "acc_step": 0, "speed/wps": 12970.055295403905, "speed/FLOPS": 203712541137982.2, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.273341178894043, "optim/lr": 0.00299552932881939, "optim/total_tokens": 1665400832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457380, "loss/out": 6.120906829833984, "created_at": "2025-01-14T21:57:32.526372+00:00"} {"global_step": 6354, "acc_step": 0, "speed/wps": 12965.934904767424, "speed/FLOPS": 203647824741026.2, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.918029308319092, "optim/lr": 0.002995522716247895, "optim/total_tokens": 1665662976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 6.021032333374023, "created_at": "2025-01-14T21:57:37.590738+00:00"} {"global_step": 6355, "acc_step": 0, "speed/wps": 12968.120916075475, "speed/FLOPS": 203682159052513.44, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.408426284790039, "optim/lr": 0.0029955160987969937, "optim/total_tokens": 1665925120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434515, "loss/out": 5.914761543273926, "created_at": "2025-01-14T21:57:42.650257+00:00"} {"global_step": 6356, "acc_step": 0, "speed/wps": 12969.26888665241, "speed/FLOPS": 203700189507901.06, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.050201892852783, "optim/lr": 0.0029955094764667085, "optim/total_tokens": 1666187264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 6.030873775482178, "created_at": "2025-01-14T21:57:47.708427+00:00"} {"global_step": 6357, "acc_step": 0, "speed/wps": 12971.040101576975, "speed/FLOPS": 203728008872195.72, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.788314342498779, "optim/lr": 0.0029955028492570607, "optim/total_tokens": 1666449408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451547, "loss/out": 5.998250961303711, "created_at": "2025-01-14T21:57:52.765840+00:00"} {"global_step": 6358, "acc_step": 0, "speed/wps": 12968.212207367389, "speed/FLOPS": 203683592907700.3, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.316696166992188, "optim/lr": 0.0029954962171680716, "optim/total_tokens": 1666711552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448605, "loss/out": 6.012360095977783, "created_at": "2025-01-14T21:57:57.829997+00:00"} {"global_step": 6359, "acc_step": 0, "speed/wps": 12971.219558917577, "speed/FLOPS": 203730827496330.12, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.882810592651367, "optim/lr": 0.002995489580199763, "optim/total_tokens": 1666973696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316527, "loss/out": 6.042198657989502, "created_at": "2025-01-14T21:58:02.887355+00:00"} {"global_step": 6360, "acc_step": 0, "speed/wps": 12967.35101493524, "speed/FLOPS": 203670066697151.03, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.3100178241729736, "optim/lr": 0.0029954829383521562, "optim/total_tokens": 1667235840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326727, "loss/out": 5.971153259277344, "created_at": "2025-01-14T21:58:07.950856+00:00"} {"global_step": 6361, "acc_step": 0, "speed/wps": 12969.881963707627, "speed/FLOPS": 203709818725509.4, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.077502727508545, "optim/lr": 0.002995476291625274, "optim/total_tokens": 1667497984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 387212, "loss/out": 5.816418647766113, "created_at": "2025-01-14T21:58:13.010626+00:00"} {"global_step": 6362, "acc_step": 0, "speed/wps": 12973.307490229941, "speed/FLOPS": 203763621326709.1, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.6685636043548584, "optim/lr": 0.0029954696400191373, "optim/total_tokens": 1667760128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440628, "loss/out": 6.0238847732543945, "created_at": "2025-01-14T21:58:18.066493+00:00"} {"global_step": 6363, "acc_step": 0, "speed/wps": 12970.148959128252, "speed/FLOPS": 203714012255480.47, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.495007276535034, "optim/lr": 0.002995462983533768, "optim/total_tokens": 1668022272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 6.033010482788086, "created_at": "2025-01-14T21:58:23.128353+00:00"} {"global_step": 6364, "acc_step": 0, "speed/wps": 12975.954845848208, "speed/FLOPS": 203805201684541.78, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.826726913452148, "optim/lr": 0.0029954563221691872, "optim/total_tokens": 1668284416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439586, "loss/out": 6.070611000061035, "created_at": "2025-01-14T21:58:28.184650+00:00"} {"global_step": 6365, "acc_step": 0, "speed/wps": 12976.814450999324, "speed/FLOPS": 203818702964661.34, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1832497119903564, "optim/lr": 0.0029954496559254176, "optim/total_tokens": 1668546560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447269, "loss/out": 6.013702869415283, "created_at": "2025-01-14T21:58:33.242125+00:00"} {"global_step": 6366, "acc_step": 0, "speed/wps": 12965.49016938789, "speed/FLOPS": 203640839560760.12, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.118613243103027, "optim/lr": 0.0029954429848024805, "optim/total_tokens": 1668808704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 5.948002815246582, "created_at": "2025-01-14T21:58:38.301565+00:00"} {"global_step": 6367, "acc_step": 0, "speed/wps": 12968.162667241138, "speed/FLOPS": 203682814811942.28, "speed/curr_iter_time": 1.2719, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.330704212188721, "optim/lr": 0.0029954363088003973, "optim/total_tokens": 1669070848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443307, "loss/out": 6.080873966217041, "created_at": "2025-01-14T21:58:43.369239+00:00"} {"global_step": 6368, "acc_step": 0, "speed/wps": 12969.628867639143, "speed/FLOPS": 203705843503963.44, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 13.76736068725586, "optim/lr": 0.0029954296279191906, "optim/total_tokens": 1669332992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356561, "loss/out": 6.046164512634277, "created_at": "2025-01-14T21:58:48.435569+00:00"} {"global_step": 6369, "acc_step": 0, "speed/wps": 12968.868536840384, "speed/FLOPS": 203693901463964.9, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.541633367538452, "optim/lr": 0.002995422942158881, "optim/total_tokens": 1669595136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428932, "loss/out": 6.1461501121521, "created_at": "2025-01-14T21:58:53.497518+00:00"} {"global_step": 6370, "acc_step": 0, "speed/wps": 12966.313615997207, "speed/FLOPS": 203653772921293.7, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.4375345706939697, "optim/lr": 0.0029954162515194917, "optim/total_tokens": 1669857280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 6.100761890411377, "created_at": "2025-01-14T21:58:58.555586+00:00"} {"global_step": 6371, "acc_step": 0, "speed/wps": 12975.152125064069, "speed/FLOPS": 203792593851572.94, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.810023784637451, "optim/lr": 0.002995409556001044, "optim/total_tokens": 1670119424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448605, "loss/out": 5.883883953094482, "created_at": "2025-01-14T21:59:03.616595+00:00"} {"global_step": 6372, "acc_step": 0, "speed/wps": 12967.258882697548, "speed/FLOPS": 203668619633755.1, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.568452835083008, "optim/lr": 0.0029954028556035592, "optim/total_tokens": 1670381568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421768, "loss/out": 6.0540080070495605, "created_at": "2025-01-14T21:59:08.676462+00:00"} {"global_step": 6373, "acc_step": 0, "speed/wps": 12971.348782467743, "speed/FLOPS": 203732857129757.25, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 80.53137969970703, "optim/lr": 0.00299539615032706, "optim/total_tokens": 1670643712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428118, "loss/out": 6.126671314239502, "created_at": "2025-01-14T21:59:13.734533+00:00"} {"global_step": 6374, "acc_step": 0, "speed/wps": 12970.313495477392, "speed/FLOPS": 203716596524940.38, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.313906192779541, "optim/lr": 0.0029953894401715674, "optim/total_tokens": 1670905856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 6.074143409729004, "created_at": "2025-01-14T21:59:18.791190+00:00"} {"global_step": 6375, "acc_step": 0, "speed/wps": 12971.809795219351, "speed/FLOPS": 203740097968519.16, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.563766479492188, "optim/lr": 0.002995382725137104, "optim/total_tokens": 1671168000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463222, "loss/out": 5.89078426361084, "created_at": "2025-01-14T21:59:23.847104+00:00"} {"global_step": 6376, "acc_step": 0, "speed/wps": 12975.491222402956, "speed/FLOPS": 203797919841248.75, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.46318244934082, "optim/lr": 0.002995376005223692, "optim/total_tokens": 1671430144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441971, "loss/out": 5.98283576965332, "created_at": "2025-01-14T21:59:28.901403+00:00"} {"global_step": 6377, "acc_step": 0, "speed/wps": 12974.491537632268, "speed/FLOPS": 203782218418214.25, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.819829225540161, "optim/lr": 0.002995369280431351, "optim/total_tokens": 1671692288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325239, "loss/out": 6.131769180297852, "created_at": "2025-01-14T21:59:33.958803+00:00"} {"global_step": 6378, "acc_step": 0, "speed/wps": 12966.916772334951, "speed/FLOPS": 203663246320401.6, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.840930700302124, "optim/lr": 0.002995362550760107, "optim/total_tokens": 1671954432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352687, "loss/out": 6.038156986236572, "created_at": "2025-01-14T21:59:39.016822+00:00"} {"global_step": 6379, "acc_step": 0, "speed/wps": 12970.114340686012, "speed/FLOPS": 203713468525274.38, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.604785919189453, "optim/lr": 0.0029953558162099784, "optim/total_tokens": 1672216576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447032, "loss/out": 5.956853866577148, "created_at": "2025-01-14T21:59:44.073479+00:00"} {"global_step": 6380, "acc_step": 0, "speed/wps": 12971.865666548023, "speed/FLOPS": 203740975504511.2, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0008, "optim/grad_norm": 11.236658096313477, "optim/lr": 0.0029953490767809887, "optim/total_tokens": 1672478720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456567, "loss/out": 6.035478115081787, "created_at": "2025-01-14T21:59:49.137431+00:00"} {"global_step": 6381, "acc_step": 0, "speed/wps": 12971.38088491871, "speed/FLOPS": 203733361342863.06, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.6410624980926514, "optim/lr": 0.00299534233247316, "optim/total_tokens": 1672740864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457380, "loss/out": 6.078225135803223, "created_at": "2025-01-14T21:59:54.193535+00:00"} {"global_step": 6382, "acc_step": 0, "speed/wps": 12971.92948694819, "speed/FLOPS": 203741977891594.22, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 202.474365234375, "optim/lr": 0.0029953355832865137, "optim/total_tokens": 1673003008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428403, "loss/out": 6.149592876434326, "created_at": "2025-01-14T21:59:59.253083+00:00"} {"global_step": 6383, "acc_step": 0, "speed/wps": 12971.048998293878, "speed/FLOPS": 203728148607360.06, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.93590784072876, "optim/lr": 0.002995328829221072, "optim/total_tokens": 1673265152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463770, "loss/out": 6.187055587768555, "created_at": "2025-01-14T22:00:04.310268+00:00"} {"global_step": 6384, "acc_step": 0, "speed/wps": 12972.118926448002, "speed/FLOPS": 203744953299255.94, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 18.114124298095703, "optim/lr": 0.0029953220702768567, "optim/total_tokens": 1673527296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 6.148375034332275, "created_at": "2025-01-14T22:00:09.366785+00:00"} {"global_step": 6385, "acc_step": 0, "speed/wps": 12967.878659465632, "speed/FLOPS": 203678354079560.38, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.938248634338379, "optim/lr": 0.0029953153064538906, "optim/total_tokens": 1673789440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478811, "loss/out": 5.9930949211120605, "created_at": "2025-01-14T22:00:14.427299+00:00"} {"global_step": 6386, "acc_step": 0, "speed/wps": 12969.046851331534, "speed/FLOPS": 203696702138078.22, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 2863.6708984375, "optim/lr": 0.002995308537752195, "optim/total_tokens": 1674051584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493435, "loss/out": 6.038874626159668, "created_at": "2025-01-14T22:00:19.488549+00:00"} {"global_step": 6387, "acc_step": 0, "speed/wps": 12975.164970207265, "speed/FLOPS": 203792795602197.9, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.129242420196533, "optim/lr": 0.0029953017641717924, "optim/total_tokens": 1674313728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 6.053881645202637, "created_at": "2025-01-14T22:00:24.544321+00:00"} {"global_step": 6388, "acc_step": 0, "speed/wps": 12967.689807394876, "speed/FLOPS": 203675387898279.22, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.22461986541748, "optim/lr": 0.0029952949857127054, "optim/total_tokens": 1674575872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300577, "loss/out": 6.097798824310303, "created_at": "2025-01-14T22:00:29.604576+00:00"} {"global_step": 6389, "acc_step": 0, "speed/wps": 12971.633300937374, "speed/FLOPS": 203737325883292.12, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.277125835418701, "optim/lr": 0.002995288202374955, "optim/total_tokens": 1674838016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300577, "loss/out": 6.276771068572998, "created_at": "2025-01-14T22:00:34.661489+00:00"} {"global_step": 6390, "acc_step": 0, "speed/wps": 12963.090384877873, "speed/FLOPS": 203603147647380.8, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.6642045974731445, "optim/lr": 0.002995281414158564, "optim/total_tokens": 1675100160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446503, "loss/out": 6.10420036315918, "created_at": "2025-01-14T22:00:39.726746+00:00"} {"global_step": 6391, "acc_step": 0, "speed/wps": 12970.202499622848, "speed/FLOPS": 203714853182520.47, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.400146961212158, "optim/lr": 0.0029952746210635543, "optim/total_tokens": 1675362304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 6.040234088897705, "created_at": "2025-01-14T22:00:44.791136+00:00"} {"global_step": 6392, "acc_step": 0, "speed/wps": 12968.166213832224, "speed/FLOPS": 203682870516028.8, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0012, "optim/grad_norm": 4.088523864746094, "optim/lr": 0.0029952678230899484, "optim/total_tokens": 1675624448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432127, "loss/out": 5.963266372680664, "created_at": "2025-01-14T22:00:49.855814+00:00"} {"global_step": 6393, "acc_step": 0, "speed/wps": 12967.827217247228, "speed/FLOPS": 203677546108831.34, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.931513786315918, "optim/lr": 0.0029952610202377684, "optim/total_tokens": 1675886592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 6.078656196594238, "created_at": "2025-01-14T22:00:54.913561+00:00"} {"global_step": 6394, "acc_step": 0, "speed/wps": 12961.862158181697, "speed/FLOPS": 203583856659357.03, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.118459701538086, "optim/lr": 0.002995254212507036, "optim/total_tokens": 1676148736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438531, "loss/out": 5.856845378875732, "created_at": "2025-01-14T22:00:59.978023+00:00"} {"global_step": 6395, "acc_step": 0, "speed/wps": 12973.236500131989, "speed/FLOPS": 203762506329670.12, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.919438362121582, "optim/lr": 0.002995247399897774, "optim/total_tokens": 1676410880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316527, "loss/out": 6.151273727416992, "created_at": "2025-01-14T22:01:05.035827+00:00"} {"global_step": 6396, "acc_step": 0, "speed/wps": 12967.05619447131, "speed/FLOPS": 203665436136639.22, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.246230602264404, "optim/lr": 0.002995240582410004, "optim/total_tokens": 1676673024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 381239, "loss/out": 6.037361145019531, "created_at": "2025-01-14T22:01:10.096919+00:00"} {"global_step": 6397, "acc_step": 0, "speed/wps": 12961.194856500217, "speed/FLOPS": 203573375769480.78, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 15.312487602233887, "optim/lr": 0.002995233760043749, "optim/total_tokens": 1676935168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440390, "loss/out": 5.977132797241211, "created_at": "2025-01-14T22:01:15.161970+00:00"} {"global_step": 6398, "acc_step": 0, "speed/wps": 12957.045808812018, "speed/FLOPS": 203508209274148.66, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.381465911865234, "optim/lr": 0.002995226932799031, "optim/total_tokens": 1677197312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479622, "loss/out": 5.983925819396973, "created_at": "2025-01-14T22:01:20.229736+00:00"} {"global_step": 6399, "acc_step": 0, "speed/wps": 12964.364060626258, "speed/FLOPS": 203623152475223.1, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.755728244781494, "optim/lr": 0.002995220100675872, "optim/total_tokens": 1677459456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443022, "loss/out": 6.027347087860107, "created_at": "2025-01-14T22:01:25.294736+00:00"} {"global_step": 6400, "acc_step": 0, "speed/wps": 12958.87010933615, "speed/FLOPS": 203536862420731.78, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.782074928283691, "optim/lr": 0.0029952132636742942, "optim/total_tokens": 1677721600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420954, "loss/out": 5.987433433532715, "created_at": "2025-01-14T22:01:30.356853+00:00"} {"global_step": 6401, "acc_step": 0, "speed/wps": 12969.60847494733, "speed/FLOPS": 203705523208716.03, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0008, "optim/grad_norm": 193.13461303710938, "optim/lr": 0.00299520642179432, "optim/total_tokens": 1677983744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467205, "loss/out": 5.960538864135742, "created_at": "2025-01-14T22:01:35.437085+00:00"} {"global_step": 6402, "acc_step": 0, "speed/wps": 12964.027911483889, "speed/FLOPS": 203617872791024.88, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 8.35343074798584, "optim/lr": 0.002995199575035973, "optim/total_tokens": 1678245888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292430, "loss/out": 6.037735939025879, "created_at": "2025-01-14T22:01:40.498066+00:00"} {"global_step": 6403, "acc_step": 0, "speed/wps": 12966.891403845171, "speed/FLOPS": 203662847873409.66, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.017488479614258, "optim/lr": 0.0029951927233992735, "optim/total_tokens": 1678508032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 391480, "loss/out": 6.03142786026001, "created_at": "2025-01-14T22:01:45.559154+00:00"} {"global_step": 6404, "acc_step": 0, "speed/wps": 12959.170975164203, "speed/FLOPS": 203541587931994.3, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 17.601381301879883, "optim/lr": 0.0029951858668842454, "optim/total_tokens": 1678770176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 6.111556529998779, "created_at": "2025-01-14T22:01:50.621053+00:00"} {"global_step": 6405, "acc_step": 0, "speed/wps": 12962.201948464819, "speed/FLOPS": 203589193532674.25, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 20.890060424804688, "optim/lr": 0.0029951790054909095, "optim/total_tokens": 1679032320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 6.0987229347229, "created_at": "2025-01-14T22:01:55.680641+00:00"} {"global_step": 6406, "acc_step": 0, "speed/wps": 12968.452750864943, "speed/FLOPS": 203687370973870.8, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.847292423248291, "optim/lr": 0.00299517213921929, "optim/total_tokens": 1679294464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435852, "loss/out": 6.135341644287109, "created_at": "2025-01-14T22:02:00.742080+00:00"} {"global_step": 6407, "acc_step": 0, "speed/wps": 12970.069899492833, "speed/FLOPS": 203712770515266.9, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.545666694641113, "optim/lr": 0.0029951652680694075, "optim/total_tokens": 1679556608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501962, "loss/out": 5.9831719398498535, "created_at": "2025-01-14T22:02:05.799632+00:00"} {"global_step": 6408, "acc_step": 0, "speed/wps": 12958.15236345333, "speed/FLOPS": 203525589235355.88, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0009, "optim/grad_norm": 3.4937305450439453, "optim/lr": 0.0029951583920412865, "optim/total_tokens": 1679818752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431618, "loss/out": 6.018746852874756, "created_at": "2025-01-14T22:02:10.860992+00:00"} {"global_step": 6409, "acc_step": 0, "speed/wps": 12976.81859341423, "speed/FLOPS": 203818768026979.7, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.79386568069458, "optim/lr": 0.002995151511134948, "optim/total_tokens": 1680080896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326344, "loss/out": 6.066654205322266, "created_at": "2025-01-14T22:02:15.915015+00:00"} {"global_step": 6410, "acc_step": 0, "speed/wps": 12962.557578239763, "speed/FLOPS": 203594779186976.28, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 48.51903533935547, "optim/lr": 0.002995144625350414, "optim/total_tokens": 1680343040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357199, "loss/out": 6.038431167602539, "created_at": "2025-01-14T22:02:20.977625+00:00"} {"global_step": 6411, "acc_step": 0, "speed/wps": 12971.138327172488, "speed/FLOPS": 203729551640149.38, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.0817413330078125, "optim/lr": 0.002995137734687709, "optim/total_tokens": 1680605184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438536, "loss/out": 6.277499675750732, "created_at": "2025-01-14T22:02:26.034892+00:00"} {"global_step": 6412, "acc_step": 0, "speed/wps": 12968.117211777826, "speed/FLOPS": 203682100871428.62, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.56764030456543, "optim/lr": 0.002995130839146853, "optim/total_tokens": 1680867328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443022, "loss/out": 6.301443099975586, "created_at": "2025-01-14T22:02:31.093054+00:00"} {"global_step": 6413, "acc_step": 0, "speed/wps": 12969.80201089338, "speed/FLOPS": 203708562956695.0, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.639001846313477, "optim/lr": 0.0029951239387278708, "optim/total_tokens": 1681129472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432656, "loss/out": 6.253279685974121, "created_at": "2025-01-14T22:02:36.151591+00:00"} {"global_step": 6414, "acc_step": 0, "speed/wps": 12955.660941416254, "speed/FLOPS": 203486458028690.94, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0008, "optim/grad_norm": 9.347463607788086, "optim/lr": 0.0029951170334307823, "optim/total_tokens": 1681391616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 6.131274223327637, "created_at": "2025-01-14T22:02:41.213789+00:00"} {"global_step": 6415, "acc_step": 0, "speed/wps": 12965.103487191269, "speed/FLOPS": 203634766185505.2, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.974086046218872, "optim/lr": 0.002995110123255613, "optim/total_tokens": 1681653760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442256, "loss/out": 6.09736442565918, "created_at": "2025-01-14T22:02:46.273472+00:00"} {"global_step": 6416, "acc_step": 0, "speed/wps": 12968.022960818329, "speed/FLOPS": 203680620530595.3, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 89.99505615234375, "optim/lr": 0.0029951032082023835, "optim/total_tokens": 1681915904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445933, "loss/out": 6.071526527404785, "created_at": "2025-01-14T22:02:51.333364+00:00"} {"global_step": 6417, "acc_step": 0, "speed/wps": 12964.056968358533, "speed/FLOPS": 203618329169169.47, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.349595069885254, "optim/lr": 0.0029950962882711167, "optim/total_tokens": 1682178048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479331, "loss/out": 6.052727222442627, "created_at": "2025-01-14T22:02:56.392436+00:00"} {"global_step": 6418, "acc_step": 0, "speed/wps": 12968.961509394227, "speed/FLOPS": 203695361725680.7, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9799296855926514, "optim/lr": 0.0029950893634618355, "optim/total_tokens": 1682440192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451275, "loss/out": 6.030488014221191, "created_at": "2025-01-14T22:03:01.449299+00:00"} {"global_step": 6419, "acc_step": 0, "speed/wps": 12964.792401909297, "speed/FLOPS": 203629880163675.84, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9568960666656494, "optim/lr": 0.002995082433774562, "optim/total_tokens": 1682702336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293783, "loss/out": 6.0348615646362305, "created_at": "2025-01-14T22:03:06.508343+00:00"} {"global_step": 6420, "acc_step": 0, "speed/wps": 12969.955792924666, "speed/FLOPS": 203710978314815.0, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.9149627685546875, "optim/lr": 0.0029950754992093195, "optim/total_tokens": 1682964480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350820, "loss/out": 6.018794536590576, "created_at": "2025-01-14T22:03:11.564959+00:00"} {"global_step": 6421, "acc_step": 0, "speed/wps": 12973.427470308234, "speed/FLOPS": 203765505778708.56, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8903318643569946, "optim/lr": 0.0029950685597661298, "optim/total_tokens": 1683226624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438251, "loss/out": 6.041739463806152, "created_at": "2025-01-14T22:03:16.623311+00:00"} {"global_step": 6422, "acc_step": 0, "speed/wps": 12960.563035836698, "speed/FLOPS": 203563452157747.66, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7345131635665894, "optim/lr": 0.002995061615445016, "optim/total_tokens": 1683488768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 6.042779445648193, "created_at": "2025-01-14T22:03:21.683678+00:00"} {"global_step": 6423, "acc_step": 0, "speed/wps": 12963.04817443814, "speed/FLOPS": 203602484674421.6, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3619818687438965, "optim/lr": 0.002995054666246001, "optim/total_tokens": 1683750912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 5.904462814331055, "created_at": "2025-01-14T22:03:26.749641+00:00"} {"global_step": 6424, "acc_step": 0, "speed/wps": 12968.607122314339, "speed/FLOPS": 203689795589612.2, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.8267781734466553, "optim/lr": 0.0029950477121691074, "optim/total_tokens": 1684013056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465057, "loss/out": 5.9895172119140625, "created_at": "2025-01-14T22:03:31.810383+00:00"} {"global_step": 6425, "acc_step": 0, "speed/wps": 12969.54694108768, "speed/FLOPS": 203704556734893.8, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.777420997619629, "optim/lr": 0.0029950407532143572, "optim/total_tokens": 1684275200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 6.224695205688477, "created_at": "2025-01-14T22:03:36.871579+00:00"} {"global_step": 6426, "acc_step": 0, "speed/wps": 12969.453001605047, "speed/FLOPS": 203703081286232.7, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 16.430870056152344, "optim/lr": 0.0029950337893817737, "optim/total_tokens": 1684537344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481190, "loss/out": 6.2078166007995605, "created_at": "2025-01-14T22:03:41.931101+00:00"} {"global_step": 6427, "acc_step": 0, "speed/wps": 12958.059133006444, "speed/FLOPS": 203524124923075.47, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.4557271003723145, "optim/lr": 0.0029950268206713803, "optim/total_tokens": 1684799488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440921, "loss/out": 6.174167633056641, "created_at": "2025-01-14T22:03:46.994820+00:00"} {"global_step": 6428, "acc_step": 0, "speed/wps": 12967.321787970157, "speed/FLOPS": 203669607647502.3, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.026168346405029, "optim/lr": 0.0029950198470831985, "optim/total_tokens": 1685061632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316000, "loss/out": 6.110555171966553, "created_at": "2025-01-14T22:03:52.056645+00:00"} {"global_step": 6429, "acc_step": 0, "speed/wps": 12962.786201715831, "speed/FLOPS": 203598370032829.5, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.075934648513794, "optim/lr": 0.0029950128686172515, "optim/total_tokens": 1685323776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448890, "loss/out": 6.064119815826416, "created_at": "2025-01-14T22:03:57.118491+00:00"} {"global_step": 6430, "acc_step": 0, "speed/wps": 12961.200882093739, "speed/FLOPS": 203573470409707.1, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.8243753910064697, "optim/lr": 0.0029950058852735617, "optim/total_tokens": 1685585920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476746, "loss/out": 5.962490081787109, "created_at": "2025-01-14T22:04:02.183469+00:00"} {"global_step": 6431, "acc_step": 0, "speed/wps": 12967.8467822573, "speed/FLOPS": 203677853404199.66, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5685410499572754, "optim/lr": 0.0029949988970521526, "optim/total_tokens": 1685848064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452081, "loss/out": 6.041357517242432, "created_at": "2025-01-14T22:04:07.243576+00:00"} {"global_step": 6432, "acc_step": 0, "speed/wps": 12964.629106743158, "speed/FLOPS": 203627315388700.53, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.9794294834136963, "optim/lr": 0.0029949919039530468, "optim/total_tokens": 1686110208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 5.957472324371338, "created_at": "2025-01-14T22:04:12.304975+00:00"} {"global_step": 6433, "acc_step": 0, "speed/wps": 12966.89638817092, "speed/FLOPS": 203662926159094.72, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.8198014497756958, "optim/lr": 0.002994984905976267, "optim/total_tokens": 1686372352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 6.0613813400268555, "created_at": "2025-01-14T22:04:17.364967+00:00"} {"global_step": 6434, "acc_step": 0, "speed/wps": 12967.18760035116, "speed/FLOPS": 203667500046552.88, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.009890079498291, "optim/lr": 0.002994977903121836, "optim/total_tokens": 1686634496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331250, "loss/out": 5.996547698974609, "created_at": "2025-01-14T22:04:22.429713+00:00"} {"global_step": 6435, "acc_step": 0, "speed/wps": 12970.67566945118, "speed/FLOPS": 203722284964877.9, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.842844009399414, "optim/lr": 0.0029949708953897767, "optim/total_tokens": 1686896640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372547, "loss/out": 6.014062881469727, "created_at": "2025-01-14T22:04:27.485732+00:00"} {"global_step": 6436, "acc_step": 0, "speed/wps": 12968.363332304263, "speed/FLOPS": 203685966532501.34, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0012, "optim/grad_norm": 2.299999237060547, "optim/lr": 0.002994963882780112, "optim/total_tokens": 1687158784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443306, "loss/out": 6.04878044128418, "created_at": "2025-01-14T22:04:32.544602+00:00"} {"global_step": 6437, "acc_step": 0, "speed/wps": 12966.830433378604, "speed/FLOPS": 203661890248450.0, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.7111783027648926, "optim/lr": 0.0029949568652928643, "optim/total_tokens": 1687420928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449411, "loss/out": 6.062620639801025, "created_at": "2025-01-14T22:04:37.605029+00:00"} {"global_step": 6438, "acc_step": 0, "speed/wps": 12969.01234394102, "speed/FLOPS": 203696160152092.06, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.398144721984863, "optim/lr": 0.0029949498429280574, "optim/total_tokens": 1687683072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 6.033451080322266, "created_at": "2025-01-14T22:04:42.668428+00:00"} {"global_step": 6439, "acc_step": 0, "speed/wps": 12971.239184376693, "speed/FLOPS": 203731135741133.75, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.670943260192871, "optim/lr": 0.0029949428156857134, "optim/total_tokens": 1687945216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476509, "loss/out": 6.175406455993652, "created_at": "2025-01-14T22:04:47.728884+00:00"} {"global_step": 6440, "acc_step": 0, "speed/wps": 12959.712947880967, "speed/FLOPS": 203550100358259.66, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.2559893131256104, "optim/lr": 0.002994935783565855, "optim/total_tokens": 1688207360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 6.003039360046387, "created_at": "2025-01-14T22:04:52.792499+00:00"} {"global_step": 6441, "acc_step": 0, "speed/wps": 12971.304361507242, "speed/FLOPS": 203732159437531.78, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.7893896102905273, "optim/lr": 0.002994928746568507, "optim/total_tokens": 1688469504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453660, "loss/out": 6.098511219024658, "created_at": "2025-01-14T22:04:57.852603+00:00"} {"global_step": 6442, "acc_step": 0, "speed/wps": 12966.787350349217, "speed/FLOPS": 203661213570271.22, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.1867763996124268, "optim/lr": 0.0029949217046936903, "optim/total_tokens": 1688731648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327927, "loss/out": 5.966959476470947, "created_at": "2025-01-14T22:05:02.927434+00:00"} {"global_step": 6443, "acc_step": 0, "speed/wps": 12967.461116062312, "speed/FLOPS": 203671795986646.47, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.646071672439575, "optim/lr": 0.0029949146579414288, "optim/total_tokens": 1688993792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 346260, "loss/out": 6.076999664306641, "created_at": "2025-01-14T22:05:07.984782+00:00"} {"global_step": 6444, "acc_step": 0, "speed/wps": 12971.839967829173, "speed/FLOPS": 203740571870817.94, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.062896251678467, "optim/lr": 0.0029949076063117454, "optim/total_tokens": 1689255936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 381895, "loss/out": 5.993803977966309, "created_at": "2025-01-14T22:05:13.042984+00:00"} {"global_step": 6445, "acc_step": 0, "speed/wps": 12965.402690826968, "speed/FLOPS": 203639465589752.72, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.4120678901672363, "optim/lr": 0.0029949005498046627, "optim/total_tokens": 1689518080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441442, "loss/out": 5.965622901916504, "created_at": "2025-01-14T22:05:18.104779+00:00"} {"global_step": 6446, "acc_step": 0, "speed/wps": 12974.81142583627, "speed/FLOPS": 203787242702030.56, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.537260055541992, "optim/lr": 0.0029948934884202042, "optim/total_tokens": 1689780224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444644, "loss/out": 6.158466339111328, "created_at": "2025-01-14T22:05:23.159356+00:00"} {"global_step": 6447, "acc_step": 0, "speed/wps": 12969.089019578585, "speed/FLOPS": 203697364448343.4, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.441136837005615, "optim/lr": 0.002994886422158393, "optim/total_tokens": 1690042368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465057, "loss/out": 6.066162109375, "created_at": "2025-01-14T22:05:28.216549+00:00"} {"global_step": 6448, "acc_step": 0, "speed/wps": 12975.851857558298, "speed/FLOPS": 203803584111928.03, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.02624797821045, "optim/lr": 0.002994879351019252, "optim/total_tokens": 1690304512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 5.997087001800537, "created_at": "2025-01-14T22:05:33.279111+00:00"} {"global_step": 6449, "acc_step": 0, "speed/wps": 12969.0335856717, "speed/FLOPS": 203696493782661.6, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 22.671464920043945, "optim/lr": 0.002994872275002804, "optim/total_tokens": 1690566656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438531, "loss/out": 6.106533527374268, "created_at": "2025-01-14T22:05:38.337534+00:00"} {"global_step": 6450, "acc_step": 0, "speed/wps": 12964.089445529076, "speed/FLOPS": 203618839267761.2, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.7613937854766846, "optim/lr": 0.0029948651941090723, "optim/total_tokens": 1690828800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 6.057867050170898, "created_at": "2025-01-14T22:05:43.396560+00:00"} {"global_step": 6451, "acc_step": 0, "speed/wps": 12967.441886762728, "speed/FLOPS": 203671493964072.7, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.221738815307617, "optim/lr": 0.00299485810833808, "optim/total_tokens": 1691090944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444881, "loss/out": 6.001988410949707, "created_at": "2025-01-14T22:05:48.454381+00:00"} {"global_step": 6452, "acc_step": 0, "speed/wps": 12966.25555859868, "speed/FLOPS": 203652861050070.56, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0009, "optim/grad_norm": 7.512727737426758, "optim/lr": 0.00299485101768985, "optim/total_tokens": 1691353088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432668, "loss/out": 6.053430557250977, "created_at": "2025-01-14T22:05:53.512291+00:00"} {"global_step": 6453, "acc_step": 0, "speed/wps": 12968.38426121324, "speed/FLOPS": 203686295249776.16, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.84379768371582, "optim/lr": 0.002994843922164406, "optim/total_tokens": 1691615232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432127, "loss/out": 6.06348991394043, "created_at": "2025-01-14T22:05:58.569984+00:00"} {"global_step": 6454, "acc_step": 0, "speed/wps": 12971.955547105563, "speed/FLOPS": 203742387202176.2, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.883650064468384, "optim/lr": 0.002994836821761771, "optim/total_tokens": 1691877376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327927, "loss/out": 6.125157356262207, "created_at": "2025-01-14T22:06:03.633495+00:00"} {"global_step": 6455, "acc_step": 0, "speed/wps": 12969.5446247946, "speed/FLOPS": 203704520354328.1, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.033871173858643, "optim/lr": 0.0029948297164819676, "optim/total_tokens": 1692139520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 378315, "loss/out": 6.045031547546387, "created_at": "2025-01-14T22:06:08.696631+00:00"} {"global_step": 6456, "acc_step": 0, "speed/wps": 12967.438082489016, "speed/FLOPS": 203671434212726.4, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0008, "optim/grad_norm": 10.009906768798828, "optim/lr": 0.0029948226063250196, "optim/total_tokens": 1692401664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447554, "loss/out": 5.987978935241699, "created_at": "2025-01-14T22:06:13.758351+00:00"} {"global_step": 6457, "acc_step": 0, "speed/wps": 12973.086524414775, "speed/FLOPS": 203760150754943.06, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.378849029541016, "optim/lr": 0.00299481549129095, "optim/total_tokens": 1692663808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479095, "loss/out": 6.177483081817627, "created_at": "2025-01-14T22:06:18.816294+00:00"} {"global_step": 6458, "acc_step": 0, "speed/wps": 12972.270912483691, "speed/FLOPS": 203747340448797.12, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.909411907196045, "optim/lr": 0.0029948083713797814, "optim/total_tokens": 1692925952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438769, "loss/out": 5.962316036224365, "created_at": "2025-01-14T22:06:23.877030+00:00"} {"global_step": 6459, "acc_step": 0, "speed/wps": 12972.149975731942, "speed/FLOPS": 203745440970926.4, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.0326480865478516, "optim/lr": 0.0029948012465915385, "optim/total_tokens": 1693188096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 6.045805931091309, "created_at": "2025-01-14T22:06:28.934781+00:00"} {"global_step": 6460, "acc_step": 0, "speed/wps": 12965.199520756181, "speed/FLOPS": 203636274524606.06, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.0169148445129395, "optim/lr": 0.002994794116926243, "optim/total_tokens": 1693450240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446742, "loss/out": 5.921786308288574, "created_at": "2025-01-14T22:06:33.999594+00:00"} {"global_step": 6461, "acc_step": 0, "speed/wps": 12969.423480712232, "speed/FLOPS": 203702617620045.1, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.306010723114014, "optim/lr": 0.0029947869823839187, "optim/total_tokens": 1693712384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302343, "loss/out": 6.086040496826172, "created_at": "2025-01-14T22:06:39.057552+00:00"} {"global_step": 6462, "acc_step": 0, "speed/wps": 12966.171328511264, "speed/FLOPS": 203651538100803.44, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.833185195922852, "optim/lr": 0.002994779842964589, "optim/total_tokens": 1693974528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 384005, "loss/out": 6.200405120849609, "created_at": "2025-01-14T22:06:44.115440+00:00"} {"global_step": 6463, "acc_step": 0, "speed/wps": 12973.275198859319, "speed/FLOPS": 203763114146360.47, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 34.606929779052734, "optim/lr": 0.002994772698668278, "optim/total_tokens": 1694236672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442778, "loss/out": 5.965541362762451, "created_at": "2025-01-14T22:06:49.171686+00:00"} {"global_step": 6464, "acc_step": 0, "speed/wps": 12961.366179868948, "speed/FLOPS": 203576066638409.7, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.607843399047852, "optim/lr": 0.0029947655494950076, "optim/total_tokens": 1694498816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433179, "loss/out": 6.052475929260254, "created_at": "2025-01-14T22:06:54.236828+00:00"} {"global_step": 6465, "acc_step": 0, "speed/wps": 12967.017174417571, "speed/FLOPS": 203664823273076.1, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.639533042907715, "optim/lr": 0.0029947583954448016, "optim/total_tokens": 1694760960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475696, "loss/out": 6.0571513175964355, "created_at": "2025-01-14T22:06:59.299316+00:00"} {"global_step": 6466, "acc_step": 0, "speed/wps": 12965.06566370568, "speed/FLOPS": 203634172115689.1, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.905313491821289, "optim/lr": 0.0029947512365176834, "optim/total_tokens": 1695023104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 6.043618202209473, "created_at": "2025-01-14T22:07:04.358474+00:00"} {"global_step": 6467, "acc_step": 0, "speed/wps": 12970.897710339315, "speed/FLOPS": 203725772422142.22, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.162899494171143, "optim/lr": 0.002994744072713677, "optim/total_tokens": 1695285248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 6.013150691986084, "created_at": "2025-01-14T22:07:09.415153+00:00"} {"global_step": 6468, "acc_step": 0, "speed/wps": 12958.969076659634, "speed/FLOPS": 203538416838542.44, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.458251953125, "optim/lr": 0.0029947369040328044, "optim/total_tokens": 1695547392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435044, "loss/out": 5.900237083435059, "created_at": "2025-01-14T22:07:14.476031+00:00"} {"global_step": 6469, "acc_step": 0, "speed/wps": 12969.156419548886, "speed/FLOPS": 203698423057494.25, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.499911308288574, "optim/lr": 0.0029947297304750904, "optim/total_tokens": 1695809536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 6.033512592315674, "created_at": "2025-01-14T22:07:19.533344+00:00"} {"global_step": 6470, "acc_step": 0, "speed/wps": 12970.217182934492, "speed/FLOPS": 203715083804106.8, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.427430152893066, "optim/lr": 0.0029947225520405574, "optim/total_tokens": 1696071680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438773, "loss/out": 6.0080132484436035, "created_at": "2025-01-14T22:07:24.592767+00:00"} {"global_step": 6471, "acc_step": 0, "speed/wps": 12966.001906427291, "speed/FLOPS": 203648877094164.3, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 15.11411190032959, "optim/lr": 0.0029947153687292292, "optim/total_tokens": 1696333824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 282855, "loss/out": 6.098548889160156, "created_at": "2025-01-14T22:07:29.655450+00:00"} {"global_step": 6472, "acc_step": 0, "speed/wps": 12975.510114300772, "speed/FLOPS": 203798216564464.12, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.122796058654785, "optim/lr": 0.0029947081805411294, "optim/total_tokens": 1696595968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 408354, "loss/out": 6.138224124908447, "created_at": "2025-01-14T22:07:34.716635+00:00"} {"global_step": 6473, "acc_step": 0, "speed/wps": 12970.545479098759, "speed/FLOPS": 203720240146492.84, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 15.32479476928711, "optim/lr": 0.002994700987476281, "optim/total_tokens": 1696858112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 6.053574562072754, "created_at": "2025-01-14T22:07:39.774581+00:00"} {"global_step": 6474, "acc_step": 0, "speed/wps": 12975.22933985419, "speed/FLOPS": 203793806615956.03, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 42.42509078979492, "optim/lr": 0.0029946937895347085, "optim/total_tokens": 1697120256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 6.006589889526367, "created_at": "2025-01-14T22:07:44.829657+00:00"} {"global_step": 6475, "acc_step": 0, "speed/wps": 12969.414299383412, "speed/FLOPS": 203702473414659.5, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.988635063171387, "optim/lr": 0.0029946865867164344, "optim/total_tokens": 1697382400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429217, "loss/out": 6.188011169433594, "created_at": "2025-01-14T22:07:49.886489+00:00"} {"global_step": 6476, "acc_step": 0, "speed/wps": 12969.246850576217, "speed/FLOPS": 203699843401043.72, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.618809700012207, "optim/lr": 0.0029946793790214823, "optim/total_tokens": 1697644544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445404, "loss/out": 5.965228080749512, "created_at": "2025-01-14T22:07:54.946885+00:00"} {"global_step": 6477, "acc_step": 0, "speed/wps": 12969.023374717424, "speed/FLOPS": 203696333405593.38, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.940964698791504, "optim/lr": 0.0029946721664498755, "optim/total_tokens": 1697906688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 5.849605560302734, "created_at": "2025-01-14T22:08:00.005057+00:00"} {"global_step": 6478, "acc_step": 0, "speed/wps": 12973.492542663766, "speed/FLOPS": 203766527829470.97, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 19.443923950195312, "optim/lr": 0.0029946649490016383, "optim/total_tokens": 1698168832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441687, "loss/out": 6.087623596191406, "created_at": "2025-01-14T22:08:05.065590+00:00"} {"global_step": 6479, "acc_step": 0, "speed/wps": 12973.411435523842, "speed/FLOPS": 203765253930385.6, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.63152027130127, "optim/lr": 0.002994657726676794, "optim/total_tokens": 1698430976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461889, "loss/out": 6.1638898849487305, "created_at": "2025-01-14T22:08:10.128773+00:00"} {"global_step": 6480, "acc_step": 0, "speed/wps": 12966.073149458136, "speed/FLOPS": 203649996063861.9, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0014, "optim/grad_norm": 16.206340789794922, "optim/lr": 0.0029946504994753658, "optim/total_tokens": 1698693120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327927, "loss/out": 6.081629276275635, "created_at": "2025-01-14T22:08:15.192387+00:00"} {"global_step": 6481, "acc_step": 0, "speed/wps": 12970.52559992778, "speed/FLOPS": 203719927916795.78, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.3780035972595215, "optim/lr": 0.002994643267397377, "optim/total_tokens": 1698955264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336478, "loss/out": 5.945568084716797, "created_at": "2025-01-14T22:08:20.248607+00:00"} {"global_step": 6482, "acc_step": 0, "speed/wps": 12972.01401903808, "speed/FLOPS": 203743305584225.25, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.623716354370117, "optim/lr": 0.0029946360304428524, "optim/total_tokens": 1699217408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 393639, "loss/out": 6.00648307800293, "created_at": "2025-01-14T22:08:25.304644+00:00"} {"global_step": 6483, "acc_step": 0, "speed/wps": 12967.677414364227, "speed/FLOPS": 203675193248703.06, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 18.938913345336914, "optim/lr": 0.0029946287886118144, "optim/total_tokens": 1699479552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428403, "loss/out": 5.983305931091309, "created_at": "2025-01-14T22:08:30.367093+00:00"} {"global_step": 6484, "acc_step": 0, "speed/wps": 12975.657600280694, "speed/FLOPS": 203800533034445.78, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.09540843963623, "optim/lr": 0.0029946215419042873, "optim/total_tokens": 1699741696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 6.100046157836914, "created_at": "2025-01-14T22:08:35.427096+00:00"} {"global_step": 6485, "acc_step": 0, "speed/wps": 12966.853127887443, "speed/FLOPS": 203662246696896.88, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 183.84815979003906, "optim/lr": 0.002994614290320295, "optim/total_tokens": 1700003840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424678, "loss/out": 6.01566743850708, "created_at": "2025-01-14T22:08:40.485954+00:00"} {"global_step": 6486, "acc_step": 0, "speed/wps": 12980.354815881621, "speed/FLOPS": 203874309260108.97, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 25.032459259033203, "optim/lr": 0.00299460703385986, "optim/total_tokens": 1700265984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455003, "loss/out": 5.962571144104004, "created_at": "2025-01-14T22:08:45.544024+00:00"} {"global_step": 6487, "acc_step": 0, "speed/wps": 12974.775598300292, "speed/FLOPS": 203786679981345.84, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.579936981201172, "optim/lr": 0.0029945997725230075, "optim/total_tokens": 1700528128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 5.96157169342041, "created_at": "2025-01-14T22:08:50.598560+00:00"} {"global_step": 6488, "acc_step": 0, "speed/wps": 12974.033271684784, "speed/FLOPS": 203775020721785.62, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.457725524902344, "optim/lr": 0.0029945925063097597, "optim/total_tokens": 1700790272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427589, "loss/out": 5.931076526641846, "created_at": "2025-01-14T22:08:55.655885+00:00"} {"global_step": 6489, "acc_step": 0, "speed/wps": 12975.806711866404, "speed/FLOPS": 203802875036800.16, "speed/curr_iter_time": 1.2719, "speed/data_load_time": 0.0008, "optim/grad_norm": 19.490976333618164, "optim/lr": 0.0029945852352201417, "optim/total_tokens": 1701052416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307513, "loss/out": 5.878702163696289, "created_at": "2025-01-14T22:09:00.710406+00:00"} {"global_step": 6490, "acc_step": 0, "speed/wps": 12970.717131042786, "speed/FLOPS": 203722936176148.28, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.365367889404297, "optim/lr": 0.0029945779592541757, "optim/total_tokens": 1701314560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 378602, "loss/out": 6.017604351043701, "created_at": "2025-01-14T22:09:05.769868+00:00"} {"global_step": 6491, "acc_step": 0, "speed/wps": 12981.224870278063, "speed/FLOPS": 203887974659985.12, "speed/curr_iter_time": 1.2721, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.217230796813965, "optim/lr": 0.0029945706784118863, "optim/total_tokens": 1701576704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427020, "loss/out": 6.050652980804443, "created_at": "2025-01-14T22:09:10.824633+00:00"} {"global_step": 6492, "acc_step": 0, "speed/wps": 12969.284160644376, "speed/FLOPS": 203700429406933.56, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 23.886112213134766, "optim/lr": 0.0029945633926932976, "optim/total_tokens": 1701838848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 412655, "loss/out": 5.972644805908203, "created_at": "2025-01-14T22:09:15.896044+00:00"} {"global_step": 6493, "acc_step": 0, "speed/wps": 12976.827943172671, "speed/FLOPS": 203818914877783.94, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.279055595397949, "optim/lr": 0.002994556102098433, "optim/total_tokens": 1702100992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444641, "loss/out": 6.009871482849121, "created_at": "2025-01-14T22:09:20.961897+00:00"} {"global_step": 6494, "acc_step": 0, "speed/wps": 12976.036082897443, "speed/FLOPS": 203806477624031.44, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 24.14804458618164, "optim/lr": 0.0029945488066273162, "optim/total_tokens": 1702363136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 6.060296058654785, "created_at": "2025-01-14T22:09:26.018040+00:00"} {"global_step": 6495, "acc_step": 0, "speed/wps": 12972.904188500363, "speed/FLOPS": 203757286918851.4, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.879465579986572, "optim/lr": 0.0029945415062799714, "optim/total_tokens": 1702625280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437195, "loss/out": 5.9536452293396, "created_at": "2025-01-14T22:09:31.076452+00:00"} {"global_step": 6496, "acc_step": 0, "speed/wps": 12967.008462440092, "speed/FLOPS": 203664686439498.25, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.950693130493164, "optim/lr": 0.0029945342010564216, "optim/total_tokens": 1702887424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 5.93702507019043, "created_at": "2025-01-14T22:09:36.138520+00:00"} {"global_step": 6497, "acc_step": 0, "speed/wps": 12974.103342377555, "speed/FLOPS": 203776121278302.8, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 23.574989318847656, "optim/lr": 0.002994526890956691, "optim/total_tokens": 1703149568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448361, "loss/out": 6.022655963897705, "created_at": "2025-01-14T22:09:41.194842+00:00"} {"global_step": 6498, "acc_step": 0, "speed/wps": 12968.837448289823, "speed/FLOPS": 203693413175558.22, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 26.286298751831055, "optim/lr": 0.0029945195759808045, "optim/total_tokens": 1703411712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444644, "loss/out": 6.044012069702148, "created_at": "2025-01-14T22:09:46.252404+00:00"} {"global_step": 6499, "acc_step": 0, "speed/wps": 12975.538172315886, "speed/FLOPS": 203798657254147.25, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.349813461303711, "optim/lr": 0.0029945122561287844, "optim/total_tokens": 1703673856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325907, "loss/out": 6.115323543548584, "created_at": "2025-01-14T22:09:51.307580+00:00"} {"global_step": 6500, "acc_step": 0, "speed/wps": 12976.822278753803, "speed/FLOPS": 203818825910301.9, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.789462089538574, "optim/lr": 0.002994504931400655, "optim/total_tokens": 1703936000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318166, "loss/out": 6.008636474609375, "created_at": "2025-01-14T22:09:56.366481+00:00"} {"global_step": 6501, "acc_step": 0, "speed/wps": 12973.04178302428, "speed/FLOPS": 203759448029924.22, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.702875137329102, "optim/lr": 0.002994497601796441, "optim/total_tokens": 1704198144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449942, "loss/out": 6.033199310302734, "created_at": "2025-01-14T22:10:01.425236+00:00"} {"global_step": 6502, "acc_step": 0, "speed/wps": 12975.199792659503, "speed/FLOPS": 203793342536661.38, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 22.845746994018555, "optim/lr": 0.0029944902673161656, "optim/total_tokens": 1704460288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 5.979736328125, "created_at": "2025-01-14T22:10:06.480086+00:00"} {"global_step": 6503, "acc_step": 0, "speed/wps": 12972.602811419856, "speed/FLOPS": 203752553377666.53, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.865396499633789, "optim/lr": 0.0029944829279598523, "optim/total_tokens": 1704722432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 5.979509353637695, "created_at": "2025-01-14T22:10:11.535606+00:00"} {"global_step": 6504, "acc_step": 0, "speed/wps": 12973.417745149365, "speed/FLOPS": 203765353031724.78, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.926401615142822, "optim/lr": 0.0029944755837275263, "optim/total_tokens": 1704984576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 5.888835906982422, "created_at": "2025-01-14T22:10:16.594513+00:00"} {"global_step": 6505, "acc_step": 0, "speed/wps": 12976.725996354851, "speed/FLOPS": 203817313662920.47, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0009, "optim/grad_norm": 9.014894485473633, "optim/lr": 0.002994468234619211, "optim/total_tokens": 1705246720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465633, "loss/out": 5.934223651885986, "created_at": "2025-01-14T22:10:21.649572+00:00"} {"global_step": 6506, "acc_step": 0, "speed/wps": 12972.362137859822, "speed/FLOPS": 203748773268686.06, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.628775596618652, "optim/lr": 0.0029944608806349296, "optim/total_tokens": 1705508864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 5.932294845581055, "created_at": "2025-01-14T22:10:26.705549+00:00"} {"global_step": 6507, "acc_step": 0, "speed/wps": 12967.38651711722, "speed/FLOPS": 203670624307704.12, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.0328755378723145, "optim/lr": 0.0029944535217747066, "optim/total_tokens": 1705771008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427332, "loss/out": 5.984457492828369, "created_at": "2025-01-14T22:10:31.764335+00:00"} {"global_step": 6508, "acc_step": 0, "speed/wps": 12975.03188341197, "speed/FLOPS": 203790705291197.38, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 24.60877799987793, "optim/lr": 0.0029944461580385666, "optim/total_tokens": 1706033152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 6.091419219970703, "created_at": "2025-01-14T22:10:36.820263+00:00"} {"global_step": 6509, "acc_step": 0, "speed/wps": 12976.737786887508, "speed/FLOPS": 203817498849437.75, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.657071590423584, "optim/lr": 0.0029944387894265336, "optim/total_tokens": 1706295296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454979, "loss/out": 6.309060573577881, "created_at": "2025-01-14T22:10:41.879279+00:00"} {"global_step": 6510, "acc_step": 0, "speed/wps": 12977.934818123864, "speed/FLOPS": 203836299869898.0, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.190608024597168, "optim/lr": 0.0029944314159386304, "optim/total_tokens": 1706557440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435329, "loss/out": 6.1354146003723145, "created_at": "2025-01-14T22:10:46.933761+00:00"} {"global_step": 6511, "acc_step": 0, "speed/wps": 12971.370930894313, "speed/FLOPS": 203733205001231.66, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.60536766052246, "optim/lr": 0.0029944240375748827, "optim/total_tokens": 1706819584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298409, "loss/out": 6.269806861877441, "created_at": "2025-01-14T22:10:51.994074+00:00"} {"global_step": 6512, "acc_step": 0, "speed/wps": 12974.071097278304, "speed/FLOPS": 203775614824709.66, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.22208023071289, "optim/lr": 0.002994416654335313, "optim/total_tokens": 1707081728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 404376, "loss/out": 6.284021377563477, "created_at": "2025-01-14T22:10:57.049006+00:00"} {"global_step": 6513, "acc_step": 0, "speed/wps": 12970.853854324387, "speed/FLOPS": 203725083603163.28, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 18.247438430786133, "optim/lr": 0.0029944092662199467, "optim/total_tokens": 1707343872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450471, "loss/out": 6.298910140991211, "created_at": "2025-01-14T22:11:02.105547+00:00"} {"global_step": 6514, "acc_step": 0, "speed/wps": 12969.29902114965, "speed/FLOPS": 203700662811589.38, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.450533866882324, "optim/lr": 0.0029944018732288072, "optim/total_tokens": 1707606016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 6.2717437744140625, "created_at": "2025-01-14T22:11:07.170330+00:00"} {"global_step": 6515, "acc_step": 0, "speed/wps": 12972.920731450098, "speed/FLOPS": 203757546748609.38, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0012, "optim/grad_norm": 6.889544486999512, "optim/lr": 0.002994394475361919, "optim/total_tokens": 1707868160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455517, "loss/out": 6.284148693084717, "created_at": "2025-01-14T22:11:12.226293+00:00"} {"global_step": 6516, "acc_step": 0, "speed/wps": 12975.19510818545, "speed/FLOPS": 203793268960559.22, "speed/curr_iter_time": 1.2713, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.9785540103912354, "optim/lr": 0.002994387072619306, "optim/total_tokens": 1708130304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454481, "loss/out": 6.174878120422363, "created_at": "2025-01-14T22:11:17.281045+00:00"} {"global_step": 6517, "acc_step": 0, "speed/wps": 12975.280350352097, "speed/FLOPS": 203794607805921.44, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.379507541656494, "optim/lr": 0.0029943796650009915, "optim/total_tokens": 1708392448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464535, "loss/out": 6.150430679321289, "created_at": "2025-01-14T22:11:22.335517+00:00"} {"global_step": 6518, "acc_step": 0, "speed/wps": 12970.067279408328, "speed/FLOPS": 203712729363239.38, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.483245372772217, "optim/lr": 0.0029943722525070018, "optim/total_tokens": 1708654592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 6.117565155029297, "created_at": "2025-01-14T22:11:27.392256+00:00"} {"global_step": 6519, "acc_step": 0, "speed/wps": 12973.574133977667, "speed/FLOPS": 203767809333174.03, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0008, "optim/grad_norm": 10.170557975769043, "optim/lr": 0.0029943648351373592, "optim/total_tokens": 1708916736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438769, "loss/out": 6.060391902923584, "created_at": "2025-01-14T22:11:32.456481+00:00"} {"global_step": 6520, "acc_step": 0, "speed/wps": 12966.488238105492, "speed/FLOPS": 203656515601460.22, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.526402473449707, "optim/lr": 0.002994357412892088, "optim/total_tokens": 1709178880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 6.2105631828308105, "created_at": "2025-01-14T22:11:37.517720+00:00"} {"global_step": 6521, "acc_step": 0, "speed/wps": 12967.35841548847, "speed/FLOPS": 203670182933008.9, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 22.04842758178711, "optim/lr": 0.0029943499857712137, "optim/total_tokens": 1709441024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435044, "loss/out": 6.146442413330078, "created_at": "2025-01-14T22:11:42.576328+00:00"} {"global_step": 6522, "acc_step": 0, "speed/wps": 12975.930468471362, "speed/FLOPS": 203804818804340.4, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.6193628311157227, "optim/lr": 0.0029943425537747592, "optim/total_tokens": 1709703168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309845, "loss/out": 6.2200469970703125, "created_at": "2025-01-14T22:11:47.631635+00:00"} {"global_step": 6523, "acc_step": 0, "speed/wps": 12964.164231054738, "speed/FLOPS": 203620013877207.78, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.0274341106414795, "optim/lr": 0.00299433511690275, "optim/total_tokens": 1709965312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 392005, "loss/out": 6.221524238586426, "created_at": "2025-01-14T22:11:52.690410+00:00"} {"global_step": 6524, "acc_step": 0, "speed/wps": 12970.409854874235, "speed/FLOPS": 203718109981679.47, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.86584186553955, "optim/lr": 0.0029943276751552093, "optim/total_tokens": 1710227456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445976, "loss/out": 5.983391761779785, "created_at": "2025-01-14T22:11:57.752047+00:00"} {"global_step": 6525, "acc_step": 0, "speed/wps": 12966.781018498918, "speed/FLOPS": 203661114119861.38, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.023591995239258, "optim/lr": 0.0029943202285321615, "optim/total_tokens": 1710489600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 6.222442626953125, "created_at": "2025-01-14T22:12:02.814232+00:00"} {"global_step": 6526, "acc_step": 0, "speed/wps": 12973.221362726, "speed/FLOPS": 203762268575908.1, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.076690673828125, "optim/lr": 0.0029943127770336317, "optim/total_tokens": 1710751744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 6.3098249435424805, "created_at": "2025-01-14T22:12:07.872784+00:00"} {"global_step": 6527, "acc_step": 0, "speed/wps": 12977.617726918183, "speed/FLOPS": 203831319516783.78, "speed/curr_iter_time": 1.2723, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.0867133140563965, "optim/lr": 0.0029943053206596434, "optim/total_tokens": 1711013888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468777, "loss/out": 6.272226810455322, "created_at": "2025-01-14T22:12:12.929419+00:00"} {"global_step": 6528, "acc_step": 0, "speed/wps": 12966.332574359087, "speed/FLOPS": 203654070688418.53, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.623140811920166, "optim/lr": 0.0029942978594102213, "optim/total_tokens": 1711276032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 6.133282661437988, "created_at": "2025-01-14T22:12:17.989738+00:00"} {"global_step": 6529, "acc_step": 0, "speed/wps": 12967.944438517052, "speed/FLOPS": 203679387229952.8, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 11.321317672729492, "optim/lr": 0.002994290393285389, "optim/total_tokens": 1711538176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434515, "loss/out": 6.09186315536499, "created_at": "2025-01-14T22:12:23.048309+00:00"} {"global_step": 6530, "acc_step": 0, "speed/wps": 12977.17324514742, "speed/FLOPS": 203824338319792.0, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.652919769287109, "optim/lr": 0.0029942829222851723, "optim/total_tokens": 1711800320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324468, "loss/out": 6.1302289962768555, "created_at": "2025-01-14T22:12:28.102416+00:00"} {"global_step": 6531, "acc_step": 0, "speed/wps": 12967.991539356632, "speed/FLOPS": 203680127013361.75, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.243535995483398, "optim/lr": 0.002994275446409594, "optim/total_tokens": 1712062464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295180, "loss/out": 6.160088539123535, "created_at": "2025-01-14T22:12:33.159500+00:00"} {"global_step": 6532, "acc_step": 0, "speed/wps": 12974.899227580858, "speed/FLOPS": 203788621749081.56, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 21.089818954467773, "optim/lr": 0.0029942679656586803, "optim/total_tokens": 1712324608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322066, "loss/out": 6.155143737792969, "created_at": "2025-01-14T22:12:38.216603+00:00"} {"global_step": 6533, "acc_step": 0, "speed/wps": 12970.873036363228, "speed/FLOPS": 203725384883442.16, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 21.739990234375, "optim/lr": 0.002994260480032454, "optim/total_tokens": 1712586752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292963, "loss/out": 6.243371486663818, "created_at": "2025-01-14T22:12:43.272756+00:00"} {"global_step": 6534, "acc_step": 0, "speed/wps": 12978.410218290655, "speed/FLOPS": 203843766682785.75, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.089264392852783, "optim/lr": 0.00299425298953094, "optim/total_tokens": 1712848896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298060, "loss/out": 6.14195442199707, "created_at": "2025-01-14T22:12:48.325931+00:00"} {"global_step": 6535, "acc_step": 0, "speed/wps": 12977.010047182226, "speed/FLOPS": 203821775071491.12, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.955805778503418, "optim/lr": 0.0029942454941541633, "optim/total_tokens": 1713111040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351344, "loss/out": 6.023741245269775, "created_at": "2025-01-14T22:12:53.379868+00:00"} {"global_step": 6536, "acc_step": 0, "speed/wps": 12973.657828807573, "speed/FLOPS": 203769123875486.44, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.001, "optim/grad_norm": 2.6163840293884277, "optim/lr": 0.0029942379939021475, "optim/total_tokens": 1713373184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 399826, "loss/out": 6.0185065269470215, "created_at": "2025-01-14T22:12:58.435403+00:00"} {"global_step": 6537, "acc_step": 0, "speed/wps": 12971.769091830683, "speed/FLOPS": 203739458665867.8, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.959512710571289, "optim/lr": 0.0029942304887749173, "optim/total_tokens": 1713635328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440628, "loss/out": 5.942602157592773, "created_at": "2025-01-14T22:13:03.497740+00:00"} {"global_step": 6538, "acc_step": 0, "speed/wps": 12967.936110051665, "speed/FLOPS": 203679256419959.4, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.585771560668945, "optim/lr": 0.0029942229787724975, "optim/total_tokens": 1713897472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 6.137053489685059, "created_at": "2025-01-14T22:13:08.555336+00:00"} {"global_step": 6539, "acc_step": 0, "speed/wps": 12967.529129981402, "speed/FLOPS": 203672864238706.7, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.439849376678467, "optim/lr": 0.0029942154638949127, "optim/total_tokens": 1714159616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444878, "loss/out": 6.040811538696289, "created_at": "2025-01-14T22:13:13.615965+00:00"} {"global_step": 6540, "acc_step": 0, "speed/wps": 12967.73800088057, "speed/FLOPS": 203676144843196.84, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.001, "optim/grad_norm": 7.867047309875488, "optim/lr": 0.0029942079441421868, "optim/total_tokens": 1714421760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445976, "loss/out": 5.97515869140625, "created_at": "2025-01-14T22:13:18.678238+00:00"} {"global_step": 6541, "acc_step": 0, "speed/wps": 12978.922875203569, "speed/FLOPS": 203851818664067.47, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.005394220352173, "optim/lr": 0.002994200419514345, "optim/total_tokens": 1714683904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 6.065576553344727, "created_at": "2025-01-14T22:13:23.732399+00:00"} {"global_step": 6542, "acc_step": 0, "speed/wps": 12966.946066768012, "speed/FLOPS": 203663706429727.6, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.319795608520508, "optim/lr": 0.0029941928900114115, "optim/total_tokens": 1714946048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452081, "loss/out": 5.985688209533691, "created_at": "2025-01-14T22:13:28.795019+00:00"} {"global_step": 6543, "acc_step": 0, "speed/wps": 12981.225069038364, "speed/FLOPS": 203887977781788.78, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.89626407623291, "optim/lr": 0.0029941853556334106, "optim/total_tokens": 1715208192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318033, "loss/out": 6.007657527923584, "created_at": "2025-01-14T22:13:33.848370+00:00"} {"global_step": 6544, "acc_step": 0, "speed/wps": 12976.42438424897, "speed/FLOPS": 203812576430339.16, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.150744438171387, "optim/lr": 0.0029941778163803676, "optim/total_tokens": 1715470336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356847, "loss/out": 6.116838455200195, "created_at": "2025-01-14T22:13:38.910383+00:00"} {"global_step": 6545, "acc_step": 0, "speed/wps": 12971.30139661383, "speed/FLOPS": 203732112869806.53, "speed/curr_iter_time": 1.2721, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.8122191429138184, "optim/lr": 0.002994170272252307, "optim/total_tokens": 1715732480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 398956, "loss/out": 6.114287853240967, "created_at": "2025-01-14T22:13:43.973075+00:00"} {"global_step": 6546, "acc_step": 0, "speed/wps": 12966.841630090978, "speed/FLOPS": 203662066108203.75, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0009, "optim/grad_norm": 2.1437156200408936, "optim/lr": 0.0029941627232492525, "optim/total_tokens": 1715994624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 6.0875630378723145, "created_at": "2025-01-14T22:13:49.032083+00:00"} {"global_step": 6547, "acc_step": 0, "speed/wps": 12984.04631826057, "speed/FLOPS": 203932289377626.22, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.37560749053955, "optim/lr": 0.0029941551693712294, "optim/total_tokens": 1716256768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432953, "loss/out": 6.023562431335449, "created_at": "2025-01-14T22:13:54.083237+00:00"} {"global_step": 6548, "acc_step": 0, "speed/wps": 12967.706843668691, "speed/FLOPS": 203675655476369.78, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.7536585330963135, "optim/lr": 0.0029941476106182625, "optim/total_tokens": 1716518912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470641, "loss/out": 5.988442420959473, "created_at": "2025-01-14T22:13:59.145782+00:00"} {"global_step": 6549, "acc_step": 0, "speed/wps": 12968.971863662111, "speed/FLOPS": 203695524353686.06, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1287312507629395, "optim/lr": 0.002994140046990376, "optim/total_tokens": 1716781056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449420, "loss/out": 5.96143102645874, "created_at": "2025-01-14T22:14:04.206779+00:00"} {"global_step": 6550, "acc_step": 0, "speed/wps": 12970.355973189162, "speed/FLOPS": 203717263695774.22, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.12941837310791, "optim/lr": 0.002994132478487595, "optim/total_tokens": 1717043200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462672, "loss/out": 5.99558162689209, "created_at": "2025-01-14T22:14:09.263023+00:00"} {"global_step": 6551, "acc_step": 0, "speed/wps": 12966.811712499593, "speed/FLOPS": 203661596211319.97, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.268461227416992, "optim/lr": 0.002994124905109944, "optim/total_tokens": 1717305344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 6.042442321777344, "created_at": "2025-01-14T22:14:14.322588+00:00"} {"global_step": 6552, "acc_step": 0, "speed/wps": 12969.625748134014, "speed/FLOPS": 203705794507848.75, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.283850193023682, "optim/lr": 0.002994117326857448, "optim/total_tokens": 1717567488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436096, "loss/out": 5.984220027923584, "created_at": "2025-01-14T22:14:19.379492+00:00"} {"global_step": 6553, "acc_step": 0, "speed/wps": 12969.390998324003, "speed/FLOPS": 203702107439502.47, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 20.70833969116211, "optim/lr": 0.0029941097437301313, "optim/total_tokens": 1717829632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293833, "loss/out": 6.006601333618164, "created_at": "2025-01-14T22:14:24.436071+00:00"} {"global_step": 6554, "acc_step": 0, "speed/wps": 12970.896208858245, "speed/FLOPS": 203725748839318.75, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.541499614715576, "optim/lr": 0.002994102155728019, "optim/total_tokens": 1718091776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361759, "loss/out": 6.0373921394348145, "created_at": "2025-01-14T22:14:29.492891+00:00"} {"global_step": 6555, "acc_step": 0, "speed/wps": 12973.997441270018, "speed/FLOPS": 203774457955885.56, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.2826602458953857, "optim/lr": 0.002994094562851135, "optim/total_tokens": 1718353920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446505, "loss/out": 6.224117279052734, "created_at": "2025-01-14T22:14:34.552229+00:00"} {"global_step": 6556, "acc_step": 0, "speed/wps": 12960.901079801628, "speed/FLOPS": 203568761602737.97, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.1963653564453125, "optim/lr": 0.002994086965099505, "optim/total_tokens": 1718616064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 6.013435363769531, "created_at": "2025-01-14T22:14:39.612313+00:00"} {"global_step": 6557, "acc_step": 0, "speed/wps": 12969.993281582829, "speed/FLOPS": 203711567125706.25, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.391604900360107, "optim/lr": 0.0029940793624731537, "optim/total_tokens": 1718878208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 6.012796401977539, "created_at": "2025-01-14T22:14:44.696848+00:00"} {"global_step": 6558, "acc_step": 0, "speed/wps": 12969.411182371845, "speed/FLOPS": 203702424457709.62, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 8.284936904907227, "optim/lr": 0.0029940717549721053, "optim/total_tokens": 1719140352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 5.984464645385742, "created_at": "2025-01-14T22:14:49.755108+00:00"} {"global_step": 6559, "acc_step": 0, "speed/wps": 12966.941749056552, "speed/FLOPS": 203663638614135.97, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0008, "optim/grad_norm": 118.87429809570312, "optim/lr": 0.002994064142596386, "optim/total_tokens": 1719402496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 6.028650283813477, "created_at": "2025-01-14T22:14:54.819040+00:00"} {"global_step": 6560, "acc_step": 0, "speed/wps": 12961.345063517098, "speed/FLOPS": 203575734977086.97, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.770124912261963, "optim/lr": 0.002994056525346019, "optim/total_tokens": 1719664640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431076, "loss/out": 6.103118896484375, "created_at": "2025-01-14T22:14:59.881363+00:00"} {"global_step": 6561, "acc_step": 0, "speed/wps": 12974.102509935292, "speed/FLOPS": 203776108203653.16, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.601618528366089, "optim/lr": 0.0029940489032210297, "optim/total_tokens": 1719926784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 284402, "loss/out": 6.085869789123535, "created_at": "2025-01-14T22:15:04.937149+00:00"} {"global_step": 6562, "acc_step": 0, "speed/wps": 12976.413293020603, "speed/FLOPS": 203812402227356.7, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 25.36644172668457, "optim/lr": 0.0029940412762214435, "optim/total_tokens": 1720188928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367376, "loss/out": 5.979275226593018, "created_at": "2025-01-14T22:15:09.991194+00:00"} {"global_step": 6563, "acc_step": 0, "speed/wps": 12969.540422481714, "speed/FLOPS": 203704454351229.62, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.43366003036499, "optim/lr": 0.0029940336443472844, "optim/total_tokens": 1720451072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443307, "loss/out": 6.0970988273620605, "created_at": "2025-01-14T22:15:15.047775+00:00"} {"global_step": 6564, "acc_step": 0, "speed/wps": 12976.15165321206, "speed/FLOPS": 203808292814632.75, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.277268886566162, "optim/lr": 0.0029940260075985783, "optim/total_tokens": 1720713216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450994, "loss/out": 6.047868251800537, "created_at": "2025-01-14T22:15:20.102355+00:00"} {"global_step": 6565, "acc_step": 0, "speed/wps": 12966.35611768742, "speed/FLOPS": 203654440468741.9, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.723023414611816, "optim/lr": 0.002994018365975349, "optim/total_tokens": 1720975360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 6.023138046264648, "created_at": "2025-01-14T22:15:25.165796+00:00"} {"global_step": 6566, "acc_step": 0, "speed/wps": 12971.930228243198, "speed/FLOPS": 203741989534651.0, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.387831687927246, "optim/lr": 0.002994010719477623, "optim/total_tokens": 1721237504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440390, "loss/out": 5.969624996185303, "created_at": "2025-01-14T22:15:30.225121+00:00"} {"global_step": 6567, "acc_step": 0, "speed/wps": 12965.602133348168, "speed/FLOPS": 203642598108610.9, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.651129245758057, "optim/lr": 0.0029940030681054235, "optim/total_tokens": 1721499648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496622, "loss/out": 5.998051643371582, "created_at": "2025-01-14T22:15:35.284575+00:00"} {"global_step": 6568, "acc_step": 0, "speed/wps": 12960.78723715269, "speed/FLOPS": 203566973547500.22, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0015, "optim/grad_norm": 7.198360443115234, "optim/lr": 0.0029939954118587764, "optim/total_tokens": 1721761792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424656, "loss/out": 6.112896919250488, "created_at": "2025-01-14T22:15:40.349833+00:00"} {"global_step": 6569, "acc_step": 0, "speed/wps": 12973.077512492837, "speed/FLOPS": 203760009210325.34, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 117.99854278564453, "optim/lr": 0.002993987750737707, "optim/total_tokens": 1722023936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322837, "loss/out": 6.101457118988037, "created_at": "2025-01-14T22:15:45.406130+00:00"} {"global_step": 6570, "acc_step": 0, "speed/wps": 12966.260585440938, "speed/FLOPS": 203652940003535.84, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.343710422515869, "optim/lr": 0.002993980084742239, "optim/total_tokens": 1722286080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350002, "loss/out": 6.102957725524902, "created_at": "2025-01-14T22:15:50.467435+00:00"} {"global_step": 6571, "acc_step": 0, "speed/wps": 12971.308163267713, "speed/FLOPS": 203732219149404.12, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 38.31040954589844, "optim/lr": 0.0029939724138723987, "optim/total_tokens": 1722548224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442492, "loss/out": 6.065572738647461, "created_at": "2025-01-14T22:15:55.524665+00:00"} {"global_step": 6572, "acc_step": 0, "speed/wps": 12975.546012886356, "speed/FLOPS": 203798780401080.56, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.497158527374268, "optim/lr": 0.0029939647381282105, "optim/total_tokens": 1722810368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367082, "loss/out": 6.038929462432861, "created_at": "2025-01-14T22:16:00.582879+00:00"} {"global_step": 6573, "acc_step": 0, "speed/wps": 12972.42809154248, "speed/FLOPS": 203749809161902.1, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 72.42365264892578, "optim/lr": 0.0029939570575097, "optim/total_tokens": 1723072512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449657, "loss/out": 6.153873443603516, "created_at": "2025-01-14T22:16:05.639572+00:00"} {"global_step": 6574, "acc_step": 0, "speed/wps": 12971.878443012345, "speed/FLOPS": 203741176176440.06, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.006199836730957, "optim/lr": 0.0029939493720168913, "optim/total_tokens": 1723334656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422314, "loss/out": 6.304920196533203, "created_at": "2025-01-14T22:16:10.697751+00:00"} {"global_step": 6575, "acc_step": 0, "speed/wps": 12965.738005840374, "speed/FLOPS": 203644732172809.38, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.860959529876709, "optim/lr": 0.0029939416816498108, "optim/total_tokens": 1723596800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 6.142941474914551, "created_at": "2025-01-14T22:16:15.761395+00:00"} {"global_step": 6576, "acc_step": 0, "speed/wps": 12973.427745368275, "speed/FLOPS": 203765510098904.5, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 21.188884735107422, "optim/lr": 0.002993933986408482, "optim/total_tokens": 1723858944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464535, "loss/out": 6.132968902587891, "created_at": "2025-01-14T22:16:20.820862+00:00"} {"global_step": 6577, "acc_step": 0, "speed/wps": 12967.181289615673, "speed/FLOPS": 203667400927780.2, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 18.7857608795166, "optim/lr": 0.0029939262862929313, "optim/total_tokens": 1724121088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438240, "loss/out": 6.309269905090332, "created_at": "2025-01-14T22:16:25.881017+00:00"} {"global_step": 6578, "acc_step": 0, "speed/wps": 12966.97138768275, "speed/FLOPS": 203664104129488.16, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.7618587017059326, "optim/lr": 0.002993918581303183, "optim/total_tokens": 1724383232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 6.318607807159424, "created_at": "2025-01-14T22:16:30.939965+00:00"} {"global_step": 6579, "acc_step": 0, "speed/wps": 12973.219485202204, "speed/FLOPS": 203762239086816.88, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5746607780456543, "optim/lr": 0.002993910871439263, "optim/total_tokens": 1724645376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445693, "loss/out": 6.1287522315979, "created_at": "2025-01-14T22:16:35.996036+00:00"} {"global_step": 6580, "acc_step": 0, "speed/wps": 12964.773407444167, "speed/FLOPS": 203629581829499.84, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 29.76224136352539, "optim/lr": 0.0029939031567011963, "optim/total_tokens": 1724907520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335714, "loss/out": 6.247225284576416, "created_at": "2025-01-14T22:16:41.063342+00:00"} {"global_step": 6581, "acc_step": 0, "speed/wps": 12970.350833199112, "speed/FLOPS": 203717182965167.4, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.0304999351501465, "optim/lr": 0.0029938954370890072, "optim/total_tokens": 1725169664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362816, "loss/out": 6.387360572814941, "created_at": "2025-01-14T22:16:46.120996+00:00"} {"global_step": 6582, "acc_step": 0, "speed/wps": 12967.721626478118, "speed/FLOPS": 203675887660705.53, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.344569206237793, "optim/lr": 0.0029938877126027223, "optim/total_tokens": 1725431808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447032, "loss/out": 6.245604515075684, "created_at": "2025-01-14T22:16:51.185782+00:00"} {"global_step": 6583, "acc_step": 0, "speed/wps": 12966.625201974317, "speed/FLOPS": 203658666807228.2, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.230278015136719, "optim/lr": 0.002993879983242365, "optim/total_tokens": 1725693952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 6.313749313354492, "created_at": "2025-01-14T22:16:56.243694+00:00"} {"global_step": 6584, "acc_step": 0, "speed/wps": 12968.576567995544, "speed/FLOPS": 203689315692049.47, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.001, "optim/grad_norm": 4.105088233947754, "optim/lr": 0.0029938722490079617, "optim/total_tokens": 1725956096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441158, "loss/out": 6.145627021789551, "created_at": "2025-01-14T22:17:01.307406+00:00"} {"global_step": 6585, "acc_step": 0, "speed/wps": 12966.589966784519, "speed/FLOPS": 203658113390154.2, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.378859758377075, "optim/lr": 0.002993864509899538, "optim/total_tokens": 1726218240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425492, "loss/out": 6.2123026847839355, "created_at": "2025-01-14T22:17:06.369678+00:00"} {"global_step": 6586, "acc_step": 0, "speed/wps": 12960.522426868958, "speed/FLOPS": 203562814338106.72, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.775113105773926, "optim/lr": 0.002993856765917119, "optim/total_tokens": 1726480384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466683, "loss/out": 6.157933235168457, "created_at": "2025-01-14T22:17:11.429859+00:00"} {"global_step": 6587, "acc_step": 0, "speed/wps": 12968.323527651233, "speed/FLOPS": 203685341345728.2, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 140.87188720703125, "optim/lr": 0.0029938490170607285, "optim/total_tokens": 1726742528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 6.1837382316589355, "created_at": "2025-01-14T22:17:16.491597+00:00"} {"global_step": 6588, "acc_step": 0, "speed/wps": 12964.681464700756, "speed/FLOPS": 203628137742372.0, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.680359125137329, "optim/lr": 0.002993841263330393, "optim/total_tokens": 1727004672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430506, "loss/out": 6.518229007720947, "created_at": "2025-01-14T22:17:21.553352+00:00"} {"global_step": 6589, "acc_step": 0, "speed/wps": 12964.385303852414, "speed/FLOPS": 203623486129281.12, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.373756408691406, "optim/lr": 0.002993833504726138, "optim/total_tokens": 1727266816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311717, "loss/out": 6.2725090980529785, "created_at": "2025-01-14T22:17:26.615225+00:00"} {"global_step": 6590, "acc_step": 0, "speed/wps": 12962.53011359162, "speed/FLOPS": 203594347816936.8, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.064191818237305, "optim/lr": 0.002993825741247989, "optim/total_tokens": 1727528960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446744, "loss/out": 6.30983829498291, "created_at": "2025-01-14T22:17:31.680184+00:00"} {"global_step": 6591, "acc_step": 0, "speed/wps": 12964.728797422735, "speed/FLOPS": 203628881167812.66, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.097973823547363, "optim/lr": 0.00299381797289597, "optim/total_tokens": 1727791104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451275, "loss/out": 6.230647087097168, "created_at": "2025-01-14T22:17:36.744508+00:00"} {"global_step": 6592, "acc_step": 0, "speed/wps": 12964.046657762674, "speed/FLOPS": 203618167227093.28, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.216337203979492, "optim/lr": 0.002993810199670108, "optim/total_tokens": 1728053248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467490, "loss/out": 6.292630195617676, "created_at": "2025-01-14T22:17:41.808122+00:00"} {"global_step": 6593, "acc_step": 0, "speed/wps": 12963.022885558914, "speed/FLOPS": 203602087477822.78, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.25954532623291, "optim/lr": 0.002993802421570427, "optim/total_tokens": 1728315392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460815, "loss/out": 6.381750106811523, "created_at": "2025-01-14T22:17:46.871474+00:00"} {"global_step": 6594, "acc_step": 0, "speed/wps": 12965.236594937298, "speed/FLOPS": 203636856825565.78, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 23.87612533569336, "optim/lr": 0.0029937946385969524, "optim/total_tokens": 1728577536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447026, "loss/out": 6.287504196166992, "created_at": "2025-01-14T22:17:51.934493+00:00"} {"global_step": 6595, "acc_step": 0, "speed/wps": 12974.208400797532, "speed/FLOPS": 203777771365153.94, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 17.069143295288086, "optim/lr": 0.0029937868507497107, "optim/total_tokens": 1728839680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329272, "loss/out": 6.328808784484863, "created_at": "2025-01-14T22:17:56.990552+00:00"} {"global_step": 6596, "acc_step": 0, "speed/wps": 12973.126884489473, "speed/FLOPS": 203760784665376.72, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 28.388975143432617, "optim/lr": 0.0029937790580287266, "optim/total_tokens": 1729101824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300816, "loss/out": 6.237102508544922, "created_at": "2025-01-14T22:18:02.045667+00:00"} {"global_step": 6597, "acc_step": 0, "speed/wps": 12968.54414960007, "speed/FLOPS": 203688806516601.0, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 11.741804122924805, "optim/lr": 0.002993771260434026, "optim/total_tokens": 1729363968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371881, "loss/out": 6.334279537200928, "created_at": "2025-01-14T22:18:07.106092+00:00"} {"global_step": 6598, "acc_step": 0, "speed/wps": 12972.102584024153, "speed/FLOPS": 203744696619032.53, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.875897407531738, "optim/lr": 0.0029937634579656333, "optim/total_tokens": 1729626112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429413, "loss/out": 6.327150344848633, "created_at": "2025-01-14T22:18:12.163503+00:00"} {"global_step": 6599, "acc_step": 0, "speed/wps": 12956.312859286318, "speed/FLOPS": 203496697294669.34, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 26.780336380004883, "optim/lr": 0.002993755650623575, "optim/total_tokens": 1729888256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444641, "loss/out": 6.3300981521606445, "created_at": "2025-01-14T22:18:17.225270+00:00"} {"global_step": 6600, "acc_step": 0, "speed/wps": 12968.454545749515, "speed/FLOPS": 203687399164999.44, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.3813090324401855, "optim/lr": 0.0029937478384078763, "optim/total_tokens": 1730150400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461913, "loss/out": 6.173192501068115, "created_at": "2025-01-14T22:18:22.282383+00:00"} {"global_step": 6601, "acc_step": 0, "speed/wps": 12968.659899598684, "speed/FLOPS": 203690624529386.4, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.790668487548828, "optim/lr": 0.002993740021318562, "optim/total_tokens": 1730412544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 6.240109920501709, "created_at": "2025-01-14T22:18:27.340464+00:00"} {"global_step": 6602, "acc_step": 0, "speed/wps": 12974.180635585544, "speed/FLOPS": 203777335274346.28, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.5765769481658936, "optim/lr": 0.0029937321993556585, "optim/total_tokens": 1730674688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 6.173417091369629, "created_at": "2025-01-14T22:18:32.395958+00:00"} {"global_step": 6603, "acc_step": 0, "speed/wps": 12976.727300563316, "speed/FLOPS": 203817334147306.6, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.707805633544922, "optim/lr": 0.0029937243725191914, "optim/total_tokens": 1730936832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434807, "loss/out": 6.171909332275391, "created_at": "2025-01-14T22:18:37.468557+00:00"} {"global_step": 6604, "acc_step": 0, "speed/wps": 12970.769021762473, "speed/FLOPS": 203723751191206.28, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.974701881408691, "optim/lr": 0.002993716540809185, "optim/total_tokens": 1731198976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443545, "loss/out": 6.1102681159973145, "created_at": "2025-01-14T22:18:42.528725+00:00"} {"global_step": 6605, "acc_step": 0, "speed/wps": 12965.721378477858, "speed/FLOPS": 203644471017232.66, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.240931987762451, "optim/lr": 0.0029937087042256664, "optim/total_tokens": 1731461120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 339985, "loss/out": 6.247093677520752, "created_at": "2025-01-14T22:18:47.587183+00:00"} {"global_step": 6606, "acc_step": 0, "speed/wps": 12971.292107420264, "speed/FLOPS": 203731966970256.9, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.54887866973877, "optim/lr": 0.00299370086276866, "optim/total_tokens": 1731723264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 404376, "loss/out": 6.209062576293945, "created_at": "2025-01-14T22:18:52.644498+00:00"} {"global_step": 6607, "acc_step": 0, "speed/wps": 12970.956294672345, "speed/FLOPS": 203726692569595.78, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 20.193632125854492, "optim/lr": 0.0029936930164381923, "optim/total_tokens": 1731985408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463485, "loss/out": 6.0716552734375, "created_at": "2025-01-14T22:18:57.707944+00:00"} {"global_step": 6608, "acc_step": 0, "speed/wps": 12971.761528422194, "speed/FLOPS": 203739339872144.28, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.001, "optim/grad_norm": 6.758263111114502, "optim/lr": 0.0029936851652342883, "optim/total_tokens": 1732247552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 6.189678192138672, "created_at": "2025-01-14T22:19:02.764881+00:00"} {"global_step": 6609, "acc_step": 0, "speed/wps": 12967.736435608262, "speed/FLOPS": 203676120258444.3, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.24919891357422, "optim/lr": 0.0029936773091569736, "optim/total_tokens": 1732509696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425179, "loss/out": 6.073700904846191, "created_at": "2025-01-14T22:19:07.823820+00:00"} {"global_step": 6610, "acc_step": 0, "speed/wps": 12974.930371516703, "speed/FLOPS": 203789110907390.25, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 17.748273849487305, "optim/lr": 0.002993669448206274, "optim/total_tokens": 1732771840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476224, "loss/out": 6.10117244720459, "created_at": "2025-01-14T22:19:12.883013+00:00"} {"global_step": 6611, "acc_step": 0, "speed/wps": 12972.200775493218, "speed/FLOPS": 203746238850983.44, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 57.24816131591797, "optim/lr": 0.0029936615823822153, "optim/total_tokens": 1733033984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 6.056365013122559, "created_at": "2025-01-14T22:19:17.940614+00:00"} {"global_step": 6612, "acc_step": 0, "speed/wps": 12973.070720075872, "speed/FLOPS": 203759902526083.12, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0012, "optim/grad_norm": 12.234017372131348, "optim/lr": 0.0029936537116848233, "optim/total_tokens": 1733296128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 6.097897052764893, "created_at": "2025-01-14T22:19:22.998014+00:00"} {"global_step": 6613, "acc_step": 0, "speed/wps": 12976.811768362313, "speed/FLOPS": 203818660830160.75, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.002673149108887, "optim/lr": 0.0029936458361141225, "optim/total_tokens": 1733558272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449409, "loss/out": 6.1617326736450195, "created_at": "2025-01-14T22:19:28.053102+00:00"} {"global_step": 6614, "acc_step": 0, "speed/wps": 12970.618409760193, "speed/FLOPS": 203721385622747.88, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 18.503013610839844, "optim/lr": 0.0029936379556701398, "optim/total_tokens": 1733820416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326058, "loss/out": 6.2563323974609375, "created_at": "2025-01-14T22:19:33.112322+00:00"} {"global_step": 6615, "acc_step": 0, "speed/wps": 12971.509824748948, "speed/FLOPS": 203735386520078.84, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 109.80612182617188, "optim/lr": 0.002993630070352901, "optim/total_tokens": 1734082560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349355, "loss/out": 6.137096405029297, "created_at": "2025-01-14T22:19:38.169360+00:00"} {"global_step": 6616, "acc_step": 0, "speed/wps": 12978.434913769448, "speed/FLOPS": 203844154559215.3, "speed/curr_iter_time": 1.2721, "speed/data_load_time": 0.0007, "optim/grad_norm": 68.8274917602539, "optim/lr": 0.0029936221801624315, "optim/total_tokens": 1734344704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432337, "loss/out": 6.277194023132324, "created_at": "2025-01-14T22:19:43.224801+00:00"} {"global_step": 6617, "acc_step": 0, "speed/wps": 12970.047700525674, "speed/FLOPS": 203712421849983.1, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0013, "optim/grad_norm": 4.632569789886475, "optim/lr": 0.0029936142850987566, "optim/total_tokens": 1734606848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445933, "loss/out": 6.3978376388549805, "created_at": "2025-01-14T22:19:48.284935+00:00"} {"global_step": 6618, "acc_step": 0, "speed/wps": 12975.848581924865, "speed/FLOPS": 203803532663603.78, "speed/curr_iter_time": 1.2717, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.51924467086792, "optim/lr": 0.002993606385161903, "optim/total_tokens": 1734868992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 6.230641841888428, "created_at": "2025-01-14T22:19:53.342510+00:00"} {"global_step": 6619, "acc_step": 0, "speed/wps": 12973.987253594258, "speed/FLOPS": 203774297944438.4, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 62.89125061035156, "optim/lr": 0.002993598480351895, "optim/total_tokens": 1735131136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424963, "loss/out": 6.248639106750488, "created_at": "2025-01-14T22:19:58.401568+00:00"} {"global_step": 6620, "acc_step": 0, "speed/wps": 12966.478387159339, "speed/FLOPS": 203656360878814.25, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 57.893470764160156, "optim/lr": 0.0029935905706687596, "optim/total_tokens": 1735393280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468777, "loss/out": 6.637345314025879, "created_at": "2025-01-14T22:20:03.463511+00:00"} {"global_step": 6621, "acc_step": 0, "speed/wps": 12972.509437493723, "speed/FLOPS": 203751086811847.4, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 26.878074645996094, "optim/lr": 0.0029935826561125227, "optim/total_tokens": 1735655424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449696, "loss/out": 7.140598773956299, "created_at": "2025-01-14T22:20:08.519283+00:00"} {"global_step": 6622, "acc_step": 0, "speed/wps": 12975.060958533451, "speed/FLOPS": 203791161955933.66, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 24.754518508911133, "optim/lr": 0.0029935747366832094, "optim/total_tokens": 1735917568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435329, "loss/out": 6.990671157836914, "created_at": "2025-01-14T22:20:13.574494+00:00"} {"global_step": 6623, "acc_step": 0, "speed/wps": 12968.007658729088, "speed/FLOPS": 203680380190256.53, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.804632186889648, "optim/lr": 0.0029935668123808458, "optim/total_tokens": 1736179712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316156, "loss/out": 6.647712230682373, "created_at": "2025-01-14T22:20:18.632259+00:00"} {"global_step": 6624, "acc_step": 0, "speed/wps": 12965.63844775511, "speed/FLOPS": 203643168476273.16, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.470335960388184, "optim/lr": 0.002993558883205458, "optim/total_tokens": 1736441856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304045, "loss/out": 6.466859817504883, "created_at": "2025-01-14T22:20:23.690411+00:00"} {"global_step": 6625, "acc_step": 0, "speed/wps": 12971.516357290351, "speed/FLOPS": 203735489122618.25, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.088430404663086, "optim/lr": 0.0029935509491570713, "optim/total_tokens": 1736704000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311430, "loss/out": 6.2944536209106445, "created_at": "2025-01-14T22:20:28.747432+00:00"} {"global_step": 6626, "acc_step": 0, "speed/wps": 12970.297745707645, "speed/FLOPS": 203716349153164.03, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.076155185699463, "optim/lr": 0.0029935430102357124, "optim/total_tokens": 1736966144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322689, "loss/out": 6.421962738037109, "created_at": "2025-01-14T22:20:33.805254+00:00"} {"global_step": 6627, "acc_step": 0, "speed/wps": 12972.4052507681, "speed/FLOPS": 203749450416153.94, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 21.36130714416504, "optim/lr": 0.002993535066441407, "optim/total_tokens": 1737228288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338116, "loss/out": 6.2193756103515625, "created_at": "2025-01-14T22:20:38.862492+00:00"} {"global_step": 6628, "acc_step": 0, "speed/wps": 12966.770455074875, "speed/FLOPS": 203660948206771.12, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0009, "optim/grad_norm": 281.1561584472656, "optim/lr": 0.00299352711777418, "optim/total_tokens": 1737490432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432637, "loss/out": 6.363064765930176, "created_at": "2025-01-14T22:20:43.924481+00:00"} {"global_step": 6629, "acc_step": 0, "speed/wps": 12973.002435155287, "speed/FLOPS": 203758830017572.03, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 20.09020233154297, "optim/lr": 0.0029935191642340583, "optim/total_tokens": 1737752576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448077, "loss/out": 6.426033020019531, "created_at": "2025-01-14T22:20:48.982374+00:00"} {"global_step": 6630, "acc_step": 0, "speed/wps": 12969.820647178529, "speed/FLOPS": 203708855665162.2, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 11.779640197753906, "optim/lr": 0.0029935112058210684, "optim/total_tokens": 1738014720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 6.425328254699707, "created_at": "2025-01-14T22:20:54.039587+00:00"} {"global_step": 6631, "acc_step": 0, "speed/wps": 12976.471929580628, "speed/FLOPS": 203813323195106.97, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.8121337890625, "optim/lr": 0.0029935032425352347, "optim/total_tokens": 1738276864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435852, "loss/out": 6.503911018371582, "created_at": "2025-01-14T22:20:59.095283+00:00"} {"global_step": 6632, "acc_step": 0, "speed/wps": 12967.057523021058, "speed/FLOPS": 203665457003338.62, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 74.39866638183594, "optim/lr": 0.002993495274376584, "optim/total_tokens": 1738539008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462672, "loss/out": 6.432051658630371, "created_at": "2025-01-14T22:21:04.157322+00:00"} {"global_step": 6633, "acc_step": 0, "speed/wps": 12973.732531020889, "speed/FLOPS": 203770297176398.16, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 40.08925247192383, "optim/lr": 0.0029934873013451433, "optim/total_tokens": 1738801152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462435, "loss/out": 6.417304992675781, "created_at": "2025-01-14T22:21:09.221249+00:00"} {"global_step": 6634, "acc_step": 0, "speed/wps": 12973.24539507861, "speed/FLOPS": 203762646037029.75, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0009, "optim/grad_norm": 9.470667839050293, "optim/lr": 0.002993479323440937, "optim/total_tokens": 1739063296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329272, "loss/out": 6.443800449371338, "created_at": "2025-01-14T22:21:14.277032+00:00"} {"global_step": 6635, "acc_step": 0, "speed/wps": 12956.547112038332, "speed/FLOPS": 203500376556036.72, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.78923225402832, "optim/lr": 0.0029934713406639915, "optim/total_tokens": 1739325440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376973, "loss/out": 6.240288257598877, "created_at": "2025-01-14T22:21:19.339002+00:00"} {"global_step": 6636, "acc_step": 0, "speed/wps": 12976.883438944806, "speed/FLOPS": 203819786515144.5, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.263299465179443, "optim/lr": 0.0029934633530143335, "optim/total_tokens": 1739587584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427047, "loss/out": 6.220422267913818, "created_at": "2025-01-14T22:21:24.396093+00:00"} {"global_step": 6637, "acc_step": 0, "speed/wps": 12970.856962465929, "speed/FLOPS": 203725132420797.25, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 17.59845542907715, "optim/lr": 0.0029934553604919886, "optim/total_tokens": 1739849728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469069, "loss/out": 6.306708335876465, "created_at": "2025-01-14T22:21:29.452542+00:00"} {"global_step": 6638, "acc_step": 0, "speed/wps": 12977.049799878436, "speed/FLOPS": 203822399442211.03, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.22079086303711, "optim/lr": 0.002993447363096983, "optim/total_tokens": 1740111872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471738, "loss/out": 6.242593288421631, "created_at": "2025-01-14T22:21:34.506507+00:00"} {"global_step": 6639, "acc_step": 0, "speed/wps": 12976.58972835581, "speed/FLOPS": 203815173386743.7, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0006, "optim/grad_norm": 64.48667907714844, "optim/lr": 0.002993439360829342, "optim/total_tokens": 1740374016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448890, "loss/out": 6.252971649169922, "created_at": "2025-01-14T22:21:39.565276+00:00"} {"global_step": 6640, "acc_step": 0, "speed/wps": 12967.472419007656, "speed/FLOPS": 203671973514934.97, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 33.91377258300781, "optim/lr": 0.002993431353689093, "optim/total_tokens": 1740636160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 6.191643714904785, "created_at": "2025-01-14T22:21:44.622813+00:00"} {"global_step": 6641, "acc_step": 0, "speed/wps": 12966.875920441584, "speed/FLOPS": 203662604685279.9, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 45.61096954345703, "optim/lr": 0.0029934233416762618, "optim/total_tokens": 1740898304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439820, "loss/out": 6.293557643890381, "created_at": "2025-01-14T22:21:49.687417+00:00"} {"global_step": 6642, "acc_step": 0, "speed/wps": 12980.61045704481, "speed/FLOPS": 203878324455864.62, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0008, "optim/grad_norm": 34.264442443847656, "optim/lr": 0.002993415324790874, "optim/total_tokens": 1741160448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465057, "loss/out": 6.177451133728027, "created_at": "2025-01-14T22:21:54.745455+00:00"} {"global_step": 6643, "acc_step": 0, "speed/wps": 12970.780956877998, "speed/FLOPS": 203723938648596.16, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 88.26953125, "optim/lr": 0.0029934073030329564, "optim/total_tokens": 1741422592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460287, "loss/out": 6.2282915115356445, "created_at": "2025-01-14T22:21:59.805605+00:00"} {"global_step": 6644, "acc_step": 0, "speed/wps": 12967.666945067745, "speed/FLOPS": 203675028814015.5, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.450126647949219, "optim/lr": 0.0029933992764025344, "optim/total_tokens": 1741684736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443307, "loss/out": 6.285731315612793, "created_at": "2025-01-14T22:22:04.864928+00:00"} {"global_step": 6645, "acc_step": 0, "speed/wps": 12970.316447965808, "speed/FLOPS": 203716642897828.1, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 99.5903091430664, "optim/lr": 0.002993391244899635, "optim/total_tokens": 1741946880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292246, "loss/out": 6.270532608032227, "created_at": "2025-01-14T22:22:09.925381+00:00"} {"global_step": 6646, "acc_step": 0, "speed/wps": 12965.071000162361, "speed/FLOPS": 203634255932074.3, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.12552547454834, "optim/lr": 0.002993383208524284, "optim/total_tokens": 1742209024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452093, "loss/out": 6.257415771484375, "created_at": "2025-01-14T22:22:14.989208+00:00"} {"global_step": 6647, "acc_step": 0, "speed/wps": 12973.855604620067, "speed/FLOPS": 203772230216394.1, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0009, "optim/grad_norm": 7.607422828674316, "optim/lr": 0.0029933751672765077, "optim/total_tokens": 1742471168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459765, "loss/out": 6.192845821380615, "created_at": "2025-01-14T22:22:20.044124+00:00"} {"global_step": 6648, "acc_step": 0, "speed/wps": 12967.397306167757, "speed/FLOPS": 203670793764568.5, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.158213138580322, "optim/lr": 0.002993367121156332, "optim/total_tokens": 1742733312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459222, "loss/out": 6.289792060852051, "created_at": "2025-01-14T22:22:25.107114+00:00"} {"global_step": 6649, "acc_step": 0, "speed/wps": 12967.195588541996, "speed/FLOPS": 203667625512066.88, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 17.14208221435547, "optim/lr": 0.0029933590701637833, "optim/total_tokens": 1742995456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444067, "loss/out": 6.232551574707031, "created_at": "2025-01-14T22:22:30.169701+00:00"} {"global_step": 6650, "acc_step": 0, "speed/wps": 12971.285835727731, "speed/FLOPS": 203731868464707.47, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0008, "optim/grad_norm": 21.211912155151367, "optim/lr": 0.0029933510142988885, "optim/total_tokens": 1743257600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 6.069755554199219, "created_at": "2025-01-14T22:22:35.225860+00:00"} {"global_step": 6651, "acc_step": 0, "speed/wps": 12971.852203821347, "speed/FLOPS": 203740764053889.03, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.71558952331543, "optim/lr": 0.002993342953561673, "optim/total_tokens": 1743519744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449409, "loss/out": 6.1948137283325195, "created_at": "2025-01-14T22:22:40.282730+00:00"} {"global_step": 6652, "acc_step": 0, "speed/wps": 12973.256459366421, "speed/FLOPS": 203762819816873.75, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.567748069763184, "optim/lr": 0.002993334887952164, "optim/total_tokens": 1743781888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 6.063148498535156, "created_at": "2025-01-14T22:22:45.344916+00:00"} {"global_step": 6653, "acc_step": 0, "speed/wps": 12966.078412681578, "speed/FLOPS": 203650078730018.5, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.886805057525635, "optim/lr": 0.002993326817470387, "optim/total_tokens": 1744044032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 6.2802557945251465, "created_at": "2025-01-14T22:22:50.405177+00:00"} {"global_step": 6654, "acc_step": 0, "speed/wps": 12968.796816135076, "speed/FLOPS": 203692774991733.5, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.134979248046875, "optim/lr": 0.0029933187421163684, "optim/total_tokens": 1744306176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313596, "loss/out": 6.155945777893066, "created_at": "2025-01-14T22:22:55.466872+00:00"} {"global_step": 6655, "acc_step": 0, "speed/wps": 12972.279576531564, "speed/FLOPS": 203747476529574.62, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.982219219207764, "optim/lr": 0.002993310661890135, "optim/total_tokens": 1744568320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 378602, "loss/out": 6.300328254699707, "created_at": "2025-01-14T22:23:00.522806+00:00"} {"global_step": 6656, "acc_step": 0, "speed/wps": 12969.601059873556, "speed/FLOPS": 203705406744793.0, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0012, "optim/grad_norm": 8.18580150604248, "optim/lr": 0.0029933025767917127, "optim/total_tokens": 1744830464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443307, "loss/out": 6.149606227874756, "created_at": "2025-01-14T22:23:05.584324+00:00"} {"global_step": 6657, "acc_step": 0, "speed/wps": 12975.615173971184, "speed/FLOPS": 203799866670955.06, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.355338096618652, "optim/lr": 0.002993294486821128, "optim/total_tokens": 1745092608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 6.155574798583984, "created_at": "2025-01-14T22:23:10.638627+00:00"} {"global_step": 6658, "acc_step": 0, "speed/wps": 12973.318415900068, "speed/FLOPS": 203763792929371.88, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.141359329223633, "optim/lr": 0.0029932863919784084, "optim/total_tokens": 1745354752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 6.140454292297363, "created_at": "2025-01-14T22:23:15.696439+00:00"} {"global_step": 6659, "acc_step": 0, "speed/wps": 12973.244419240902, "speed/FLOPS": 203762630710157.6, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.460865020751953, "optim/lr": 0.0029932782922635786, "optim/total_tokens": 1745616896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423342, "loss/out": 6.140718936920166, "created_at": "2025-01-14T22:23:20.751852+00:00"} {"global_step": 6660, "acc_step": 0, "speed/wps": 12970.72638355039, "speed/FLOPS": 203723081499494.5, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.88925313949585, "optim/lr": 0.002993270187676666, "optim/total_tokens": 1745879040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 5.991739749908447, "created_at": "2025-01-14T22:23:25.809382+00:00"} {"global_step": 6661, "acc_step": 0, "speed/wps": 12972.271986222366, "speed/FLOPS": 203747357313338.5, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.4597649574279785, "optim/lr": 0.0029932620782176966, "optim/total_tokens": 1746141184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460556, "loss/out": 6.0417866706848145, "created_at": "2025-01-14T22:23:30.865676+00:00"} {"global_step": 6662, "acc_step": 0, "speed/wps": 12967.751266279834, "speed/FLOPS": 203676353194520.84, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.767642974853516, "optim/lr": 0.002993253963886697, "optim/total_tokens": 1746403328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436916, "loss/out": 6.102919101715088, "created_at": "2025-01-14T22:23:35.929159+00:00"} {"global_step": 6663, "acc_step": 0, "speed/wps": 12975.65422186149, "speed/FLOPS": 203800479971729.75, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.730084419250488, "optim/lr": 0.0029932458446836944, "optim/total_tokens": 1746665472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 284736, "loss/out": 6.032443523406982, "created_at": "2025-01-14T22:23:40.984127+00:00"} {"global_step": 6664, "acc_step": 0, "speed/wps": 12978.646376693674, "speed/FLOPS": 203847475875020.8, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0006, "optim/grad_norm": 30.551477432250977, "optim/lr": 0.0029932377206087135, "optim/total_tokens": 1746927616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345847, "loss/out": 6.064017295837402, "created_at": "2025-01-14T22:23:46.037126+00:00"} {"global_step": 6665, "acc_step": 0, "speed/wps": 12970.737541566137, "speed/FLOPS": 203723256751464.5, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.13176155090332, "optim/lr": 0.002993229591661783, "optim/total_tokens": 1747189760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 341575, "loss/out": 6.159200668334961, "created_at": "2025-01-14T22:23:51.094576+00:00"} {"global_step": 6666, "acc_step": 0, "speed/wps": 12973.800897778487, "speed/FLOPS": 203771370970272.0, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.125962734222412, "optim/lr": 0.002993221457842928, "optim/total_tokens": 1747451904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329132, "loss/out": 6.027597904205322, "created_at": "2025-01-14T22:23:56.149695+00:00"} {"global_step": 6667, "acc_step": 0, "speed/wps": 12977.159807299699, "speed/FLOPS": 203824127259928.0, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.433743000030518, "optim/lr": 0.002993213319152175, "optim/total_tokens": 1747714048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366558, "loss/out": 6.023688793182373, "created_at": "2025-01-14T22:24:01.212122+00:00"} {"global_step": 6668, "acc_step": 0, "speed/wps": 12969.900420933087, "speed/FLOPS": 203710108621598.16, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.70062780380249, "optim/lr": 0.0029932051755895512, "optim/total_tokens": 1747976192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446747, "loss/out": 5.941079139709473, "created_at": "2025-01-14T22:24:06.269655+00:00"} {"global_step": 6669, "acc_step": 0, "speed/wps": 12976.498345399625, "speed/FLOPS": 203813738091843.44, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.7027530670166016, "optim/lr": 0.0029931970271550824, "optim/total_tokens": 1748238336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 6.204178810119629, "created_at": "2025-01-14T22:24:11.323701+00:00"} {"global_step": 6670, "acc_step": 0, "speed/wps": 12959.768168477001, "speed/FLOPS": 203550967673601.4, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.396735429763794, "optim/lr": 0.0029931888738487966, "optim/total_tokens": 1748500480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429217, "loss/out": 5.973216533660889, "created_at": "2025-01-14T22:24:16.384695+00:00"} {"global_step": 6671, "acc_step": 0, "speed/wps": 12974.921493398386, "speed/FLOPS": 203788971464342.28, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.004786968231201, "optim/lr": 0.0029931807156707195, "optim/total_tokens": 1748762624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471454, "loss/out": 6.096678256988525, "created_at": "2025-01-14T22:24:21.442889+00:00"} {"global_step": 6672, "acc_step": 0, "speed/wps": 12976.791946556932, "speed/FLOPS": 203818349501469.22, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0011, "optim/grad_norm": 10.77381420135498, "optim/lr": 0.002993172552620877, "optim/total_tokens": 1749024768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 6.145512104034424, "created_at": "2025-01-14T22:24:26.497201+00:00"} {"global_step": 6673, "acc_step": 0, "speed/wps": 12976.632618103145, "speed/FLOPS": 203815847029162.28, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.153781414031982, "optim/lr": 0.0029931643846992967, "optim/total_tokens": 1749286912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430791, "loss/out": 5.994682788848877, "created_at": "2025-01-14T22:24:31.551723+00:00"} {"global_step": 6674, "acc_step": 0, "speed/wps": 12975.040000891428, "speed/FLOPS": 203790832787366.84, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.923758506774902, "optim/lr": 0.002993156211906005, "optim/total_tokens": 1749549056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332206, "loss/out": 6.113647937774658, "created_at": "2025-01-14T22:24:36.606769+00:00"} {"global_step": 6675, "acc_step": 0, "speed/wps": 12973.53455326019, "speed/FLOPS": 203767187663616.44, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.710141181945801, "optim/lr": 0.002993148034241029, "optim/total_tokens": 1749811200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363102, "loss/out": 6.024453163146973, "created_at": "2025-01-14T22:24:41.665943+00:00"} {"global_step": 6676, "acc_step": 0, "speed/wps": 12974.422326328011, "speed/FLOPS": 203781131359574.6, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 31.164674758911133, "optim/lr": 0.0029931398517043946, "optim/total_tokens": 1750073344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450746, "loss/out": 6.110716342926025, "created_at": "2025-01-14T22:24:46.720638+00:00"} {"global_step": 6677, "acc_step": 0, "speed/wps": 12966.90579054081, "speed/FLOPS": 203663073836234.06, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 25.64402198791504, "optim/lr": 0.002993131664296129, "optim/total_tokens": 1750335488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464007, "loss/out": 6.115650177001953, "created_at": "2025-01-14T22:24:51.781461+00:00"} {"global_step": 6678, "acc_step": 0, "speed/wps": 12979.577813804752, "speed/FLOPS": 203862105374779.0, "speed/curr_iter_time": 1.2723, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.7584400177001953, "optim/lr": 0.002993123472016259, "optim/total_tokens": 1750597632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447791, "loss/out": 6.172529697418213, "created_at": "2025-01-14T22:24:56.838677+00:00"} {"global_step": 6679, "acc_step": 0, "speed/wps": 12975.308292532583, "speed/FLOPS": 203795046676262.56, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.0996453762054443, "optim/lr": 0.0029931152748648104, "optim/total_tokens": 1750859776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455532, "loss/out": 6.082883358001709, "created_at": "2025-01-14T22:25:01.894363+00:00"} {"global_step": 6680, "acc_step": 0, "speed/wps": 12970.012747944515, "speed/FLOPS": 203711872871666.12, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.883459091186523, "optim/lr": 0.0029931070728418108, "optim/total_tokens": 1751121920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 6.0999884605407715, "created_at": "2025-01-14T22:25:06.952202+00:00"} {"global_step": 6681, "acc_step": 0, "speed/wps": 12979.707282951465, "speed/FLOPS": 203864138865637.38, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 16.660879135131836, "optim/lr": 0.002993098865947287, "optim/total_tokens": 1751384064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 6.033515930175781, "created_at": "2025-01-14T22:25:12.007996+00:00"} {"global_step": 6682, "acc_step": 0, "speed/wps": 12972.135199563878, "speed/FLOPS": 203745208890902.38, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 19.921497344970703, "optim/lr": 0.0029930906541812655, "optim/total_tokens": 1751646208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439302, "loss/out": 6.11362361907959, "created_at": "2025-01-14T22:25:17.065396+00:00"} {"global_step": 6683, "acc_step": 0, "speed/wps": 12983.896300934039, "speed/FLOPS": 203929933149367.62, "speed/curr_iter_time": 1.2721, "speed/data_load_time": 0.0006, "optim/grad_norm": 15.825496673583984, "optim/lr": 0.0029930824375437736, "optim/total_tokens": 1751908352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437948, "loss/out": 6.237558364868164, "created_at": "2025-01-14T22:25:22.120314+00:00"} {"global_step": 6684, "acc_step": 0, "speed/wps": 12983.751152605982, "speed/FLOPS": 203927653395420.8, "speed/curr_iter_time": 1.2723, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.473712205886841, "optim/lr": 0.002993074216034837, "optim/total_tokens": 1752170496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 6.094121932983398, "created_at": "2025-01-14T22:25:27.172198+00:00"} {"global_step": 6685, "acc_step": 0, "speed/wps": 12978.202315144463, "speed/FLOPS": 203840501278185.94, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 18.844820022583008, "optim/lr": 0.0029930659896544834, "optim/total_tokens": 1752432640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450990, "loss/out": 6.117632865905762, "created_at": "2025-01-14T22:25:32.233795+00:00"} {"global_step": 6686, "acc_step": 0, "speed/wps": 12979.680420728202, "speed/FLOPS": 203863716957508.47, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.41730785369873, "optim/lr": 0.0029930577584027394, "optim/total_tokens": 1752694784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337583, "loss/out": 6.222874164581299, "created_at": "2025-01-14T22:25:37.286926+00:00"} {"global_step": 6687, "acc_step": 0, "speed/wps": 12974.918967048425, "speed/FLOPS": 203788931784544.56, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.872573852539062, "optim/lr": 0.002993049522279632, "optim/total_tokens": 1752956928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345847, "loss/out": 6.042019367218018, "created_at": "2025-01-14T22:25:42.348375+00:00"} {"global_step": 6688, "acc_step": 0, "speed/wps": 12972.549291762221, "speed/FLOPS": 203751712777899.66, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 24.05415153503418, "optim/lr": 0.002993041281285188, "optim/total_tokens": 1753219072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432431, "loss/out": 6.109479904174805, "created_at": "2025-01-14T22:25:47.404858+00:00"} {"global_step": 6689, "acc_step": 0, "speed/wps": 12976.117313785411, "speed/FLOPS": 203807753466749.9, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 11.938729286193848, "optim/lr": 0.0029930330354194343, "optim/total_tokens": 1753481216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471976, "loss/out": 6.168374061584473, "created_at": "2025-01-14T22:25:52.459250+00:00"} {"global_step": 6690, "acc_step": 0, "speed/wps": 12972.072099320148, "speed/FLOPS": 203744217814865.78, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 17.802053451538086, "optim/lr": 0.002993024784682398, "optim/total_tokens": 1753743360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 6.048178672790527, "created_at": "2025-01-14T22:25:57.514980+00:00"} {"global_step": 6691, "acc_step": 0, "speed/wps": 12969.549424709352, "speed/FLOPS": 203704595743585.03, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.310711860656738, "optim/lr": 0.002993016529074105, "optim/total_tokens": 1754005504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 6.130252361297607, "created_at": "2025-01-14T22:26:02.577739+00:00"} {"global_step": 6692, "acc_step": 0, "speed/wps": 12972.929649880145, "speed/FLOPS": 203757686824808.47, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.230323314666748, "optim/lr": 0.0029930082685945838, "optim/total_tokens": 1754267648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431076, "loss/out": 6.042981147766113, "created_at": "2025-01-14T22:26:07.637290+00:00"} {"global_step": 6693, "acc_step": 0, "speed/wps": 12973.184108928639, "speed/FLOPS": 203761683453828.47, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.05198860168457, "optim/lr": 0.00299300000324386, "optim/total_tokens": 1754529792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440919, "loss/out": 6.053725719451904, "created_at": "2025-01-14T22:26:12.693210+00:00"} {"global_step": 6694, "acc_step": 0, "speed/wps": 12966.541516620593, "speed/FLOPS": 203657352413752.22, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 31.428129196166992, "optim/lr": 0.0029929917330219612, "optim/total_tokens": 1754791936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 6.1486711502075195, "created_at": "2025-01-14T22:26:17.756404+00:00"} {"global_step": 6695, "acc_step": 0, "speed/wps": 12981.432889433643, "speed/FLOPS": 203891241886671.78, "speed/curr_iter_time": 1.2719, "speed/data_load_time": 0.0008, "optim/grad_norm": 10.896547317504883, "optim/lr": 0.0029929834579289142, "optim/total_tokens": 1755054080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 6.052610874176025, "created_at": "2025-01-14T22:26:22.816563+00:00"} {"global_step": 6696, "acc_step": 0, "speed/wps": 12967.523985062642, "speed/FLOPS": 203672783430687.72, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.719059944152832, "optim/lr": 0.002992975177964747, "optim/total_tokens": 1755316224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322837, "loss/out": 6.14640998840332, "created_at": "2025-01-14T22:26:27.875911+00:00"} {"global_step": 6697, "acc_step": 0, "speed/wps": 12975.458089936892, "speed/FLOPS": 203797399450340.44, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 62.178916931152344, "optim/lr": 0.002992966893129485, "optim/total_tokens": 1755578368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416907, "loss/out": 6.044043064117432, "created_at": "2025-01-14T22:26:32.931497+00:00"} {"global_step": 6698, "acc_step": 0, "speed/wps": 12969.919012080923, "speed/FLOPS": 203710400621121.84, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.177743911743164, "optim/lr": 0.0029929586034231556, "optim/total_tokens": 1755840512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427880, "loss/out": 6.131493091583252, "created_at": "2025-01-14T22:26:37.991718+00:00"} {"global_step": 6699, "acc_step": 0, "speed/wps": 12964.675376155581, "speed/FLOPS": 203628042113403.5, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.441476821899414, "optim/lr": 0.002992950308845787, "optim/total_tokens": 1756102656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458430, "loss/out": 6.1070356369018555, "created_at": "2025-01-14T22:26:43.050171+00:00"} {"global_step": 6700, "acc_step": 0, "speed/wps": 12983.119296417135, "speed/FLOPS": 203917729225713.56, "speed/curr_iter_time": 1.2722, "speed/data_load_time": 0.0012, "optim/grad_norm": 4.320882320404053, "optim/lr": 0.002992942009397405, "optim/total_tokens": 1756364800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 6.158785343170166, "created_at": "2025-01-14T22:26:48.103239+00:00"} {"global_step": 6701, "acc_step": 0, "speed/wps": 12975.926675964645, "speed/FLOPS": 203804759237810.97, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.380506992340088, "optim/lr": 0.002992933705078037, "optim/total_tokens": 1756626944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 5.977347373962402, "created_at": "2025-01-14T22:26:53.160422+00:00"} {"global_step": 6702, "acc_step": 0, "speed/wps": 12971.63875040445, "speed/FLOPS": 203737411474660.94, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 15.191514015197754, "optim/lr": 0.0029929253958877106, "optim/total_tokens": 1756889088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433708, "loss/out": 6.053400993347168, "created_at": "2025-01-14T22:26:58.221689+00:00"} {"global_step": 6703, "acc_step": 0, "speed/wps": 12978.741093330107, "speed/FLOPS": 203848963529950.28, "speed/curr_iter_time": 1.2715, "speed/data_load_time": 0.0007, "optim/grad_norm": 46.21462631225586, "optim/lr": 0.0029929170818264526, "optim/total_tokens": 1757151232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463485, "loss/out": 5.961859226226807, "created_at": "2025-01-14T22:27:03.280484+00:00"} {"global_step": 6704, "acc_step": 0, "speed/wps": 12978.530706881862, "speed/FLOPS": 203845659121679.3, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 21.72724723815918, "optim/lr": 0.0029929087628942897, "optim/total_tokens": 1757413376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 6.18717098236084, "created_at": "2025-01-14T22:27:08.333709+00:00"} {"global_step": 6705, "acc_step": 0, "speed/wps": 12978.245679911235, "speed/FLOPS": 203841182381437.53, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.398370742797852, "optim/lr": 0.0029929004390912496, "optim/total_tokens": 1757675520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434807, "loss/out": 6.01070499420166, "created_at": "2025-01-14T22:27:13.390833+00:00"} {"global_step": 6706, "acc_step": 0, "speed/wps": 12966.388558593886, "speed/FLOPS": 203654949997756.44, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 12.251235961914062, "optim/lr": 0.0029928921104173593, "optim/total_tokens": 1757937664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453131, "loss/out": 6.102208614349365, "created_at": "2025-01-14T22:27:18.449725+00:00"} {"global_step": 6707, "acc_step": 0, "speed/wps": 12966.862245855247, "speed/FLOPS": 203662389907110.62, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.5296549797058105, "optim/lr": 0.002992883776872646, "optim/total_tokens": 1758199808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 6.001501083374023, "created_at": "2025-01-14T22:27:23.509797+00:00"} {"global_step": 6708, "acc_step": 0, "speed/wps": 12969.426148343662, "speed/FLOPS": 203702659518862.4, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 72.6129150390625, "optim/lr": 0.002992875438457137, "optim/total_tokens": 1758461952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421476, "loss/out": 6.076193332672119, "created_at": "2025-01-14T22:27:28.566658+00:00"} {"global_step": 6709, "acc_step": 0, "speed/wps": 12966.370515141414, "speed/FLOPS": 203654666600541.06, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 23.255855560302734, "optim/lr": 0.002992867095170859, "optim/total_tokens": 1758724096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302343, "loss/out": 6.115200042724609, "created_at": "2025-01-14T22:27:33.624897+00:00"} {"global_step": 6710, "acc_step": 0, "speed/wps": 12967.445963455502, "speed/FLOPS": 203671557994134.66, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.03179931640625, "optim/lr": 0.00299285874701384, "optim/total_tokens": 1758986240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 400828, "loss/out": 6.146049499511719, "created_at": "2025-01-14T22:27:38.685657+00:00"} {"global_step": 6711, "acc_step": 0, "speed/wps": 12965.77573529973, "speed/FLOPS": 203645324765814.28, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.447257041931152, "optim/lr": 0.002992850393986106, "optim/total_tokens": 1759248384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428118, "loss/out": 6.036644458770752, "created_at": "2025-01-14T22:27:43.743770+00:00"} {"global_step": 6712, "acc_step": 0, "speed/wps": 12962.756701613367, "speed/FLOPS": 203597906693182.9, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.174778938293457, "optim/lr": 0.0029928420360876857, "optim/total_tokens": 1759510528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 6.098613739013672, "created_at": "2025-01-14T22:27:48.807704+00:00"} {"global_step": 6713, "acc_step": 0, "speed/wps": 12969.893347370327, "speed/FLOPS": 203709997521574.9, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.0787410736084, "optim/lr": 0.002992833673318605, "optim/total_tokens": 1759772672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 6.114572525024414, "created_at": "2025-01-14T22:27:53.864538+00:00"} {"global_step": 6714, "acc_step": 0, "speed/wps": 12973.586272782475, "speed/FLOPS": 203767999989784.0, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.951242923736572, "optim/lr": 0.0029928253056788925, "optim/total_tokens": 1760034816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431605, "loss/out": 6.155837059020996, "created_at": "2025-01-14T22:27:58.923415+00:00"} {"global_step": 6715, "acc_step": 0, "speed/wps": 12964.939868880554, "speed/FLOPS": 203632196335100.28, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.060616493225098, "optim/lr": 0.002992816933168575, "optim/total_tokens": 1760296960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 6.028931617736816, "created_at": "2025-01-14T22:28:03.984097+00:00"} {"global_step": 6716, "acc_step": 0, "speed/wps": 12976.677439304807, "speed/FLOPS": 203816551007725.16, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.001, "optim/grad_norm": 73.19525909423828, "optim/lr": 0.0029928085557876798, "optim/total_tokens": 1760559104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 6.077795505523682, "created_at": "2025-01-14T22:28:09.038959+00:00"} {"global_step": 6717, "acc_step": 0, "speed/wps": 12975.306337844067, "speed/FLOPS": 203795015975193.38, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.263419151306152, "optim/lr": 0.002992800173536234, "optim/total_tokens": 1760821248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447791, "loss/out": 6.182671070098877, "created_at": "2025-01-14T22:28:14.099144+00:00"} {"global_step": 6718, "acc_step": 0, "speed/wps": 12965.450893592068, "speed/FLOPS": 203640222680416.16, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 173.42977905273438, "optim/lr": 0.0029927917864142647, "optim/total_tokens": 1761083392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317214, "loss/out": 6.144913196563721, "created_at": "2025-01-14T22:28:19.158751+00:00"} {"global_step": 6719, "acc_step": 0, "speed/wps": 12970.615325640872, "speed/FLOPS": 203721337182415.94, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 16.640958786010742, "optim/lr": 0.0029927833944218, "optim/total_tokens": 1761345536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335181, "loss/out": 6.293272972106934, "created_at": "2025-01-14T22:28:24.216188+00:00"} {"global_step": 6720, "acc_step": 0, "speed/wps": 12971.126310106032, "speed/FLOPS": 203729362895607.5, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0006, "optim/grad_norm": 23.285024642944336, "optim/lr": 0.0029927749975588672, "optim/total_tokens": 1761607680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292724, "loss/out": 6.22401237487793, "created_at": "2025-01-14T22:28:29.272187+00:00"} {"global_step": 6721, "acc_step": 0, "speed/wps": 12969.011567126487, "speed/FLOPS": 203696147951152.38, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.020632743835449, "optim/lr": 0.0029927665958254935, "optim/total_tokens": 1761869824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364158, "loss/out": 6.234567642211914, "created_at": "2025-01-14T22:28:34.330061+00:00"} {"global_step": 6722, "acc_step": 0, "speed/wps": 12969.050085288396, "speed/FLOPS": 203696752931814.66, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 1686.37646484375, "optim/lr": 0.002992758189221706, "optim/total_tokens": 1762131968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452330, "loss/out": 6.095161437988281, "created_at": "2025-01-14T22:28:39.390326+00:00"} {"global_step": 6723, "acc_step": 0, "speed/wps": 12974.37927656116, "speed/FLOPS": 203780455203829.62, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 47.38300704956055, "optim/lr": 0.002992749777747532, "optim/total_tokens": 1762394112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445976, "loss/out": 6.194327354431152, "created_at": "2025-01-14T22:28:44.445894+00:00"} {"global_step": 6724, "acc_step": 0, "speed/wps": 12968.127949467096, "speed/FLOPS": 203682269521593.75, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.904901504516602, "optim/lr": 0.0029927413614029997, "optim/total_tokens": 1762656256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 6.2269744873046875, "created_at": "2025-01-14T22:28:49.509786+00:00"} {"global_step": 6725, "acc_step": 0, "speed/wps": 12969.791406689359, "speed/FLOPS": 203708396403099.12, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.242284774780273, "optim/lr": 0.0029927329401881362, "optim/total_tokens": 1762918400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435567, "loss/out": 6.235514163970947, "created_at": "2025-01-14T22:28:54.566366+00:00"} {"global_step": 6726, "acc_step": 0, "speed/wps": 12976.613387916346, "speed/FLOPS": 203815544992653.53, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.315275192260742, "optim/lr": 0.0029927245141029694, "optim/total_tokens": 1763180544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461100, "loss/out": 6.175619602203369, "created_at": "2025-01-14T22:28:59.625028+00:00"} {"global_step": 6727, "acc_step": 0, "speed/wps": 12968.688362990433, "speed/FLOPS": 203691071586066.62, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 154.93270874023438, "optim/lr": 0.0029927160831475256, "optim/total_tokens": 1763442688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444878, "loss/out": 6.0590715408325195, "created_at": "2025-01-14T22:29:04.683184+00:00"} {"global_step": 6728, "acc_step": 0, "speed/wps": 12969.85114042556, "speed/FLOPS": 203709334603508.94, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.68769359588623, "optim/lr": 0.0029927076473218336, "optim/total_tokens": 1763704832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329272, "loss/out": 6.157620429992676, "created_at": "2025-01-14T22:29:09.740211+00:00"} {"global_step": 6729, "acc_step": 0, "speed/wps": 12969.52320928198, "speed/FLOPS": 203704183994274.4, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.3325047492980957, "optim/lr": 0.00299269920662592, "optim/total_tokens": 1763966976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300465, "loss/out": 6.185274600982666, "created_at": "2025-01-14T22:29:14.803462+00:00"} {"global_step": 6730, "acc_step": 0, "speed/wps": 12971.307687449771, "speed/FLOPS": 203732211676029.53, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0007, "optim/grad_norm": 141.17849731445312, "optim/lr": 0.002992690761059813, "optim/total_tokens": 1764229120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443015, "loss/out": 6.231962203979492, "created_at": "2025-01-14T22:29:19.862819+00:00"} {"global_step": 6731, "acc_step": 0, "speed/wps": 12972.561447781363, "speed/FLOPS": 203751903704884.4, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 15.930948257446289, "optim/lr": 0.00299268231062354, "optim/total_tokens": 1764491264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438240, "loss/out": 6.191708564758301, "created_at": "2025-01-14T22:29:24.918439+00:00"} {"global_step": 6732, "acc_step": 0, "speed/wps": 12968.323307776314, "speed/FLOPS": 203685337892290.47, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 114.73207092285156, "optim/lr": 0.002992673855317128, "optim/total_tokens": 1764753408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432365, "loss/out": 6.320830345153809, "created_at": "2025-01-14T22:29:29.976642+00:00"} {"global_step": 6733, "acc_step": 0, "speed/wps": 12974.17524617719, "speed/FLOPS": 203777250626282.25, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 23.62384605407715, "optim/lr": 0.0029926653951406054, "optim/total_tokens": 1765015552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441971, "loss/out": 6.415868282318115, "created_at": "2025-01-14T22:29:35.036359+00:00"} {"global_step": 6734, "acc_step": 0, "speed/wps": 12973.136370040362, "speed/FLOPS": 203760933648987.97, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0009, "optim/grad_norm": 65.9732437133789, "optim/lr": 0.002992656930093999, "optim/total_tokens": 1765277696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 6.285056114196777, "created_at": "2025-01-14T22:29:40.095782+00:00"} {"global_step": 6735, "acc_step": 0, "speed/wps": 12962.035230221818, "speed/FLOPS": 203586574993573.06, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.530888557434082, "optim/lr": 0.0029926484601773373, "optim/total_tokens": 1765539840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 6.337257385253906, "created_at": "2025-01-14T22:29:45.160251+00:00"} {"global_step": 6736, "acc_step": 0, "speed/wps": 12974.313616408908, "speed/FLOPS": 203779423920912.0, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 2462.781494140625, "optim/lr": 0.0029926399853906476, "optim/total_tokens": 1765801984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 6.1793365478515625, "created_at": "2025-01-14T22:29:50.223414+00:00"} {"global_step": 6737, "acc_step": 0, "speed/wps": 12978.175009157518, "speed/FLOPS": 203840072400138.6, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.37373161315918, "optim/lr": 0.0029926315057339573, "optim/total_tokens": 1766064128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444644, "loss/out": 6.021543502807617, "created_at": "2025-01-14T22:29:55.276935+00:00"} {"global_step": 6738, "acc_step": 0, "speed/wps": 12970.29547456914, "speed/FLOPS": 203716313481812.97, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.8843228816986084, "optim/lr": 0.0029926230212072944, "optim/total_tokens": 1766326272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320571, "loss/out": 6.156652450561523, "created_at": "2025-01-14T22:30:00.334059+00:00"} {"global_step": 6739, "acc_step": 0, "speed/wps": 12971.1484697278, "speed/FLOPS": 203729710942917.88, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 25.93204116821289, "optim/lr": 0.002992614531810686, "optim/total_tokens": 1766588416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292773, "loss/out": 6.086246967315674, "created_at": "2025-01-14T22:30:05.395445+00:00"} {"global_step": 6740, "acc_step": 0, "speed/wps": 12975.02216063238, "speed/FLOPS": 203790552581583.4, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.235635757446289, "optim/lr": 0.0029926060375441606, "optim/total_tokens": 1766850560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420921, "loss/out": 6.209461212158203, "created_at": "2025-01-14T22:30:10.450071+00:00"} {"global_step": 6741, "acc_step": 0, "speed/wps": 12968.516575927228, "speed/FLOPS": 203688373434178.8, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0009, "optim/grad_norm": 17.522302627563477, "optim/lr": 0.002992597538407745, "optim/total_tokens": 1767112704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435567, "loss/out": 6.182950973510742, "created_at": "2025-01-14T22:30:15.513339+00:00"} {"global_step": 6742, "acc_step": 0, "speed/wps": 12973.957417837562, "speed/FLOPS": 203773829332880.9, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 129.4888153076172, "optim/lr": 0.002992589034401468, "optim/total_tokens": 1767374848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 6.106878757476807, "created_at": "2025-01-14T22:30:20.568895+00:00"} {"global_step": 6743, "acc_step": 0, "speed/wps": 12973.287427992174, "speed/FLOPS": 203763306221696.56, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0006, "optim/grad_norm": 11.205846786499023, "optim/lr": 0.0029925805255253566, "optim/total_tokens": 1767636992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465870, "loss/out": 6.137787818908691, "created_at": "2025-01-14T22:30:25.624053+00:00"} {"global_step": 6744, "acc_step": 0, "speed/wps": 12971.067525787843, "speed/FLOPS": 203728439607112.22, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0012, "optim/grad_norm": 10.745762825012207, "optim/lr": 0.0029925720117794383, "optim/total_tokens": 1767899136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448605, "loss/out": 6.121153354644775, "created_at": "2025-01-14T22:30:30.686296+00:00"} {"global_step": 6745, "acc_step": 0, "speed/wps": 12968.343749005824, "speed/FLOPS": 203685658949889.1, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.3762788772583, "optim/lr": 0.0029925634931637418, "optim/total_tokens": 1768161280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449657, "loss/out": 6.081177234649658, "created_at": "2025-01-14T22:30:35.747670+00:00"} {"global_step": 6746, "acc_step": 0, "speed/wps": 12970.98401022174, "speed/FLOPS": 203727127880384.2, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.525635719299316, "optim/lr": 0.002992554969678295, "optim/total_tokens": 1768423424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423864, "loss/out": 6.078484535217285, "created_at": "2025-01-14T22:30:40.805922+00:00"} {"global_step": 6747, "acc_step": 0, "speed/wps": 12957.966603036833, "speed/FLOPS": 203522671612753.22, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0012, "optim/grad_norm": 8.905847549438477, "optim/lr": 0.002992546441323124, "optim/total_tokens": 1768685568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 6.120905876159668, "created_at": "2025-01-14T22:30:45.872820+00:00"} {"global_step": 6748, "acc_step": 0, "speed/wps": 12967.735472547778, "speed/FLOPS": 203676105132256.03, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.6338348388671875, "optim/lr": 0.002992537908098258, "optim/total_tokens": 1768947712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418245, "loss/out": 5.927820682525635, "created_at": "2025-01-14T22:30:50.931081+00:00"} {"global_step": 6749, "acc_step": 0, "speed/wps": 12965.48881249249, "speed/FLOPS": 203640818248853.28, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.576094627380371, "optim/lr": 0.002992529370003725, "optim/total_tokens": 1769209856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328845, "loss/out": 6.018786430358887, "created_at": "2025-01-14T22:30:55.997187+00:00"} {"global_step": 6750, "acc_step": 0, "speed/wps": 12973.832615299974, "speed/FLOPS": 203771869137531.4, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 25.105228424072266, "optim/lr": 0.002992520827039552, "optim/total_tokens": 1769472000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452325, "loss/out": 6.096602439880371, "created_at": "2025-01-14T22:31:01.054998+00:00"} {"global_step": 6751, "acc_step": 0, "speed/wps": 12970.279968691084, "speed/FLOPS": 203716069940690.78, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.270669460296631, "optim/lr": 0.0029925122792057676, "optim/total_tokens": 1769734144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 6.119750022888184, "created_at": "2025-01-14T22:31:06.116496+00:00"} {"global_step": 6752, "acc_step": 0, "speed/wps": 12971.134116653831, "speed/FLOPS": 203729485508168.03, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.243974685668945, "optim/lr": 0.0029925037265023996, "optim/total_tokens": 1769996288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461622, "loss/out": 5.933413982391357, "created_at": "2025-01-14T22:31:11.179735+00:00"} {"global_step": 6753, "acc_step": 0, "speed/wps": 12967.663771543184, "speed/FLOPS": 203674978969451.4, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.0816798210144043, "optim/lr": 0.0029924951689294754, "optim/total_tokens": 1770258432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 6.050373077392578, "created_at": "2025-01-14T22:31:16.237177+00:00"} {"global_step": 6754, "acc_step": 0, "speed/wps": 12978.40802090745, "speed/FLOPS": 203843732169862.97, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.026794910430908, "optim/lr": 0.0029924866064870233, "optim/total_tokens": 1770520576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450990, "loss/out": 5.981456756591797, "created_at": "2025-01-14T22:31:21.299512+00:00"} {"global_step": 6755, "acc_step": 0, "speed/wps": 12971.36615593295, "speed/FLOPS": 203733130003901.97, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.560513973236084, "optim/lr": 0.0029924780391750714, "optim/total_tokens": 1770782720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320721, "loss/out": 6.014289855957031, "created_at": "2025-01-14T22:31:26.355411+00:00"} {"global_step": 6756, "acc_step": 0, "speed/wps": 12970.357689707678, "speed/FLOPS": 203717290656056.25, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 18.79985237121582, "optim/lr": 0.0029924694669936475, "optim/total_tokens": 1771044864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348659, "loss/out": 5.978726863861084, "created_at": "2025-01-14T22:31:31.417596+00:00"} {"global_step": 6757, "acc_step": 0, "speed/wps": 12975.919973880911, "speed/FLOPS": 203804653972376.3, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.925556659698486, "optim/lr": 0.002992460889942779, "optim/total_tokens": 1771307008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402220, "loss/out": 5.953145980834961, "created_at": "2025-01-14T22:31:36.472112+00:00"} {"global_step": 6758, "acc_step": 0, "speed/wps": 12974.48096403392, "speed/FLOPS": 203782052345322.56, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0006, "optim/grad_norm": 25.750965118408203, "optim/lr": 0.0029924523080224947, "optim/total_tokens": 1771569152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 392536, "loss/out": 5.942419052124023, "created_at": "2025-01-14T22:31:41.529918+00:00"} {"global_step": 6759, "acc_step": 0, "speed/wps": 12975.524845521171, "speed/FLOPS": 203798447938523.3, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.885894775390625, "optim/lr": 0.002992443721232822, "optim/total_tokens": 1771831296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357724, "loss/out": 6.086271286010742, "created_at": "2025-01-14T22:31:46.584578+00:00"} {"global_step": 6760, "acc_step": 0, "speed/wps": 12969.315205920775, "speed/FLOPS": 203700917015660.1, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.001, "optim/grad_norm": 2.393601894378662, "optim/lr": 0.0029924351295737895, "optim/total_tokens": 1772093440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451559, "loss/out": 6.0448713302612305, "created_at": "2025-01-14T22:31:51.663409+00:00"} {"global_step": 6761, "acc_step": 0, "speed/wps": 12976.408205658767, "speed/FLOPS": 203812322323348.28, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.938663482666016, "optim/lr": 0.002992426533045425, "optim/total_tokens": 1772355584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466369, "loss/out": 6.063723564147949, "created_at": "2025-01-14T22:31:56.718862+00:00"} {"global_step": 6762, "acc_step": 0, "speed/wps": 12959.601572501955, "speed/FLOPS": 203548351054886.44, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 537.4558715820312, "optim/lr": 0.002992417931647756, "optim/total_tokens": 1772617728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440105, "loss/out": 6.0944623947143555, "created_at": "2025-01-14T22:32:01.783857+00:00"} {"global_step": 6763, "acc_step": 0, "speed/wps": 12964.357798019915, "speed/FLOPS": 203623054112384.78, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.295192718505859, "optim/lr": 0.0029924093253808114, "optim/total_tokens": 1772879872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 6.184113502502441, "created_at": "2025-01-14T22:32:06.842580+00:00"} {"global_step": 6764, "acc_step": 0, "speed/wps": 12970.822048558512, "speed/FLOPS": 203724584049904.53, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.072954177856445, "optim/lr": 0.002992400714244619, "optim/total_tokens": 1773142016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443022, "loss/out": 5.983604907989502, "created_at": "2025-01-14T22:32:11.901606+00:00"} {"global_step": 6765, "acc_step": 0, "speed/wps": 12972.636237813726, "speed/FLOPS": 203753078385114.94, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.140215873718262, "optim/lr": 0.0029923920982392067, "optim/total_tokens": 1773404160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336478, "loss/out": 6.109611511230469, "created_at": "2025-01-14T22:32:16.961787+00:00"} {"global_step": 6766, "acc_step": 0, "speed/wps": 12969.222199612374, "speed/FLOPS": 203699456223782.6, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 13.232406616210938, "optim/lr": 0.0029923834773646026, "optim/total_tokens": 1773666304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366558, "loss/out": 6.1022844314575195, "created_at": "2025-01-14T22:32:22.019764+00:00"} {"global_step": 6767, "acc_step": 0, "speed/wps": 12966.195033716944, "speed/FLOPS": 203651910423633.0, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 34.500770568847656, "optim/lr": 0.002992374851620835, "optim/total_tokens": 1773928448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444067, "loss/out": 6.1551666259765625, "created_at": "2025-01-14T22:32:27.081948+00:00"} {"global_step": 6768, "acc_step": 0, "speed/wps": 12966.483440456308, "speed/FLOPS": 203656440247787.12, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.331214904785156, "optim/lr": 0.0029923662210079322, "optim/total_tokens": 1774190592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 6.136930465698242, "created_at": "2025-01-14T22:32:32.140268+00:00"} {"global_step": 6769, "acc_step": 0, "speed/wps": 12971.8661376244, "speed/FLOPS": 203740982903413.03, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.823062896728516, "optim/lr": 0.0029923575855259218, "optim/total_tokens": 1774452736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 6.0951128005981445, "created_at": "2025-01-14T22:32:37.198601+00:00"} {"global_step": 6770, "acc_step": 0, "speed/wps": 12971.504023895954, "speed/FLOPS": 203735295409711.38, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.500834465026855, "optim/lr": 0.0029923489451748326, "optim/total_tokens": 1774714880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445166, "loss/out": 6.117920875549316, "created_at": "2025-01-14T22:32:42.254505+00:00"} {"global_step": 6771, "acc_step": 0, "speed/wps": 12975.479312113685, "speed/FLOPS": 203797732773789.28, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0006, "optim/grad_norm": 116.12968444824219, "optim/lr": 0.002992340299954693, "optim/total_tokens": 1774977024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447269, "loss/out": 6.081783294677734, "created_at": "2025-01-14T22:32:47.314740+00:00"} {"global_step": 6772, "acc_step": 0, "speed/wps": 12966.383338124786, "speed/FLOPS": 203654868003115.56, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.76493501663208, "optim/lr": 0.00299233164986553, "optim/total_tokens": 1775239168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419902, "loss/out": 6.059991836547852, "created_at": "2025-01-14T22:32:52.374708+00:00"} {"global_step": 6773, "acc_step": 0, "speed/wps": 12967.810613202912, "speed/FLOPS": 203677285319499.0, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 27.061494827270508, "optim/lr": 0.002992322994907373, "optim/total_tokens": 1775501312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455769, "loss/out": 6.120255947113037, "created_at": "2025-01-14T22:32:57.437702+00:00"} {"global_step": 6774, "acc_step": 0, "speed/wps": 12966.031237341524, "speed/FLOPS": 203649337776477.22, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.182828426361084, "optim/lr": 0.0029923143350802494, "optim/total_tokens": 1775763456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 6.032567977905273, "created_at": "2025-01-14T22:33:02.495906+00:00"} {"global_step": 6775, "acc_step": 0, "speed/wps": 12964.575691074679, "speed/FLOPS": 203626476422226.25, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.875205039978027, "optim/lr": 0.0029923056703841876, "optim/total_tokens": 1776025600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430791, "loss/out": 6.058568477630615, "created_at": "2025-01-14T22:33:07.565999+00:00"} {"global_step": 6776, "acc_step": 0, "speed/wps": 12971.56383318699, "speed/FLOPS": 203736234796813.7, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.703136444091797, "optim/lr": 0.0029922970008192163, "optim/total_tokens": 1776287744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 343977, "loss/out": 6.108346939086914, "created_at": "2025-01-14T22:33:12.622633+00:00"} {"global_step": 6777, "acc_step": 0, "speed/wps": 12959.188208627416, "speed/FLOPS": 203541858607215.34, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0019, "optim/grad_norm": 9.144647598266602, "optim/lr": 0.002992288326385364, "optim/total_tokens": 1776549888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452847, "loss/out": 5.989692687988281, "created_at": "2025-01-14T22:33:17.684450+00:00"} {"global_step": 6778, "acc_step": 0, "speed/wps": 12974.185771435132, "speed/FLOPS": 203777415939921.44, "speed/curr_iter_time": 1.2722, "speed/data_load_time": 0.0007, "optim/grad_norm": 162.9126434326172, "optim/lr": 0.0029922796470826583, "optim/total_tokens": 1776812032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456330, "loss/out": 6.071618556976318, "created_at": "2025-01-14T22:33:22.739859+00:00"} {"global_step": 6779, "acc_step": 0, "speed/wps": 12973.677813257635, "speed/FLOPS": 203769437758736.2, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.562468528747559, "optim/lr": 0.002992270962911128, "optim/total_tokens": 1777074176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 6.193623065948486, "created_at": "2025-01-14T22:33:27.799824+00:00"} {"global_step": 6780, "acc_step": 0, "speed/wps": 12972.29546264223, "speed/FLOPS": 203747726042772.44, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 32.53439712524414, "optim/lr": 0.0029922622738708013, "optim/total_tokens": 1777336320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421165, "loss/out": 6.123588562011719, "created_at": "2025-01-14T22:33:32.861137+00:00"} {"global_step": 6781, "acc_step": 0, "speed/wps": 12976.42476233217, "speed/FLOPS": 203812582368655.34, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.957388877868652, "optim/lr": 0.0029922535799617065, "optim/total_tokens": 1777598464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495289, "loss/out": 6.033451080322266, "created_at": "2025-01-14T22:33:37.916415+00:00"} {"global_step": 6782, "acc_step": 0, "speed/wps": 12966.56946896164, "speed/FLOPS": 203657791443678.9, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 106.3086166381836, "optim/lr": 0.0029922448811838718, "optim/total_tokens": 1777860608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450990, "loss/out": 6.02096700668335, "created_at": "2025-01-14T22:33:42.978362+00:00"} {"global_step": 6783, "acc_step": 0, "speed/wps": 12970.56562609705, "speed/FLOPS": 203720556582785.88, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 15.877230644226074, "optim/lr": 0.0029922361775373257, "optim/total_tokens": 1778122752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322689, "loss/out": 6.090091228485107, "created_at": "2025-01-14T22:33:48.040421+00:00"} {"global_step": 6784, "acc_step": 0, "speed/wps": 12966.459101042756, "speed/FLOPS": 203656057963851.4, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.220505714416504, "optim/lr": 0.0029922274690220964, "optim/total_tokens": 1778384896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375475, "loss/out": 6.033409118652344, "created_at": "2025-01-14T22:33:53.098382+00:00"} {"global_step": 6785, "acc_step": 0, "speed/wps": 12968.738875122808, "speed/FLOPS": 203691864948516.53, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 37.36701583862305, "optim/lr": 0.0029922187556382135, "optim/total_tokens": 1778647040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359599, "loss/out": 6.058555603027344, "created_at": "2025-01-14T22:33:58.158762+00:00"} {"global_step": 6786, "acc_step": 0, "speed/wps": 12972.358514838925, "speed/FLOPS": 203748716364164.3, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 65.3088150024414, "optim/lr": 0.002992210037385704, "optim/total_tokens": 1778909184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367183, "loss/out": 6.111544132232666, "created_at": "2025-01-14T22:34:03.220345+00:00"} {"global_step": 6787, "acc_step": 0, "speed/wps": 12972.09418565086, "speed/FLOPS": 203744564711039.4, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 15.570939064025879, "optim/lr": 0.002992201314264597, "optim/total_tokens": 1779171328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421768, "loss/out": 6.1167073249816895, "created_at": "2025-01-14T22:34:08.280804+00:00"} {"global_step": 6788, "acc_step": 0, "speed/wps": 12974.034716483015, "speed/FLOPS": 203775043414327.2, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0013, "optim/grad_norm": 5.704115390777588, "optim/lr": 0.00299219258627492, "optim/total_tokens": 1779433472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 6.083859443664551, "created_at": "2025-01-14T22:34:13.359234+00:00"} {"global_step": 6789, "acc_step": 0, "speed/wps": 12970.531392733428, "speed/FLOPS": 203720018900768.62, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.538262844085693, "optim/lr": 0.0029921838534167026, "optim/total_tokens": 1779695616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495052, "loss/out": 5.964956760406494, "created_at": "2025-01-14T22:34:18.417848+00:00"} {"global_step": 6790, "acc_step": 0, "speed/wps": 12966.553453514634, "speed/FLOPS": 203657539899076.16, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.054966449737549, "optim/lr": 0.002992175115689973, "optim/total_tokens": 1779957760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 6.115819931030273, "created_at": "2025-01-14T22:34:23.475844+00:00"} {"global_step": 6791, "acc_step": 0, "speed/wps": 12974.359347362075, "speed/FLOPS": 203780142188372.4, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.914186477661133, "optim/lr": 0.00299216637309476, "optim/total_tokens": 1780219904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428932, "loss/out": 6.06770133972168, "created_at": "2025-01-14T22:34:28.538678+00:00"} {"global_step": 6792, "acc_step": 0, "speed/wps": 12969.575568530514, "speed/FLOPS": 203705006368221.75, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.1957831382751465, "optim/lr": 0.0029921576256310917, "optim/total_tokens": 1780482048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305392, "loss/out": 6.098445892333984, "created_at": "2025-01-14T22:34:33.599069+00:00"} {"global_step": 6793, "acc_step": 0, "speed/wps": 12975.642856795641, "speed/FLOPS": 203800301467753.34, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.584377765655518, "optim/lr": 0.0029921488732989965, "optim/total_tokens": 1780744192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416384, "loss/out": 6.024827003479004, "created_at": "2025-01-14T22:34:38.653663+00:00"} {"global_step": 6794, "acc_step": 0, "speed/wps": 12974.239474437525, "speed/FLOPS": 203778259419369.7, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 46.988677978515625, "optim/lr": 0.0029921401160985035, "optim/total_tokens": 1781006336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428688, "loss/out": 5.9842729568481445, "created_at": "2025-01-14T22:34:43.714207+00:00"} {"global_step": 6795, "acc_step": 0, "speed/wps": 12971.149705867756, "speed/FLOPS": 203729730358194.53, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0009, "optim/grad_norm": 10.35348129272461, "optim/lr": 0.0029921313540296405, "optim/total_tokens": 1781268480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 6.032902717590332, "created_at": "2025-01-14T22:34:48.770168+00:00"} {"global_step": 6796, "acc_step": 0, "speed/wps": 12966.86377507542, "speed/FLOPS": 203662413925614.8, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.429095268249512, "optim/lr": 0.002992122587092437, "optim/total_tokens": 1781530624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445976, "loss/out": 6.037248134613037, "created_at": "2025-01-14T22:34:53.830064+00:00"} {"global_step": 6797, "acc_step": 0, "speed/wps": 12966.332027229782, "speed/FLOPS": 203654062095000.97, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.066164493560791, "optim/lr": 0.002992113815286921, "optim/total_tokens": 1781792768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435567, "loss/out": 5.985104560852051, "created_at": "2025-01-14T22:34:58.889044+00:00"} {"global_step": 6798, "acc_step": 0, "speed/wps": 12965.494237686891, "speed/FLOPS": 203640903458986.38, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 48.17171096801758, "optim/lr": 0.0029921050386131215, "optim/total_tokens": 1782054912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 5.972278118133545, "created_at": "2025-01-14T22:35:03.948018+00:00"} {"global_step": 6799, "acc_step": 0, "speed/wps": 12963.071481367933, "speed/FLOPS": 203602850741781.1, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 25.151968002319336, "optim/lr": 0.0029920962570710668, "optim/total_tokens": 1782317056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 6.110867023468018, "created_at": "2025-01-14T22:35:09.007549+00:00"} {"global_step": 6800, "acc_step": 0, "speed/wps": 12965.628435651415, "speed/FLOPS": 203643011222426.56, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.20365047454834, "optim/lr": 0.0029920874706607856, "optim/total_tokens": 1782579200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 6.100152492523193, "created_at": "2025-01-14T22:35:14.065708+00:00"} {"global_step": 6801, "acc_step": 0, "speed/wps": 12965.277572664842, "speed/FLOPS": 203637500437086.16, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0009, "optim/grad_norm": 7.8336334228515625, "optim/lr": 0.002992078679382306, "optim/total_tokens": 1782841344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445407, "loss/out": 6.007120609283447, "created_at": "2025-01-14T22:35:19.124062+00:00"} {"global_step": 6802, "acc_step": 0, "speed/wps": 12970.177225853315, "speed/FLOPS": 203714456223240.16, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 27.392126083374023, "optim/lr": 0.0029920698832356578, "optim/total_tokens": 1783103488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325819, "loss/out": 6.050208568572998, "created_at": "2025-01-14T22:35:24.181686+00:00"} {"global_step": 6803, "acc_step": 0, "speed/wps": 12969.349404181003, "speed/FLOPS": 203701454146330.3, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 27.297426223754883, "optim/lr": 0.0029920610822208697, "optim/total_tokens": 1783365632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353457, "loss/out": 6.202848434448242, "created_at": "2025-01-14T22:35:29.240034+00:00"} {"global_step": 6804, "acc_step": 0, "speed/wps": 12968.585456553254, "speed/FLOPS": 203689455299062.47, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 25.609037399291992, "optim/lr": 0.0029920522763379693, "optim/total_tokens": 1783627776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438769, "loss/out": 6.187745094299316, "created_at": "2025-01-14T22:35:34.301344+00:00"} {"global_step": 6805, "acc_step": 0, "speed/wps": 12966.041006378746, "speed/FLOPS": 203649491212630.88, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 11.969744682312012, "optim/lr": 0.0029920434655869863, "optim/total_tokens": 1783889920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 6.298915863037109, "created_at": "2025-01-14T22:35:39.359400+00:00"} {"global_step": 6806, "acc_step": 0, "speed/wps": 12963.262794788045, "speed/FLOPS": 203605855581936.12, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.014822006225586, "optim/lr": 0.002992034649967949, "optim/total_tokens": 1784152064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 6.239914417266846, "created_at": "2025-01-14T22:35:44.423985+00:00"} {"global_step": 6807, "acc_step": 0, "speed/wps": 12966.763762446202, "speed/FLOPS": 203660843089841.2, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 17.324726104736328, "optim/lr": 0.0029920258294808865, "optim/total_tokens": 1784414208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468255, "loss/out": 6.113152027130127, "created_at": "2025-01-14T22:35:49.486791+00:00"} {"global_step": 6808, "acc_step": 0, "speed/wps": 12965.5945914631, "speed/FLOPS": 203642479652942.25, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 26.337072372436523, "optim/lr": 0.002992017004125827, "optim/total_tokens": 1784676352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449420, "loss/out": 6.257666110992432, "created_at": "2025-01-14T22:35:54.544950+00:00"} {"global_step": 6809, "acc_step": 0, "speed/wps": 12966.27044322372, "speed/FLOPS": 203653094833560.47, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.751944065093994, "optim/lr": 0.0029920081739027997, "optim/total_tokens": 1784938496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323417, "loss/out": 6.246840476989746, "created_at": "2025-01-14T22:35:59.602799+00:00"} {"global_step": 6810, "acc_step": 0, "speed/wps": 12968.026385451398, "speed/FLOPS": 203680674319163.62, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.935049533843994, "optim/lr": 0.002991999338811833, "optim/total_tokens": 1785200640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 343445, "loss/out": 6.049423694610596, "created_at": "2025-01-14T22:36:04.660588+00:00"} {"global_step": 6811, "acc_step": 0, "speed/wps": 12972.86861065859, "speed/FLOPS": 203756728119957.28, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.088361740112305, "optim/lr": 0.0029919904988529567, "optim/total_tokens": 1785462784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361759, "loss/out": 6.074228763580322, "created_at": "2025-01-14T22:36:09.720325+00:00"} {"global_step": 6812, "acc_step": 0, "speed/wps": 12966.976235871438, "speed/FLOPS": 203664180276953.6, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.823686599731445, "optim/lr": 0.002991981654026199, "optim/total_tokens": 1785724928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 392536, "loss/out": 6.066181182861328, "created_at": "2025-01-14T22:36:14.778549+00:00"} {"global_step": 6813, "acc_step": 0, "speed/wps": 12978.656590837838, "speed/FLOPS": 203847636302190.6, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.375176429748535, "optim/lr": 0.002991972804331588, "optim/total_tokens": 1785987072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442256, "loss/out": 6.196681976318359, "created_at": "2025-01-14T22:36:19.839914+00:00"} {"global_step": 6814, "acc_step": 0, "speed/wps": 12971.106158974928, "speed/FLOPS": 203729046394402.97, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.935683250427246, "optim/lr": 0.002991963949769154, "optim/total_tokens": 1786249216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458734, "loss/out": 6.018583297729492, "created_at": "2025-01-14T22:36:24.898113+00:00"} {"global_step": 6815, "acc_step": 0, "speed/wps": 12969.515158463515, "speed/FLOPS": 203704057545107.4, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.6958119869232178, "optim/lr": 0.0029919550903389253, "optim/total_tokens": 1786511360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 6.096257209777832, "created_at": "2025-01-14T22:36:29.957455+00:00"} {"global_step": 6816, "acc_step": 0, "speed/wps": 12970.065971757058, "speed/FLOPS": 203712708824779.28, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.447419166564941, "optim/lr": 0.0029919462260409304, "optim/total_tokens": 1786773504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431605, "loss/out": 6.012577533721924, "created_at": "2025-01-14T22:36:35.014249+00:00"} {"global_step": 6817, "acc_step": 0, "speed/wps": 12963.083865596487, "speed/FLOPS": 203603045253108.25, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.264955520629883, "optim/lr": 0.0029919373568751987, "optim/total_tokens": 1787035648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474937, "loss/out": 6.206027030944824, "created_at": "2025-01-14T22:36:40.073791+00:00"} {"global_step": 6818, "acc_step": 0, "speed/wps": 12966.11870049306, "speed/FLOPS": 203650711505459.06, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.8883867263793945, "optim/lr": 0.0029919284828417592, "optim/total_tokens": 1787297792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 6.115121841430664, "created_at": "2025-01-14T22:36:45.134080+00:00"} {"global_step": 6819, "acc_step": 0, "speed/wps": 12969.421508673573, "speed/FLOPS": 203702586646468.1, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 22.85791778564453, "optim/lr": 0.0029919196039406406, "optim/total_tokens": 1787559936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 340231, "loss/out": 6.119697570800781, "created_at": "2025-01-14T22:36:50.190956+00:00"} {"global_step": 6820, "acc_step": 0, "speed/wps": 12964.478104653672, "speed/FLOPS": 203624943693386.56, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.4309568405151367, "optim/lr": 0.0029919107201718718, "optim/total_tokens": 1787822080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361473, "loss/out": 6.206136226654053, "created_at": "2025-01-14T22:36:55.249575+00:00"} {"global_step": 6821, "acc_step": 0, "speed/wps": 12969.680716185523, "speed/FLOPS": 203706657856631.7, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.389335632324219, "optim/lr": 0.002991901831535482, "optim/total_tokens": 1788084224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448890, "loss/out": 6.074042797088623, "created_at": "2025-01-14T22:37:00.307273+00:00"} {"global_step": 6822, "acc_step": 0, "speed/wps": 12973.341539730196, "speed/FLOPS": 203764156120899.5, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.719689846038818, "optim/lr": 0.0029918929380315, "optim/total_tokens": 1788346368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463770, "loss/out": 6.176101207733154, "created_at": "2025-01-14T22:37:05.362316+00:00"} {"global_step": 6823, "acc_step": 0, "speed/wps": 12968.617722191833, "speed/FLOPS": 203689962075254.0, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 17.91705322265625, "optim/lr": 0.0029918840396599552, "optim/total_tokens": 1788608512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 6.198704719543457, "created_at": "2025-01-14T22:37:10.419207+00:00"} {"global_step": 6824, "acc_step": 0, "speed/wps": 12968.055668169362, "speed/FLOPS": 203681134244487.9, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.463741302490234, "optim/lr": 0.0029918751364208756, "optim/total_tokens": 1788870656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465083, "loss/out": 6.367733001708984, "created_at": "2025-01-14T22:37:15.481916+00:00"} {"global_step": 6825, "acc_step": 0, "speed/wps": 12965.87071331143, "speed/FLOPS": 203646816526001.47, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.445817708969116, "optim/lr": 0.002991866228314292, "optim/total_tokens": 1789132800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445691, "loss/out": 6.37263822555542, "created_at": "2025-01-14T22:37:20.547363+00:00"} {"global_step": 6826, "acc_step": 0, "speed/wps": 12975.343797626005, "speed/FLOPS": 203795604332543.84, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.624778985977173, "optim/lr": 0.002991857315340232, "optim/total_tokens": 1789394944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 6.237281322479248, "created_at": "2025-01-14T22:37:25.602923+00:00"} {"global_step": 6827, "acc_step": 0, "speed/wps": 12970.19546159011, "speed/FLOPS": 203714742640545.06, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 8.447280883789062, "optim/lr": 0.002991848397498725, "optim/total_tokens": 1789657088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 6.195890426635742, "created_at": "2025-01-14T22:37:30.659386+00:00"} {"global_step": 6828, "acc_step": 0, "speed/wps": 12968.445643031315, "speed/FLOPS": 203687259335576.5, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0008, "optim/grad_norm": 21.353960037231445, "optim/lr": 0.0029918394747898004, "optim/total_tokens": 1789919232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446218, "loss/out": 6.39706563949585, "created_at": "2025-01-14T22:37:35.721400+00:00"} {"global_step": 6829, "acc_step": 0, "speed/wps": 12972.857623564047, "speed/FLOPS": 203756555552539.62, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 103.3343505859375, "optim/lr": 0.002991830547213487, "optim/total_tokens": 1790181376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309845, "loss/out": 6.253693580627441, "created_at": "2025-01-14T22:37:40.776713+00:00"} {"global_step": 6830, "acc_step": 0, "speed/wps": 12976.687210032851, "speed/FLOPS": 203816704470435.47, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.4954447746276855, "optim/lr": 0.0029918216147698145, "optim/total_tokens": 1790443520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352115, "loss/out": 6.408364295959473, "created_at": "2025-01-14T22:37:45.833482+00:00"} {"global_step": 6831, "acc_step": 0, "speed/wps": 12969.356914580052, "speed/FLOPS": 203701572107467.75, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0006, "optim/grad_norm": 46.45949172973633, "optim/lr": 0.002991812677458811, "optim/total_tokens": 1790705664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430283, "loss/out": 6.283332824707031, "created_at": "2025-01-14T22:37:50.895667+00:00"} {"global_step": 6832, "acc_step": 0, "speed/wps": 12971.79629674027, "speed/FLOPS": 203739885956356.3, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.226996421813965, "optim/lr": 0.002991803735280507, "optim/total_tokens": 1790967808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471976, "loss/out": 6.437471866607666, "created_at": "2025-01-14T22:37:55.957847+00:00"} {"global_step": 6833, "acc_step": 0, "speed/wps": 12975.994929460847, "speed/FLOPS": 203805831252758.94, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.521575927734375, "optim/lr": 0.0029917947882349303, "optim/total_tokens": 1791229952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442208, "loss/out": 6.4950432777404785, "created_at": "2025-01-14T22:38:01.013887+00:00"} {"global_step": 6834, "acc_step": 0, "speed/wps": 12968.651120881943, "speed/FLOPS": 203690486647576.78, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0006, "optim/grad_norm": 40.48609161376953, "optim/lr": 0.0029917858363221112, "optim/total_tokens": 1791492096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 6.236504554748535, "created_at": "2025-01-14T22:38:06.075177+00:00"} {"global_step": 6835, "acc_step": 0, "speed/wps": 12973.661550622692, "speed/FLOPS": 203769182331707.2, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 32.61623001098633, "optim/lr": 0.0029917768795420787, "optim/total_tokens": 1791754240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 6.254572868347168, "created_at": "2025-01-14T22:38:11.134188+00:00"} {"global_step": 6836, "acc_step": 0, "speed/wps": 12964.310881211459, "speed/FLOPS": 203622317219437.5, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 11.291411399841309, "optim/lr": 0.002991767917894861, "optim/total_tokens": 1792016384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323942, "loss/out": 6.585858345031738, "created_at": "2025-01-14T22:38:16.200648+00:00"} {"global_step": 6837, "acc_step": 0, "speed/wps": 12971.285897894688, "speed/FLOPS": 203731869441124.94, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 19.285999298095703, "optim/lr": 0.0029917589513804885, "optim/total_tokens": 1792278528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 340756, "loss/out": 6.421093940734863, "created_at": "2025-01-14T22:38:21.257941+00:00"} {"global_step": 6838, "acc_step": 0, "speed/wps": 12967.581714487356, "speed/FLOPS": 203673690150631.4, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.1122963428497314, "optim/lr": 0.0029917499799989905, "optim/total_tokens": 1792540672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357199, "loss/out": 6.348624229431152, "created_at": "2025-01-14T22:38:26.318429+00:00"} {"global_step": 6839, "acc_step": 0, "speed/wps": 12968.82724720492, "speed/FLOPS": 203693252953502.03, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.808637857437134, "optim/lr": 0.0029917410037503954, "optim/total_tokens": 1792802816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459237, "loss/out": 6.358457565307617, "created_at": "2025-01-14T22:38:31.382329+00:00"} {"global_step": 6840, "acc_step": 0, "speed/wps": 12974.842239206566, "speed/FLOPS": 203787726668353.06, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.518462181091309, "optim/lr": 0.002991732022634733, "optim/total_tokens": 1793064960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449942, "loss/out": 6.089352607727051, "created_at": "2025-01-14T22:38:36.440013+00:00"} {"global_step": 6841, "acc_step": 0, "speed/wps": 12967.545312725082, "speed/FLOPS": 203673118410933.62, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.074029922485352, "optim/lr": 0.002991723036652033, "optim/total_tokens": 1793327104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 6.183382034301758, "created_at": "2025-01-14T22:38:41.501101+00:00"} {"global_step": 6842, "acc_step": 0, "speed/wps": 12970.582600595359, "speed/FLOPS": 203720823190607.22, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.637929916381836, "optim/lr": 0.0029917140458023237, "optim/total_tokens": 1793589248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411117, "loss/out": 6.261198997497559, "created_at": "2025-01-14T22:38:46.562889+00:00"} {"global_step": 6843, "acc_step": 0, "speed/wps": 12968.534780704405, "speed/FLOPS": 203688659365220.38, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.022025108337402, "optim/lr": 0.0029917050500856347, "optim/total_tokens": 1793851392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 6.249085426330566, "created_at": "2025-01-14T22:38:51.619838+00:00"} {"global_step": 6844, "acc_step": 0, "speed/wps": 12968.900294784033, "speed/FLOPS": 203694400266109.9, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.0015034675598145, "optim/lr": 0.002991696049501996, "optim/total_tokens": 1794113536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444641, "loss/out": 6.218022346496582, "created_at": "2025-01-14T22:38:56.677146+00:00"} {"global_step": 6845, "acc_step": 0, "speed/wps": 12968.164524182457, "speed/FLOPS": 203682843977757.4, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.713649272918701, "optim/lr": 0.002991687044051437, "optim/total_tokens": 1794375680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448606, "loss/out": 6.147371292114258, "created_at": "2025-01-14T22:39:01.739028+00:00"} {"global_step": 6846, "acc_step": 0, "speed/wps": 12965.611323614023, "speed/FLOPS": 203642742454364.9, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.7240583896636963, "optim/lr": 0.0029916780337339863, "optim/total_tokens": 1794637824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359599, "loss/out": 6.054603576660156, "created_at": "2025-01-14T22:39:06.802908+00:00"} {"global_step": 6847, "acc_step": 0, "speed/wps": 12966.161186570584, "speed/FLOPS": 203651378807688.56, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.094419956207275, "optim/lr": 0.0029916690185496737, "optim/total_tokens": 1794899968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455532, "loss/out": 6.243198394775391, "created_at": "2025-01-14T22:39:11.866044+00:00"} {"global_step": 6848, "acc_step": 0, "speed/wps": 12961.211095057839, "speed/FLOPS": 203573630818342.38, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.001, "optim/grad_norm": 2.1865451335906982, "optim/lr": 0.002991659998498529, "optim/total_tokens": 1795162112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 6.137396335601807, "created_at": "2025-01-14T22:39:16.931909+00:00"} {"global_step": 6849, "acc_step": 0, "speed/wps": 12969.799319202439, "speed/FLOPS": 203708520679990.0, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.6707794666290283, "optim/lr": 0.002991650973580581, "optim/total_tokens": 1795424256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443545, "loss/out": 6.1703267097473145, "created_at": "2025-01-14T22:39:21.993985+00:00"} {"global_step": 6850, "acc_step": 0, "speed/wps": 12972.896728999893, "speed/FLOPS": 203757169757146.1, "speed/curr_iter_time": 1.2722, "speed/data_load_time": 0.0007, "optim/grad_norm": 32.21332931518555, "optim/lr": 0.002991641943795859, "optim/total_tokens": 1795686400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 6.002740383148193, "created_at": "2025-01-14T22:39:27.052201+00:00"} {"global_step": 6851, "acc_step": 0, "speed/wps": 12965.512070865605, "speed/FLOPS": 203641183553563.44, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.0856611728668213, "optim/lr": 0.0029916329091443927, "optim/total_tokens": 1795948544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456851, "loss/out": 6.000856876373291, "created_at": "2025-01-14T22:39:32.113470+00:00"} {"global_step": 6852, "acc_step": 0, "speed/wps": 12967.550558026482, "speed/FLOPS": 203673200795599.9, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 59.300750732421875, "optim/lr": 0.0029916238696262123, "optim/total_tokens": 1796210688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 339300, "loss/out": 6.05742073059082, "created_at": "2025-01-14T22:39:37.183880+00:00"} {"global_step": 6853, "acc_step": 0, "speed/wps": 12970.91489849693, "speed/FLOPS": 203726042385776.56, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.271095275878906, "optim/lr": 0.0029916148252413467, "optim/total_tokens": 1796472832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293306, "loss/out": 6.026239395141602, "created_at": "2025-01-14T22:39:42.239978+00:00"} {"global_step": 6854, "acc_step": 0, "speed/wps": 12965.612503753082, "speed/FLOPS": 203642760990070.5, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.022876739501953, "optim/lr": 0.002991605775989825, "optim/total_tokens": 1796734976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401411, "loss/out": 6.24753999710083, "created_at": "2025-01-14T22:39:47.298995+00:00"} {"global_step": 6855, "acc_step": 0, "speed/wps": 12977.994021336272, "speed/FLOPS": 203837229737702.6, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.616840362548828, "optim/lr": 0.0029915967218716775, "optim/total_tokens": 1796997120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436151, "loss/out": 6.196944236755371, "created_at": "2025-01-14T22:39:52.356911+00:00"} {"global_step": 6856, "acc_step": 0, "speed/wps": 12970.475917806725, "speed/FLOPS": 203719147590814.22, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.429445266723633, "optim/lr": 0.0029915876628869337, "optim/total_tokens": 1797259264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 6.371617317199707, "created_at": "2025-01-14T22:39:57.420017+00:00"} {"global_step": 6857, "acc_step": 0, "speed/wps": 12966.841362478333, "speed/FLOPS": 203662061904979.44, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.504612922668457, "optim/lr": 0.002991578599035622, "optim/total_tokens": 1797521408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444644, "loss/out": 6.187125205993652, "created_at": "2025-01-14T22:40:02.481074+00:00"} {"global_step": 6858, "acc_step": 0, "speed/wps": 12967.192365068051, "speed/FLOPS": 203667574882979.03, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.1539571285247803, "optim/lr": 0.0029915695303177735, "optim/total_tokens": 1797783552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 6.162513732910156, "created_at": "2025-01-14T22:40:07.545941+00:00"} {"global_step": 6859, "acc_step": 0, "speed/wps": 12971.061660774672, "speed/FLOPS": 203728347489021.0, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.085663318634033, "optim/lr": 0.002991560456733417, "optim/total_tokens": 1798045696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436381, "loss/out": 6.124019622802734, "created_at": "2025-01-14T22:40:12.602175+00:00"} {"global_step": 6860, "acc_step": 0, "speed/wps": 12966.744508931766, "speed/FLOPS": 203660540686939.88, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.374499559402466, "optim/lr": 0.0029915513782825818, "optim/total_tokens": 1798307840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303511, "loss/out": 6.143689155578613, "created_at": "2025-01-14T22:40:17.665225+00:00"} {"global_step": 6861, "acc_step": 0, "speed/wps": 12972.198192809694, "speed/FLOPS": 203746198286389.7, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.445683479309082, "optim/lr": 0.002991542294965298, "optim/total_tokens": 1798569984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452859, "loss/out": 6.072608470916748, "created_at": "2025-01-14T22:40:22.721129+00:00"} {"global_step": 6862, "acc_step": 0, "speed/wps": 12972.651129917173, "speed/FLOPS": 203753312286063.47, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.7187612056732178, "optim/lr": 0.0029915332067815955, "optim/total_tokens": 1798832128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445166, "loss/out": 6.179553031921387, "created_at": "2025-01-14T22:40:27.780275+00:00"} {"global_step": 6863, "acc_step": 0, "speed/wps": 12961.86229427284, "speed/FLOPS": 203583858796855.44, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.542181968688965, "optim/lr": 0.002991524113731503, "optim/total_tokens": 1799094272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 6.062779903411865, "created_at": "2025-01-14T22:40:32.841029+00:00"} {"global_step": 6864, "acc_step": 0, "speed/wps": 12958.213493346073, "speed/FLOPS": 203526549364322.7, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.342212677001953, "optim/lr": 0.0029915150158150517, "optim/total_tokens": 1799356416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438531, "loss/out": 6.037642478942871, "created_at": "2025-01-14T22:40:37.906425+00:00"} {"global_step": 6865, "acc_step": 0, "speed/wps": 12971.673594582115, "speed/FLOPS": 203737958750352.06, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.919290065765381, "optim/lr": 0.0029915059130322702, "optim/total_tokens": 1799618560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440919, "loss/out": 6.147400856018066, "created_at": "2025-01-14T22:40:42.963617+00:00"} {"global_step": 6866, "acc_step": 0, "speed/wps": 12970.89480779501, "speed/FLOPS": 203725726833695.4, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.868842840194702, "optim/lr": 0.002991496805383188, "optim/total_tokens": 1799880704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453945, "loss/out": 6.150202751159668, "created_at": "2025-01-14T22:40:48.021325+00:00"} {"global_step": 6867, "acc_step": 0, "speed/wps": 12973.362838697136, "speed/FLOPS": 203764490650443.12, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.8712849617004395, "optim/lr": 0.002991487692867835, "optim/total_tokens": 1800142848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 6.107945919036865, "created_at": "2025-01-14T22:40:53.081080+00:00"} {"global_step": 6868, "acc_step": 0, "speed/wps": 12966.254275607718, "speed/FLOPS": 203652840898934.5, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 15.793168067932129, "optim/lr": 0.0029914785754862412, "optim/total_tokens": 1800404992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328221, "loss/out": 6.1713480949401855, "created_at": "2025-01-14T22:40:58.145246+00:00"} {"global_step": 6869, "acc_step": 0, "speed/wps": 12973.440417277478, "speed/FLOPS": 203765709128651.53, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.536731243133545, "optim/lr": 0.002991469453238437, "optim/total_tokens": 1800667136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353744, "loss/out": 6.085790634155273, "created_at": "2025-01-14T22:41:03.201440+00:00"} {"global_step": 6870, "acc_step": 0, "speed/wps": 12973.970172110412, "speed/FLOPS": 203774029656262.22, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.266366481781006, "optim/lr": 0.0029914603261244504, "optim/total_tokens": 1800929280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443016, "loss/out": 6.025653839111328, "created_at": "2025-01-14T22:41:08.262789+00:00"} {"global_step": 6871, "acc_step": 0, "speed/wps": 12972.374236133703, "speed/FLOPS": 203748963288702.12, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1091.0557861328125, "optim/lr": 0.0029914511941443126, "optim/total_tokens": 1801191424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462938, "loss/out": 6.125148296356201, "created_at": "2025-01-14T22:41:13.318879+00:00"} {"global_step": 6872, "acc_step": 0, "speed/wps": 12961.319468537753, "speed/FLOPS": 203575332972765.2, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 121.3364486694336, "optim/lr": 0.0029914420572980534, "optim/total_tokens": 1801453568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451559, "loss/out": 6.059523105621338, "created_at": "2025-01-14T22:41:18.383572+00:00"} {"global_step": 6873, "acc_step": 0, "speed/wps": 12972.373258035608, "speed/FLOPS": 203748947926327.53, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 9.791168212890625, "optim/lr": 0.002991432915585702, "optim/total_tokens": 1801715712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461359, "loss/out": 6.249405384063721, "created_at": "2025-01-14T22:41:23.447250+00:00"} {"global_step": 6874, "acc_step": 0, "speed/wps": 12980.307658711563, "speed/FLOPS": 203873568591952.47, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.844196319580078, "optim/lr": 0.002991423769007288, "optim/total_tokens": 1801977856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460556, "loss/out": 6.324063301086426, "created_at": "2025-01-14T22:41:28.500235+00:00"} {"global_step": 6875, "acc_step": 0, "speed/wps": 12979.643603861648, "speed/FLOPS": 203863138698027.56, "speed/curr_iter_time": 1.272, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.604907989501953, "optim/lr": 0.002991414617562842, "optim/total_tokens": 1802240000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426252, "loss/out": 6.131568908691406, "created_at": "2025-01-14T22:41:33.557308+00:00"} {"global_step": 6876, "acc_step": 0, "speed/wps": 12976.330749512112, "speed/FLOPS": 203811105768129.16, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0012, "optim/grad_norm": 10.088457107543945, "optim/lr": 0.002991405461252394, "optim/total_tokens": 1802502144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298409, "loss/out": 5.986175537109375, "created_at": "2025-01-14T22:41:38.616990+00:00"} {"global_step": 6877, "acc_step": 0, "speed/wps": 12976.961047670564, "speed/FLOPS": 203821005466828.1, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5947246551513672, "optim/lr": 0.002991396300075972, "optim/total_tokens": 1802764288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352926, "loss/out": 6.078225135803223, "created_at": "2025-01-14T22:41:43.675132+00:00"} {"global_step": 6878, "acc_step": 0, "speed/wps": 12973.942620807498, "speed/FLOPS": 203773596925190.5, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.84946060180664, "optim/lr": 0.002991387134033608, "optim/total_tokens": 1803026432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 394694, "loss/out": 6.0066046714782715, "created_at": "2025-01-14T22:41:48.731400+00:00"} {"global_step": 6879, "acc_step": 0, "speed/wps": 12975.450357138432, "speed/FLOPS": 203797277996114.62, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3929035663604736, "optim/lr": 0.0029913779631253316, "optim/total_tokens": 1803288576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439571, "loss/out": 6.107631683349609, "created_at": "2025-01-14T22:41:53.787014+00:00"} {"global_step": 6880, "acc_step": 0, "speed/wps": 12977.53679316504, "speed/FLOPS": 203830048340976.06, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.567361831665039, "optim/lr": 0.0029913687873511724, "optim/total_tokens": 1803550720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427351, "loss/out": 6.038535118103027, "created_at": "2025-01-14T22:41:58.844165+00:00"} {"global_step": 6881, "acc_step": 0, "speed/wps": 12975.45686972365, "speed/FLOPS": 203797380285214.72, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 172.19927978515625, "optim/lr": 0.0029913596067111603, "optim/total_tokens": 1803812864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 6.064759254455566, "created_at": "2025-01-14T22:42:03.905576+00:00"} {"global_step": 6882, "acc_step": 0, "speed/wps": 12978.529821209788, "speed/FLOPS": 203845645210982.34, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.830792427062988, "optim/lr": 0.0029913504212053244, "optim/total_tokens": 1804075008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 6.114002227783203, "created_at": "2025-01-14T22:42:08.960253+00:00"} {"global_step": 6883, "acc_step": 0, "speed/wps": 12972.01523621257, "speed/FLOPS": 203743324701623.16, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.429976463317871, "optim/lr": 0.0029913412308336957, "optim/total_tokens": 1804337152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 6.256884574890137, "created_at": "2025-01-14T22:42:14.016489+00:00"} {"global_step": 6884, "acc_step": 0, "speed/wps": 12977.397274801138, "speed/FLOPS": 203827857013352.66, "speed/curr_iter_time": 1.2722, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.201539039611816, "optim/lr": 0.0029913320355963047, "optim/total_tokens": 1804599296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 6.134826183319092, "created_at": "2025-01-14T22:42:19.071982+00:00"} {"global_step": 6885, "acc_step": 0, "speed/wps": 12972.418714642754, "speed/FLOPS": 203749661884806.66, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.004980564117432, "optim/lr": 0.0029913228354931805, "optim/total_tokens": 1804861440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432365, "loss/out": 6.109576225280762, "created_at": "2025-01-14T22:42:24.129020+00:00"} {"global_step": 6886, "acc_step": 0, "speed/wps": 12975.046149416374, "speed/FLOPS": 203790929358400.1, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9535306692123413, "optim/lr": 0.002991313630524353, "optim/total_tokens": 1805123584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305561, "loss/out": 6.140848159790039, "created_at": "2025-01-14T22:42:29.183448+00:00"} {"global_step": 6887, "acc_step": 0, "speed/wps": 12977.491327008976, "speed/FLOPS": 203829334232518.03, "speed/curr_iter_time": 1.272, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5467619895935059, "optim/lr": 0.002991304420689852, "optim/total_tokens": 1805385728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359074, "loss/out": 6.00916051864624, "created_at": "2025-01-14T22:42:34.238297+00:00"} {"global_step": 6888, "acc_step": 0, "speed/wps": 12980.161793302244, "speed/FLOPS": 203871277575259.03, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4678361415863037, "optim/lr": 0.002991295205989709, "optim/total_tokens": 1805647872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445406, "loss/out": 5.986370086669922, "created_at": "2025-01-14T22:42:39.295620+00:00"} {"global_step": 6889, "acc_step": 0, "speed/wps": 12978.136180663216, "speed/FLOPS": 203839462545279.66, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.396101951599121, "optim/lr": 0.0029912859864239523, "optim/total_tokens": 1805910016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444881, "loss/out": 5.9970221519470215, "created_at": "2025-01-14T22:42:44.355182+00:00"} {"global_step": 6890, "acc_step": 0, "speed/wps": 12971.44764159675, "speed/FLOPS": 203734409848224.97, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.079930782318115, "optim/lr": 0.0029912767619926133, "optim/total_tokens": 1806172160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 6.075756072998047, "created_at": "2025-01-14T22:42:49.413322+00:00"} {"global_step": 6891, "acc_step": 0, "speed/wps": 12979.500204899648, "speed/FLOPS": 203860886420278.53, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.448596477508545, "optim/lr": 0.0029912675326957214, "optim/total_tokens": 1806434304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 6.090580940246582, "created_at": "2025-01-14T22:42:54.467179+00:00"} {"global_step": 6892, "acc_step": 0, "speed/wps": 12971.589581025864, "speed/FLOPS": 203736639202004.44, "speed/curr_iter_time": 1.2721, "speed/data_load_time": 0.001, "optim/grad_norm": 25.98946762084961, "optim/lr": 0.0029912582985333076, "optim/total_tokens": 1806696448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 6.099169731140137, "created_at": "2025-01-14T22:42:59.523460+00:00"} {"global_step": 6893, "acc_step": 0, "speed/wps": 12979.98438213177, "speed/FLOPS": 203868491089038.25, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 22.23243522644043, "optim/lr": 0.0029912490595054007, "optim/total_tokens": 1806958592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452596, "loss/out": 6.116554260253906, "created_at": "2025-01-14T22:43:04.577480+00:00"} {"global_step": 6894, "acc_step": 0, "speed/wps": 12968.232808370718, "speed/FLOPS": 203683916474766.34, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.277071952819824, "optim/lr": 0.0029912398156120316, "optim/total_tokens": 1807220736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440628, "loss/out": 6.098761558532715, "created_at": "2025-01-14T22:43:09.665075+00:00"} {"global_step": 6895, "acc_step": 0, "speed/wps": 12975.324937315325, "speed/FLOPS": 203795308105447.84, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.001, "optim/grad_norm": 3.270962715148926, "optim/lr": 0.0029912305668532303, "optim/total_tokens": 1807482880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301284, "loss/out": 6.135784149169922, "created_at": "2025-01-14T22:43:14.724381+00:00"} {"global_step": 6896, "acc_step": 0, "speed/wps": 12975.009165050127, "speed/FLOPS": 203790348468106.34, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.351808547973633, "optim/lr": 0.002991221313229027, "optim/total_tokens": 1807745024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 387743, "loss/out": 6.101472854614258, "created_at": "2025-01-14T22:43:19.779802+00:00"} {"global_step": 6897, "acc_step": 0, "speed/wps": 12975.156974587111, "speed/FLOPS": 203792670019996.28, "speed/curr_iter_time": 1.2722, "speed/data_load_time": 0.0006, "optim/grad_norm": 11.942527770996094, "optim/lr": 0.002991212054739452, "optim/total_tokens": 1808007168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450990, "loss/out": 6.104109764099121, "created_at": "2025-01-14T22:43:24.837945+00:00"} {"global_step": 6898, "acc_step": 0, "speed/wps": 12977.534067149232, "speed/FLOPS": 203830005525151.88, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.096639156341553, "optim/lr": 0.0029912027913845357, "optim/total_tokens": 1808269312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 6.12876033782959, "created_at": "2025-01-14T22:43:29.892837+00:00"} {"global_step": 6899, "acc_step": 0, "speed/wps": 12975.791633064739, "speed/FLOPS": 203802638203499.53, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.6296000480651855, "optim/lr": 0.002991193523164307, "optim/total_tokens": 1808531456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 6.076025009155273, "created_at": "2025-01-14T22:43:34.950966+00:00"} {"global_step": 6900, "acc_step": 0, "speed/wps": 12975.527958294291, "speed/FLOPS": 203798496828902.56, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.845853805541992, "optim/lr": 0.0029911842500787984, "optim/total_tokens": 1808793600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427880, "loss/out": 6.038548946380615, "created_at": "2025-01-14T22:43:40.006786+00:00"} {"global_step": 6901, "acc_step": 0, "speed/wps": 12969.112477308472, "speed/FLOPS": 203697732884225.56, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 18.960006713867188, "optim/lr": 0.0029911749721280387, "optim/total_tokens": 1809055744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 5.900398254394531, "created_at": "2025-01-14T22:43:45.067211+00:00"} {"global_step": 6902, "acc_step": 0, "speed/wps": 12974.11003780068, "speed/FLOPS": 203776226439123.47, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 24.739652633666992, "optim/lr": 0.0029911656893120584, "optim/total_tokens": 1809317888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461100, "loss/out": 6.023724555969238, "created_at": "2025-01-14T22:43:50.122267+00:00"} {"global_step": 6903, "acc_step": 0, "speed/wps": 12969.65860221585, "speed/FLOPS": 203706310526350.22, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.937820911407471, "optim/lr": 0.0029911564016308874, "optim/total_tokens": 1809580032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433708, "loss/out": 6.275960922241211, "created_at": "2025-01-14T22:43:55.184102+00:00"} {"global_step": 6904, "acc_step": 0, "speed/wps": 12972.680849654762, "speed/FLOPS": 203753779075381.5, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.0502893924713135, "optim/lr": 0.002991147109084557, "optim/total_tokens": 1809842176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300290, "loss/out": 6.1095290184021, "created_at": "2025-01-14T22:44:00.244978+00:00"} {"global_step": 6905, "acc_step": 0, "speed/wps": 12970.306251726082, "speed/FLOPS": 203716482751872.1, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.530281066894531, "optim/lr": 0.0029911378116730964, "optim/total_tokens": 1810104320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316371, "loss/out": 6.095968246459961, "created_at": "2025-01-14T22:44:05.301973+00:00"} {"global_step": 6906, "acc_step": 0, "speed/wps": 12969.864431663083, "speed/FLOPS": 203709543360658.3, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.1429901123046875, "optim/lr": 0.0029911285093965367, "optim/total_tokens": 1810366464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359599, "loss/out": 6.030447006225586, "created_at": "2025-01-14T22:44:10.364040+00:00"} {"global_step": 6907, "acc_step": 0, "speed/wps": 12971.79284859476, "speed/FLOPS": 203739831798492.84, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.895118713378906, "optim/lr": 0.0029911192022549077, "optim/total_tokens": 1810628608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429739, "loss/out": 6.146603584289551, "created_at": "2025-01-14T22:44:15.422954+00:00"} {"global_step": 6908, "acc_step": 0, "speed/wps": 12967.748503748044, "speed/FLOPS": 203676309805162.97, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.853894233703613, "optim/lr": 0.0029911098902482406, "optim/total_tokens": 1810890752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 6.005101203918457, "created_at": "2025-01-14T22:44:20.485164+00:00"} {"global_step": 6909, "acc_step": 0, "speed/wps": 12970.568659986515, "speed/FLOPS": 203720604234188.94, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.134733200073242, "optim/lr": 0.002991100573376565, "optim/total_tokens": 1811152896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433464, "loss/out": 5.98217248916626, "created_at": "2025-01-14T22:44:25.544178+00:00"} {"global_step": 6910, "acc_step": 0, "speed/wps": 12967.622307962527, "speed/FLOPS": 203674327726940.25, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.946139335632324, "optim/lr": 0.0029910912516399116, "optim/total_tokens": 1811415040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 6.0490217208862305, "created_at": "2025-01-14T22:44:30.606726+00:00"} {"global_step": 6911, "acc_step": 0, "speed/wps": 12976.182066165213, "speed/FLOPS": 203808770491853.78, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.6749494075775146, "optim/lr": 0.0029910819250383107, "optim/total_tokens": 1811677184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 6.100980758666992, "created_at": "2025-01-14T22:44:35.660752+00:00"} {"global_step": 6912, "acc_step": 0, "speed/wps": 12971.723522607388, "speed/FLOPS": 203738742938597.28, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.4415123462677, "optim/lr": 0.002991072593571793, "optim/total_tokens": 1811939328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310495, "loss/out": 6.028378963470459, "created_at": "2025-01-14T22:44:40.720721+00:00"} {"global_step": 6913, "acc_step": 0, "speed/wps": 12965.850074420508, "speed/FLOPS": 203646492363858.75, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.897510051727295, "optim/lr": 0.0029910632572403887, "optim/total_tokens": 1812201472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359782, "loss/out": 6.015254974365234, "created_at": "2025-01-14T22:44:45.779190+00:00"} {"global_step": 6914, "acc_step": 0, "speed/wps": 12968.4178188295, "speed/FLOPS": 203686822318252.56, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 47.15229034423828, "optim/lr": 0.002991053916044128, "optim/total_tokens": 1812463616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456584, "loss/out": 5.986285209655762, "created_at": "2025-01-14T22:44:50.837315+00:00"} {"global_step": 6915, "acc_step": 0, "speed/wps": 12970.77238329319, "speed/FLOPS": 203723803988665.4, "speed/curr_iter_time": 1.2723, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1937129497528076, "optim/lr": 0.002991044569983042, "optim/total_tokens": 1812725760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459237, "loss/out": 6.0931243896484375, "created_at": "2025-01-14T22:44:55.897625+00:00"} {"global_step": 6916, "acc_step": 0, "speed/wps": 12967.151363288382, "speed/FLOPS": 203666930893687.0, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 58.182884216308594, "optim/lr": 0.002991035219057161, "optim/total_tokens": 1812987904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 5.914419174194336, "created_at": "2025-01-14T22:45:00.962875+00:00"} {"global_step": 6917, "acc_step": 0, "speed/wps": 12963.33978694071, "speed/FLOPS": 203607064849492.06, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.001, "optim/grad_norm": 2.1046454906463623, "optim/lr": 0.0029910258632665147, "optim/total_tokens": 1813250048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449420, "loss/out": 5.966148376464844, "created_at": "2025-01-14T22:45:06.029436+00:00"} {"global_step": 6918, "acc_step": 0, "speed/wps": 12963.727312074338, "speed/FLOPS": 203613151464231.0, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.580021381378174, "optim/lr": 0.0029910165026111346, "optim/total_tokens": 1813512192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 6.026325225830078, "created_at": "2025-01-14T22:45:11.093267+00:00"} {"global_step": 6919, "acc_step": 0, "speed/wps": 12959.828215680987, "speed/FLOPS": 203551910797453.5, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 18.069334030151367, "optim/lr": 0.002991007137091051, "optim/total_tokens": 1813774336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441442, "loss/out": 6.049705505371094, "created_at": "2025-01-14T22:45:16.156829+00:00"} {"global_step": 6920, "acc_step": 0, "speed/wps": 12967.696524876652, "speed/FLOPS": 203675493405561.3, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0013, "optim/grad_norm": 4.2740092277526855, "optim/lr": 0.0029909977667062953, "optim/total_tokens": 1814036480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312776, "loss/out": 5.970224380493164, "created_at": "2025-01-14T22:45:21.217155+00:00"} {"global_step": 6921, "acc_step": 0, "speed/wps": 12966.990725501822, "speed/FLOPS": 203664407856509.62, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.655446529388428, "optim/lr": 0.002990988391456896, "optim/total_tokens": 1814298624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375098, "loss/out": 6.105649471282959, "created_at": "2025-01-14T22:45:26.275790+00:00"} {"global_step": 6922, "acc_step": 0, "speed/wps": 12971.251098838262, "speed/FLOPS": 203731322874124.88, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 20.88053321838379, "optim/lr": 0.002990979011342885, "optim/total_tokens": 1814560768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437717, "loss/out": 5.971709728240967, "created_at": "2025-01-14T22:45:31.336290+00:00"} {"global_step": 6923, "acc_step": 0, "speed/wps": 12969.360048297822, "speed/FLOPS": 203701621326811.5, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.09843921661377, "optim/lr": 0.0029909696263642933, "optim/total_tokens": 1814822912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 6.155763626098633, "created_at": "2025-01-14T22:45:36.393582+00:00"} {"global_step": 6924, "acc_step": 0, "speed/wps": 12967.943965335662, "speed/FLOPS": 203679379797988.88, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.053803443908691, "optim/lr": 0.0029909602365211507, "optim/total_tokens": 1815085056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449411, "loss/out": 6.097259044647217, "created_at": "2025-01-14T22:45:41.450917+00:00"} {"global_step": 6925, "acc_step": 0, "speed/wps": 12970.470182428295, "speed/FLOPS": 203719057508814.72, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.088898658752441, "optim/lr": 0.002990950841813488, "optim/total_tokens": 1815347200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431076, "loss/out": 6.107316017150879, "created_at": "2025-01-14T22:45:46.517059+00:00"} {"global_step": 6926, "acc_step": 0, "speed/wps": 12971.664823740735, "speed/FLOPS": 203737820992235.8, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.50441837310791, "optim/lr": 0.002990941442241337, "optim/total_tokens": 1815609344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 6.020038604736328, "created_at": "2025-01-14T22:45:51.572882+00:00"} {"global_step": 6927, "acc_step": 0, "speed/wps": 12966.183955189854, "speed/FLOPS": 203651736420141.56, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 29.017528533935547, "optim/lr": 0.002990932037804726, "optim/total_tokens": 1815871488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331674, "loss/out": 6.0800323486328125, "created_at": "2025-01-14T22:45:56.631025+00:00"} {"global_step": 6928, "acc_step": 0, "speed/wps": 12966.896366666151, "speed/FLOPS": 203662925821332.78, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.099761009216309, "optim/lr": 0.0029909226285036875, "optim/total_tokens": 1816133632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334608, "loss/out": 5.963010311126709, "created_at": "2025-01-14T22:46:01.697434+00:00"} {"global_step": 6929, "acc_step": 0, "speed/wps": 12963.797221335508, "speed/FLOPS": 203614249485243.06, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.558900833129883, "optim/lr": 0.0029909132143382516, "optim/total_tokens": 1816395776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 6.051901340484619, "created_at": "2025-01-14T22:46:06.761658+00:00"} {"global_step": 6930, "acc_step": 0, "speed/wps": 12963.959315707923, "speed/FLOPS": 203616795400102.88, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.1664373874664307, "optim/lr": 0.00299090379530845, "optim/total_tokens": 1816657920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 6.037834167480469, "created_at": "2025-01-14T22:46:11.820387+00:00"} {"global_step": 6931, "acc_step": 0, "speed/wps": 12971.96090597454, "speed/FLOPS": 203742471370577.3, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 17.41489028930664, "optim/lr": 0.0029908943714143117, "optim/total_tokens": 1816920064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 5.967505931854248, "created_at": "2025-01-14T22:46:16.880554+00:00"} {"global_step": 6932, "acc_step": 0, "speed/wps": 12974.26052757104, "speed/FLOPS": 203778590087761.56, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 17.56574058532715, "optim/lr": 0.0029908849426558684, "optim/total_tokens": 1817182208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 6.088557720184326, "created_at": "2025-01-14T22:46:21.938518+00:00"} {"global_step": 6933, "acc_step": 0, "speed/wps": 12966.60838119121, "speed/FLOPS": 203658402613715.3, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.1753411293029785, "optim/lr": 0.0029908755090331507, "optim/total_tokens": 1817444352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429454, "loss/out": 6.071503162384033, "created_at": "2025-01-14T22:46:27.000199+00:00"} {"global_step": 6934, "acc_step": 0, "speed/wps": 12969.661685880214, "speed/FLOPS": 203706358959536.44, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.549574851989746, "optim/lr": 0.00299086607054619, "optim/total_tokens": 1817706496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 6.034308433532715, "created_at": "2025-01-14T22:46:32.057846+00:00"} {"global_step": 6935, "acc_step": 0, "speed/wps": 12964.942861919613, "speed/FLOPS": 203632243344891.56, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.4326369762420654, "optim/lr": 0.002990856627195016, "optim/total_tokens": 1817968640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466392, "loss/out": 5.96480655670166, "created_at": "2025-01-14T22:46:37.117326+00:00"} {"global_step": 6936, "acc_step": 0, "speed/wps": 12969.56967857203, "speed/FLOPS": 203704913858330.12, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.001, "optim/grad_norm": 4.435604572296143, "optim/lr": 0.0029908471789796606, "optim/total_tokens": 1818230784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324713, "loss/out": 6.016201019287109, "created_at": "2025-01-14T22:46:42.179215+00:00"} {"global_step": 6937, "acc_step": 0, "speed/wps": 12975.88584145549, "speed/FLOPS": 203804117875732.2, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.69815731048584, "optim/lr": 0.002990837725900154, "optim/total_tokens": 1818492928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353457, "loss/out": 5.919551372528076, "created_at": "2025-01-14T22:46:47.233552+00:00"} {"global_step": 6938, "acc_step": 0, "speed/wps": 12966.9965940025, "speed/FLOPS": 203664500029376.94, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.8273403644561768, "optim/lr": 0.002990828267956527, "optim/total_tokens": 1818755072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449696, "loss/out": 6.025205612182617, "created_at": "2025-01-14T22:46:52.296952+00:00"} {"global_step": 6939, "acc_step": 0, "speed/wps": 12971.296626485742, "speed/FLOPS": 203732037948390.1, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.953070640563965, "optim/lr": 0.00299081880514881, "optim/total_tokens": 1819017216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 5.931606769561768, "created_at": "2025-01-14T22:46:57.360374+00:00"} {"global_step": 6940, "acc_step": 0, "speed/wps": 12971.413482482916, "speed/FLOPS": 203733873332402.66, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 22.19504737854004, "optim/lr": 0.002990809337477035, "optim/total_tokens": 1819279360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439820, "loss/out": 6.03420352935791, "created_at": "2025-01-14T22:47:02.418314+00:00"} {"global_step": 6941, "acc_step": 0, "speed/wps": 12976.01706504338, "speed/FLOPS": 203806178922500.1, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.9975361824035645, "optim/lr": 0.002990799864941233, "optim/total_tokens": 1819541504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 6.064723014831543, "created_at": "2025-01-14T22:47:07.472955+00:00"} {"global_step": 6942, "acc_step": 0, "speed/wps": 12975.05667369497, "speed/FLOPS": 203791094656657.3, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 9.852985382080078, "optim/lr": 0.0029907903875414335, "optim/total_tokens": 1819803648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 6.05401611328125, "created_at": "2025-01-14T22:47:12.528730+00:00"} {"global_step": 6943, "acc_step": 0, "speed/wps": 12970.882078695118, "speed/FLOPS": 203725526905689.97, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0009, "optim/grad_norm": 3.6932363510131836, "optim/lr": 0.002990780905277668, "optim/total_tokens": 1820065792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460556, "loss/out": 6.080089092254639, "created_at": "2025-01-14T22:47:17.584902+00:00"} {"global_step": 6944, "acc_step": 0, "speed/wps": 12979.83754476391, "speed/FLOPS": 203866184806400.3, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 20.661222457885742, "optim/lr": 0.002990771418149969, "optim/total_tokens": 1820327936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446220, "loss/out": 5.994392395019531, "created_at": "2025-01-14T22:47:22.639260+00:00"} {"global_step": 6945, "acc_step": 0, "speed/wps": 12983.024326633458, "speed/FLOPS": 203916237594758.75, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.491334915161133, "optim/lr": 0.0029907619261583646, "optim/total_tokens": 1820590080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 6.103970050811768, "created_at": "2025-01-14T22:47:27.690618+00:00"} {"global_step": 6946, "acc_step": 0, "speed/wps": 12982.848379779345, "speed/FLOPS": 203913474107638.6, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.884406089782715, "optim/lr": 0.002990752429302888, "optim/total_tokens": 1820852224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464271, "loss/out": 6.0990190505981445, "created_at": "2025-01-14T22:47:32.747194+00:00"} {"global_step": 6947, "acc_step": 0, "speed/wps": 12978.461091031702, "speed/FLOPS": 203844565709090.38, "speed/curr_iter_time": 1.2721, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.545645713806152, "optim/lr": 0.0029907429275835688, "optim/total_tokens": 1821114368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429217, "loss/out": 6.0908637046813965, "created_at": "2025-01-14T22:47:37.800270+00:00"} {"global_step": 6948, "acc_step": 0, "speed/wps": 12984.052259707076, "speed/FLOPS": 203932382696208.0, "speed/curr_iter_time": 1.2717, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.7018089294433594, "optim/lr": 0.002990733421000439, "optim/total_tokens": 1821376512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328460, "loss/out": 6.0400567054748535, "created_at": "2025-01-14T22:47:42.856433+00:00"} {"global_step": 6949, "acc_step": 0, "speed/wps": 12973.605018011982, "speed/FLOPS": 203768294409372.03, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.352450847625732, "optim/lr": 0.00299072390955353, "optim/total_tokens": 1821638656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350820, "loss/out": 6.031987190246582, "created_at": "2025-01-14T22:47:47.912317+00:00"} {"global_step": 6950, "acc_step": 0, "speed/wps": 12970.839308241768, "speed/FLOPS": 203724855136947.34, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.095133304595947, "optim/lr": 0.0029907143932428708, "optim/total_tokens": 1821900800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438014, "loss/out": 5.908339500427246, "created_at": "2025-01-14T22:47:52.971618+00:00"} {"global_step": 6951, "acc_step": 0, "speed/wps": 12975.957633412349, "speed/FLOPS": 203805245467067.16, "speed/curr_iter_time": 1.2713, "speed/data_load_time": 0.0008, "optim/grad_norm": 9.771098136901855, "optim/lr": 0.0029907048720684945, "optim/total_tokens": 1822162944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463248, "loss/out": 6.029837608337402, "created_at": "2025-01-14T22:47:58.027489+00:00"} {"global_step": 6952, "acc_step": 0, "speed/wps": 12978.876738636061, "speed/FLOPS": 203851094025876.47, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.372466564178467, "optim/lr": 0.0029906953460304316, "optim/total_tokens": 1822425088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 6.1629791259765625, "created_at": "2025-01-14T22:48:03.083924+00:00"} {"global_step": 6953, "acc_step": 0, "speed/wps": 12981.932761349708, "speed/FLOPS": 203899093062009.56, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.949598789215088, "optim/lr": 0.0029906858151287122, "optim/total_tokens": 1822687232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 5.906696796417236, "created_at": "2025-01-14T22:48:08.137838+00:00"} {"global_step": 6954, "acc_step": 0, "speed/wps": 12983.490419374451, "speed/FLOPS": 203923558221734.6, "speed/curr_iter_time": 1.2722, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.121833324432373, "optim/lr": 0.002990676279363369, "optim/total_tokens": 1822949376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434230, "loss/out": 6.015586853027344, "created_at": "2025-01-14T22:48:13.191216+00:00"} {"global_step": 6955, "acc_step": 0, "speed/wps": 12975.004430473153, "speed/FLOPS": 203790274105068.97, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0011, "optim/grad_norm": 2.810269832611084, "optim/lr": 0.0029906667387344316, "optim/total_tokens": 1823211520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494768, "loss/out": 5.802718162536621, "created_at": "2025-01-14T22:48:18.254910+00:00"} {"global_step": 6956, "acc_step": 0, "speed/wps": 12977.148544897023, "speed/FLOPS": 203823950368417.84, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.4036288261413574, "optim/lr": 0.0029906571932419327, "optim/total_tokens": 1823473664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423342, "loss/out": 5.946926593780518, "created_at": "2025-01-14T22:48:23.310285+00:00"} {"global_step": 6957, "acc_step": 0, "speed/wps": 12979.322268375468, "speed/FLOPS": 203858091682655.88, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.04961895942688, "optim/lr": 0.002990647642885902, "optim/total_tokens": 1823735808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330329, "loss/out": 5.90565824508667, "created_at": "2025-01-14T22:48:28.370322+00:00"} {"global_step": 6958, "acc_step": 0, "speed/wps": 12977.340015580916, "speed/FLOPS": 203826957678616.66, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 61.73603057861328, "optim/lr": 0.002990638087666371, "optim/total_tokens": 1823997952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322837, "loss/out": 6.0880208015441895, "created_at": "2025-01-14T22:48:33.423888+00:00"} {"global_step": 6959, "acc_step": 0, "speed/wps": 12983.697303756157, "speed/FLOPS": 203926807625238.22, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.1360630989074707, "optim/lr": 0.0029906285275833715, "optim/total_tokens": 1824260096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429454, "loss/out": 5.971166133880615, "created_at": "2025-01-14T22:48:38.475154+00:00"} {"global_step": 6960, "acc_step": 0, "speed/wps": 12979.64746797088, "speed/FLOPS": 203863199389172.9, "speed/curr_iter_time": 1.2718, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.4073259830474854, "optim/lr": 0.0029906189626369337, "optim/total_tokens": 1824522240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321779, "loss/out": 5.938791275024414, "created_at": "2025-01-14T22:48:43.531659+00:00"} {"global_step": 6961, "acc_step": 0, "speed/wps": 12976.94868956156, "speed/FLOPS": 203820811365744.34, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0009, "optim/grad_norm": 4.176960468292236, "optim/lr": 0.0029906093928270903, "optim/total_tokens": 1824784384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364445, "loss/out": 5.991711139678955, "created_at": "2025-01-14T22:48:48.587421+00:00"} {"global_step": 6962, "acc_step": 0, "speed/wps": 12978.337305300258, "speed/FLOPS": 203842621484078.6, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8445838689804077, "optim/lr": 0.002990599818153871, "optim/total_tokens": 1825046528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441677, "loss/out": 5.987480163574219, "created_at": "2025-01-14T22:48:53.645811+00:00"} {"global_step": 6963, "acc_step": 0, "speed/wps": 12979.41377011245, "speed/FLOPS": 203859528843171.72, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 22.773221969604492, "optim/lr": 0.0029905902386173073, "optim/total_tokens": 1825308672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454466, "loss/out": 5.9982829093933105, "created_at": "2025-01-14T22:48:58.698619+00:00"} {"global_step": 6964, "acc_step": 0, "speed/wps": 12972.460751765522, "speed/FLOPS": 203750322135584.8, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0012, "optim/grad_norm": 2.2838187217712402, "optim/lr": 0.002990580654217432, "optim/total_tokens": 1825570816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 5.979773998260498, "created_at": "2025-01-14T22:49:03.755287+00:00"} {"global_step": 6965, "acc_step": 0, "speed/wps": 12977.499536119234, "speed/FLOPS": 203829463167875.12, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.801449298858643, "optim/lr": 0.002990571064954274, "optim/total_tokens": 1825832960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445691, "loss/out": 5.941247940063477, "created_at": "2025-01-14T22:49:08.811699+00:00"} {"global_step": 6966, "acc_step": 0, "speed/wps": 12976.299653889579, "speed/FLOPS": 203810617368647.44, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.23357892036438, "optim/lr": 0.0029905614708278664, "optim/total_tokens": 1826095104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 6.084821701049805, "created_at": "2025-01-14T22:49:13.870761+00:00"} {"global_step": 6967, "acc_step": 0, "speed/wps": 12977.987501416916, "speed/FLOPS": 203837127333409.8, "speed/curr_iter_time": 1.272, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.077180862426758, "optim/lr": 0.0029905518718382398, "optim/total_tokens": 1826357248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 6.046561241149902, "created_at": "2025-01-14T22:49:18.926284+00:00"} {"global_step": 6968, "acc_step": 0, "speed/wps": 12978.498837125486, "speed/FLOPS": 203845158563361.8, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.8778653144836426, "optim/lr": 0.0029905422679854256, "optim/total_tokens": 1826619392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450471, "loss/out": 5.976466178894043, "created_at": "2025-01-14T22:49:23.980066+00:00"} {"global_step": 6969, "acc_step": 0, "speed/wps": 12973.47183644792, "speed/FLOPS": 203766202609897.8, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.001, "optim/grad_norm": 2.320343494415283, "optim/lr": 0.0029905326592694547, "optim/total_tokens": 1826881536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321630, "loss/out": 5.989068031311035, "created_at": "2025-01-14T22:49:29.039894+00:00"} {"global_step": 6970, "acc_step": 0, "speed/wps": 12973.98147683758, "speed/FLOPS": 203774207212536.7, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.049710750579834, "optim/lr": 0.0029905230456903596, "optim/total_tokens": 1827143680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303156, "loss/out": 5.963594436645508, "created_at": "2025-01-14T22:49:34.100688+00:00"} {"global_step": 6971, "acc_step": 0, "speed/wps": 12961.614391090541, "speed/FLOPS": 203579965136722.78, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.465700149536133, "optim/lr": 0.0029905134272481706, "optim/total_tokens": 1827405824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 408834, "loss/out": 6.031309127807617, "created_at": "2025-01-14T22:49:39.163919+00:00"} {"global_step": 6972, "acc_step": 0, "speed/wps": 12970.198547900402, "speed/FLOPS": 203714791115289.22, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.757858753204346, "optim/lr": 0.0029905038039429194, "optim/total_tokens": 1827667968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441677, "loss/out": 6.029689788818359, "created_at": "2025-01-14T22:49:44.224336+00:00"} {"global_step": 6973, "acc_step": 0, "speed/wps": 12970.170539269964, "speed/FLOPS": 203714351201260.3, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3029956817626953, "optim/lr": 0.0029904941757746373, "optim/total_tokens": 1827930112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464007, "loss/out": 6.0838775634765625, "created_at": "2025-01-14T22:49:49.280782+00:00"} {"global_step": 6974, "acc_step": 0, "speed/wps": 12970.343952790901, "speed/FLOPS": 203717074898901.78, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.8163671493530273, "optim/lr": 0.002990484542743356, "optim/total_tokens": 1828192256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 6.004422664642334, "created_at": "2025-01-14T22:49:54.343524+00:00"} {"global_step": 6975, "acc_step": 0, "speed/wps": 12970.415222029065, "speed/FLOPS": 203718194280221.34, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8890764713287354, "optim/lr": 0.002990474904849107, "optim/total_tokens": 1828454400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432365, "loss/out": 5.877274513244629, "created_at": "2025-01-14T22:49:59.402625+00:00"} {"global_step": 6976, "acc_step": 0, "speed/wps": 12980.844006434007, "speed/FLOPS": 203881992669952.44, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.9545819759368896, "optim/lr": 0.002990465262091921, "optim/total_tokens": 1828716544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447263, "loss/out": 5.909510612487793, "created_at": "2025-01-14T22:50:04.462744+00:00"} {"global_step": 6977, "acc_step": 0, "speed/wps": 12980.603493913215, "speed/FLOPS": 203878215090314.6, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.500535726547241, "optim/lr": 0.00299045561447183, "optim/total_tokens": 1828978688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 5.98614501953125, "created_at": "2025-01-14T22:50:09.520611+00:00"} {"global_step": 6978, "acc_step": 0, "speed/wps": 12980.011887273457, "speed/FLOPS": 203868923095084.47, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8101990222930908, "optim/lr": 0.002990445961988866, "optim/total_tokens": 1829240832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427589, "loss/out": 5.962283611297607, "created_at": "2025-01-14T22:50:14.573720+00:00"} {"global_step": 6979, "acc_step": 0, "speed/wps": 12979.226592852001, "speed/FLOPS": 203856588967087.88, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.0764145851135254, "optim/lr": 0.0029904363046430594, "optim/total_tokens": 1829502976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322837, "loss/out": 5.802196502685547, "created_at": "2025-01-14T22:50:19.626487+00:00"} {"global_step": 6980, "acc_step": 0, "speed/wps": 12976.266395258248, "speed/FLOPS": 203810094996140.22, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0009, "optim/grad_norm": 2.794872283935547, "optim/lr": 0.002990426642434443, "optim/total_tokens": 1829765120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324322, "loss/out": 5.915965557098389, "created_at": "2025-01-14T22:50:24.684537+00:00"} {"global_step": 6981, "acc_step": 0, "speed/wps": 12979.188042945292, "speed/FLOPS": 203855983487829.9, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.722293496131897, "optim/lr": 0.002990416975363047, "optim/total_tokens": 1830027264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424864, "loss/out": 5.939947128295898, "created_at": "2025-01-14T22:50:29.738577+00:00"} {"global_step": 6982, "acc_step": 0, "speed/wps": 12977.364039262446, "speed/FLOPS": 203827335003547.06, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.1360926628112793, "optim/lr": 0.0029904073034289036, "optim/total_tokens": 1830289408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435044, "loss/out": 5.940169334411621, "created_at": "2025-01-14T22:50:34.792198+00:00"} {"global_step": 6983, "acc_step": 0, "speed/wps": 12977.305490370101, "speed/FLOPS": 203826415412738.2, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.1203503608703613, "optim/lr": 0.002990397626632044, "optim/total_tokens": 1830551552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470641, "loss/out": 5.927422523498535, "created_at": "2025-01-14T22:50:39.846646+00:00"} {"global_step": 6984, "acc_step": 0, "speed/wps": 12985.172822691902, "speed/FLOPS": 203949982677699.6, "speed/curr_iter_time": 1.2708, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8576891422271729, "optim/lr": 0.0029903879449725, "optim/total_tokens": 1830813696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 5.99163818359375, "created_at": "2025-01-14T22:50:44.902324+00:00"} {"global_step": 6985, "acc_step": 0, "speed/wps": 12974.750208620448, "speed/FLOPS": 203786281201534.8, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.311340570449829, "optim/lr": 0.002990378258450304, "optim/total_tokens": 1831075840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469353, "loss/out": 5.929317474365234, "created_at": "2025-01-14T22:50:49.958745+00:00"} {"global_step": 6986, "acc_step": 0, "speed/wps": 12978.136927454898, "speed/FLOPS": 203839474274669.22, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.000845432281494, "optim/lr": 0.0029903685670654863, "optim/total_tokens": 1831337984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 5.845860958099365, "created_at": "2025-01-14T22:50:55.013133+00:00"} {"global_step": 6987, "acc_step": 0, "speed/wps": 12980.117180274092, "speed/FLOPS": 203870576866347.2, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8044122457504272, "optim/lr": 0.0029903588708180794, "optim/total_tokens": 1831600128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 5.837216377258301, "created_at": "2025-01-14T22:51:00.066699+00:00"} {"global_step": 6988, "acc_step": 0, "speed/wps": 12968.369125568639, "speed/FLOPS": 203686057523679.16, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.434279918670654, "optim/lr": 0.0029903491697081137, "optim/total_tokens": 1831862272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431320, "loss/out": 5.892839431762695, "created_at": "2025-01-14T22:51:05.128985+00:00"} {"global_step": 6989, "acc_step": 0, "speed/wps": 12980.256606283412, "speed/FLOPS": 203872766743414.97, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.072832107543945, "optim/lr": 0.0029903394637356227, "optim/total_tokens": 1832124416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 5.924707412719727, "created_at": "2025-01-14T22:51:10.181703+00:00"} {"global_step": 6990, "acc_step": 0, "speed/wps": 12985.49785304886, "speed/FLOPS": 203955087726094.12, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4369466304779053, "optim/lr": 0.0029903297529006376, "optim/total_tokens": 1832386560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 5.928591728210449, "created_at": "2025-01-14T22:51:15.235212+00:00"} {"global_step": 6991, "acc_step": 0, "speed/wps": 12974.202858270923, "speed/FLOPS": 203777684312157.3, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3664510250091553, "optim/lr": 0.0029903200372031883, "optim/total_tokens": 1832648704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441450, "loss/out": 5.775195121765137, "created_at": "2025-01-14T22:51:20.290022+00:00"} {"global_step": 6992, "acc_step": 0, "speed/wps": 12968.71534704172, "speed/FLOPS": 203691495407672.06, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.633314847946167, "optim/lr": 0.0029903103166433083, "optim/total_tokens": 1832910848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293306, "loss/out": 5.96945858001709, "created_at": "2025-01-14T22:51:25.349665+00:00"} {"global_step": 6993, "acc_step": 0, "speed/wps": 12981.560379147388, "speed/FLOPS": 203893244287815.38, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.160642623901367, "optim/lr": 0.0029903005912210287, "optim/total_tokens": 1833172992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363102, "loss/out": 5.845968246459961, "created_at": "2025-01-14T22:51:30.406373+00:00"} {"global_step": 6994, "acc_step": 0, "speed/wps": 12978.200529528553, "speed/FLOPS": 203840473232634.38, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.303267002105713, "optim/lr": 0.0029902908609363817, "optim/total_tokens": 1833435136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439583, "loss/out": 5.894037246704102, "created_at": "2025-01-14T22:51:35.464164+00:00"} {"global_step": 6995, "acc_step": 0, "speed/wps": 12977.779693190198, "speed/FLOPS": 203833863419650.78, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.4982919692993164, "optim/lr": 0.0029902811257893984, "optim/total_tokens": 1833697280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 6.101879119873047, "created_at": "2025-01-14T22:51:40.523376+00:00"} {"global_step": 6996, "acc_step": 0, "speed/wps": 12979.300789438808, "speed/FLOPS": 203857754326440.5, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7153809070587158, "optim/lr": 0.002990271385780111, "optim/total_tokens": 1833959424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461889, "loss/out": 5.947192192077637, "created_at": "2025-01-14T22:51:45.576254+00:00"} {"global_step": 6997, "acc_step": 0, "speed/wps": 12978.241290044198, "speed/FLOPS": 203841113432543.4, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0009, "optim/grad_norm": 2.9698920249938965, "optim/lr": 0.0029902616409085508, "optim/total_tokens": 1834221568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442256, "loss/out": 5.787419319152832, "created_at": "2025-01-14T22:51:50.635268+00:00"} {"global_step": 6998, "acc_step": 0, "speed/wps": 12972.957141872937, "speed/FLOPS": 203758118624333.25, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.6493992805480957, "optim/lr": 0.00299025189117475, "optim/total_tokens": 1834483712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439064, "loss/out": 5.843640327453613, "created_at": "2025-01-14T22:51:55.698006+00:00"} {"global_step": 6999, "acc_step": 0, "speed/wps": 12979.371046318332, "speed/FLOPS": 203858857807276.47, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0013, "optim/grad_norm": 2.4442999362945557, "optim/lr": 0.0029902421365787407, "optim/total_tokens": 1834745856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430791, "loss/out": 5.752133369445801, "created_at": "2025-01-14T22:52:00.752984+00:00"} {"global_step": 7000, "acc_step": 0, "speed/wps": 12980.20380400454, "speed/FLOPS": 203871937411067.38, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.9627339839935303, "optim/lr": 0.002990232377120554, "optim/total_tokens": 1835008000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462435, "loss/out": 5.821667194366455, "created_at": "2025-01-14T22:52:05.807305+00:00"} {"global_step": 7001, "acc_step": 0, "speed/wps": 12322.426836833964, "speed/FLOPS": 193540646261380.66, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.268, "optim/grad_norm": 2.350611686706543, "optim/lr": 0.0029902226128002223, "optim/total_tokens": 1835270144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 5.8591156005859375, "created_at": "2025-01-14T22:52:11.129574+00:00"} {"global_step": 7002, "acc_step": 0, "speed/wps": 12982.535259159433, "speed/FLOPS": 203908556118030.06, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6269623041152954, "optim/lr": 0.002990212843617777, "optim/total_tokens": 1835532288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 5.804522514343262, "created_at": "2025-01-14T22:52:16.182722+00:00"} {"global_step": 7003, "acc_step": 0, "speed/wps": 12983.810907193285, "speed/FLOPS": 203928591923325.78, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4354525804519653, "optim/lr": 0.0029902030695732507, "optim/total_tokens": 1835794432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430268, "loss/out": 5.842268943786621, "created_at": "2025-01-14T22:52:21.234798+00:00"} {"global_step": 7004, "acc_step": 0, "speed/wps": 12976.71454804813, "speed/FLOPS": 203817133851531.8, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3893591165542603, "optim/lr": 0.0029901932906666747, "optim/total_tokens": 1836056576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443543, "loss/out": 5.7286248207092285, "created_at": "2025-01-14T22:52:26.294324+00:00"} {"global_step": 7005, "acc_step": 0, "speed/wps": 12975.867797881348, "speed/FLOPS": 203803834476605.84, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7035068273544312, "optim/lr": 0.002990183506898081, "optim/total_tokens": 1836318720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 5.722778797149658, "created_at": "2025-01-14T22:52:31.348521+00:00"} {"global_step": 7006, "acc_step": 0, "speed/wps": 12978.262971324782, "speed/FLOPS": 203841453966848.2, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3197050094604492, "optim/lr": 0.002990173718267501, "optim/total_tokens": 1836580864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 5.756430625915527, "created_at": "2025-01-14T22:52:36.401790+00:00"} {"global_step": 7007, "acc_step": 0, "speed/wps": 12977.299442613621, "speed/FLOPS": 203826320424412.2, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.779500722885132, "optim/lr": 0.0029901639247749676, "optim/total_tokens": 1836843008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325525, "loss/out": 5.814301490783691, "created_at": "2025-01-14T22:52:41.466141+00:00"} {"global_step": 7008, "acc_step": 0, "speed/wps": 12980.489098967613, "speed/FLOPS": 203876418360498.78, "speed/curr_iter_time": 1.2721, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3250304460525513, "optim/lr": 0.0029901541264205126, "optim/total_tokens": 1837105152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369482, "loss/out": 5.847838401794434, "created_at": "2025-01-14T22:52:46.530807+00:00"} {"global_step": 7009, "acc_step": 0, "speed/wps": 12974.990918167367, "speed/FLOPS": 203790061875738.75, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.3335306644439697, "optim/lr": 0.0029901443232041676, "optim/total_tokens": 1837367296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474645, "loss/out": 5.769427299499512, "created_at": "2025-01-14T22:52:51.585320+00:00"} {"global_step": 7010, "acc_step": 0, "speed/wps": 12979.688675702626, "speed/FLOPS": 203863846613225.3, "speed/curr_iter_time": 1.272, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7854743003845215, "optim/lr": 0.0029901345151259646, "optim/total_tokens": 1837629440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 5.922183036804199, "created_at": "2025-01-14T22:52:56.638279+00:00"} {"global_step": 7011, "acc_step": 0, "speed/wps": 12981.11580207988, "speed/FLOPS": 203886261594057.25, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.0020053386688232, "optim/lr": 0.0029901247021859355, "optim/total_tokens": 1837891584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 6.029869079589844, "created_at": "2025-01-14T22:53:01.694332+00:00"} {"global_step": 7012, "acc_step": 0, "speed/wps": 12972.41012215256, "speed/FLOPS": 203749526927940.84, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.1107261180877686, "optim/lr": 0.0029901148843841127, "optim/total_tokens": 1838153728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495864, "loss/out": 5.862139701843262, "created_at": "2025-01-14T22:53:06.751072+00:00"} {"global_step": 7013, "acc_step": 0, "speed/wps": 12964.71018293346, "speed/FLOPS": 203628588801679.62, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 19.429841995239258, "optim/lr": 0.0029901050617205283, "optim/total_tokens": 1838415872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 5.933749675750732, "created_at": "2025-01-14T22:53:11.814923+00:00"} {"global_step": 7014, "acc_step": 0, "speed/wps": 12968.23522696148, "speed/FLOPS": 203683954462057.78, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 31.852083206176758, "optim/lr": 0.0029900952341952135, "optim/total_tokens": 1838678016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494003, "loss/out": 5.779881954193115, "created_at": "2025-01-14T22:53:16.874294+00:00"} {"global_step": 7015, "acc_step": 0, "speed/wps": 12967.508602903035, "speed/FLOPS": 203672541832733.8, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 103.02424621582031, "optim/lr": 0.0029900854018082013, "optim/total_tokens": 1838940160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445166, "loss/out": 5.8185601234436035, "created_at": "2025-01-14T22:53:21.932846+00:00"} {"global_step": 7016, "acc_step": 0, "speed/wps": 12972.152551222196, "speed/FLOPS": 203745481422539.94, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.92331075668335, "optim/lr": 0.0029900755645595233, "optim/total_tokens": 1839202304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308424, "loss/out": 5.940716743469238, "created_at": "2025-01-14T22:53:26.993761+00:00"} {"global_step": 7017, "acc_step": 0, "speed/wps": 12959.639908290905, "speed/FLOPS": 203548953171130.8, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.338881969451904, "optim/lr": 0.002990065722449212, "optim/total_tokens": 1839464448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434505, "loss/out": 5.931879043579102, "created_at": "2025-01-14T22:53:32.054697+00:00"} {"global_step": 7018, "acc_step": 0, "speed/wps": 12964.094310670049, "speed/FLOPS": 203618915681485.56, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.001, "optim/grad_norm": 10.24977970123291, "optim/lr": 0.0029900558754772993, "optim/total_tokens": 1839726592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 5.805418491363525, "created_at": "2025-01-14T22:53:37.117840+00:00"} {"global_step": 7019, "acc_step": 0, "speed/wps": 12968.572189447259, "speed/FLOPS": 203689246920931.9, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.536237716674805, "optim/lr": 0.002990046023643817, "optim/total_tokens": 1839988736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 5.88753604888916, "created_at": "2025-01-14T22:53:42.178983+00:00"} {"global_step": 7020, "acc_step": 0, "speed/wps": 12965.971352437748, "speed/FLOPS": 203648397201772.94, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.001, "optim/grad_norm": 1.6138110160827637, "optim/lr": 0.002990036166948798, "optim/total_tokens": 1840250880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452847, "loss/out": 5.783623695373535, "created_at": "2025-01-14T22:53:47.243978+00:00"} {"global_step": 7021, "acc_step": 0, "speed/wps": 12976.576076347643, "speed/FLOPS": 203814958963195.6, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.2163543701171875, "optim/lr": 0.0029900263053922733, "optim/total_tokens": 1840513024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 5.852062702178955, "created_at": "2025-01-14T22:53:52.297864+00:00"} {"global_step": 7022, "acc_step": 0, "speed/wps": 12967.23205285849, "speed/FLOPS": 203668198234264.62, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.1495423316955566, "optim/lr": 0.002990016438974276, "optim/total_tokens": 1840775168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431076, "loss/out": 5.7986602783203125, "created_at": "2025-01-14T22:53:57.360639+00:00"} {"global_step": 7023, "acc_step": 0, "speed/wps": 12965.467123704744, "speed/FLOPS": 203640477596638.25, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9525107145309448, "optim/lr": 0.002990006567694838, "optim/total_tokens": 1841037312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289551, "loss/out": 5.743088245391846, "created_at": "2025-01-14T22:54:02.423140+00:00"} {"global_step": 7024, "acc_step": 0, "speed/wps": 12959.565957797779, "speed/FLOPS": 203547791677017.62, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.6031020879745483, "optim/lr": 0.0029899966915539916, "optim/total_tokens": 1841299456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353744, "loss/out": 5.7802886962890625, "created_at": "2025-01-14T22:54:07.490689+00:00"} {"global_step": 7025, "acc_step": 0, "speed/wps": 12971.951591923535, "speed/FLOPS": 203742325080607.44, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.262544870376587, "optim/lr": 0.0029899868105517694, "optim/total_tokens": 1841561600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 393877, "loss/out": 5.708840847015381, "created_at": "2025-01-14T22:54:12.546724+00:00"} {"global_step": 7026, "acc_step": 0, "speed/wps": 12969.180824065803, "speed/FLOPS": 203698806363967.75, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2784857749938965, "optim/lr": 0.0029899769246882024, "optim/total_tokens": 1841823744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444067, "loss/out": 5.663682460784912, "created_at": "2025-01-14T22:54:17.605921+00:00"} {"global_step": 7027, "acc_step": 0, "speed/wps": 12968.78173689019, "speed/FLOPS": 203692538151471.5, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.184926152229309, "optim/lr": 0.0029899670339633244, "optim/total_tokens": 1842085888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 5.725808620452881, "created_at": "2025-01-14T22:54:22.666049+00:00"} {"global_step": 7028, "acc_step": 0, "speed/wps": 12960.53050472, "speed/FLOPS": 203562941211857.47, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.526389241218567, "optim/lr": 0.002989957138377166, "optim/total_tokens": 1842348032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439583, "loss/out": 5.751127243041992, "created_at": "2025-01-14T22:54:27.732920+00:00"} {"global_step": 7029, "acc_step": 0, "speed/wps": 12963.294417818983, "speed/FLOPS": 203606352265091.6, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7687485218048096, "optim/lr": 0.002989947237929761, "optim/total_tokens": 1842610176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 5.687475204467773, "created_at": "2025-01-14T22:54:32.796221+00:00"} {"global_step": 7030, "acc_step": 0, "speed/wps": 12969.166219588718, "speed/FLOPS": 203698576980586.47, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 39.004703521728516, "optim/lr": 0.0029899373326211414, "optim/total_tokens": 1842872320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331674, "loss/out": 5.689502239227295, "created_at": "2025-01-14T22:54:37.857468+00:00"} {"global_step": 7031, "acc_step": 0, "speed/wps": 12971.598739132129, "speed/FLOPS": 203736783042647.88, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.340895175933838, "optim/lr": 0.0029899274224513394, "optim/total_tokens": 1843134464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348945, "loss/out": 5.791956901550293, "created_at": "2025-01-14T22:54:42.918473+00:00"} {"global_step": 7032, "acc_step": 0, "speed/wps": 12967.46259046231, "speed/FLOPS": 203671819144124.5, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.098589897155762, "optim/lr": 0.0029899175074203864, "optim/total_tokens": 1843396608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441450, "loss/out": 5.736743450164795, "created_at": "2025-01-14T22:54:47.976065+00:00"} {"global_step": 7033, "acc_step": 0, "speed/wps": 12964.981637785197, "speed/FLOPS": 203632852373144.2, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3976243734359741, "optim/lr": 0.002989907587528316, "optim/total_tokens": 1843658752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 5.792757511138916, "created_at": "2025-01-14T22:54:53.051790+00:00"} {"global_step": 7034, "acc_step": 0, "speed/wps": 12963.93387997396, "speed/FLOPS": 203616395896948.6, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2759106159210205, "optim/lr": 0.00298989766277516, "optim/total_tokens": 1843920896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472260, "loss/out": 5.529412746429443, "created_at": "2025-01-14T22:54:58.115024+00:00"} {"global_step": 7035, "acc_step": 0, "speed/wps": 12962.968291527372, "speed/FLOPS": 203601230003537.38, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.295460820198059, "optim/lr": 0.0029898877331609505, "optim/total_tokens": 1844183040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447026, "loss/out": 5.745525360107422, "created_at": "2025-01-14T22:55:03.180019+00:00"} {"global_step": 7036, "acc_step": 0, "speed/wps": 12964.954322907175, "speed/FLOPS": 203632423355450.28, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 22.606647491455078, "optim/lr": 0.0029898777986857206, "optim/total_tokens": 1844445184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313228, "loss/out": 5.695446968078613, "created_at": "2025-01-14T22:55:08.238723+00:00"} {"global_step": 7037, "acc_step": 0, "speed/wps": 12963.490616836394, "speed/FLOPS": 203609433840265.75, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.6208267211914062, "optim/lr": 0.0029898678593495024, "optim/total_tokens": 1844707328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320810, "loss/out": 5.779098033905029, "created_at": "2025-01-14T22:55:13.299676+00:00"} {"global_step": 7038, "acc_step": 0, "speed/wps": 12966.61434491841, "speed/FLOPS": 203658496282246.0, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2551805973052979, "optim/lr": 0.002989857915152328, "optim/total_tokens": 1844969472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325000, "loss/out": 5.772323131561279, "created_at": "2025-01-14T22:55:18.363218+00:00"} {"global_step": 7039, "acc_step": 0, "speed/wps": 12965.89096524222, "speed/FLOPS": 203647134610403.62, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2521599531173706, "optim/lr": 0.0029898479660942306, "optim/total_tokens": 1845231616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 406528, "loss/out": 5.743208885192871, "created_at": "2025-01-14T22:55:23.424460+00:00"} {"global_step": 7040, "acc_step": 0, "speed/wps": 12974.594944504977, "speed/FLOPS": 203783842565246.1, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4340720176696777, "optim/lr": 0.002989838012175242, "optim/total_tokens": 1845493760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431076, "loss/out": 5.820662021636963, "created_at": "2025-01-14T22:55:28.487132+00:00"} {"global_step": 7041, "acc_step": 0, "speed/wps": 12960.705768399404, "speed/FLOPS": 203565693968780.06, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.686750411987305, "optim/lr": 0.002989828053395395, "optim/total_tokens": 1845755904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 5.606165409088135, "created_at": "2025-01-14T22:55:33.547871+00:00"} {"global_step": 7042, "acc_step": 0, "speed/wps": 12964.997441531066, "speed/FLOPS": 203633100592689.7, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.169276237487793, "optim/lr": 0.002989818089754722, "optim/total_tokens": 1846018048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423864, "loss/out": 5.652412414550781, "created_at": "2025-01-14T22:55:38.610731+00:00"} {"global_step": 7043, "acc_step": 0, "speed/wps": 12962.128594297199, "speed/FLOPS": 203588041404672.6, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.877762794494629, "optim/lr": 0.0029898081212532553, "optim/total_tokens": 1846280192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 5.845937252044678, "created_at": "2025-01-14T22:55:43.677808+00:00"} {"global_step": 7044, "acc_step": 0, "speed/wps": 12966.937675070234, "speed/FLOPS": 203663574626582.6, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.7114235162734985, "optim/lr": 0.002989798147891028, "optim/total_tokens": 1846542336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316975, "loss/out": 5.75830078125, "created_at": "2025-01-14T22:55:48.735432+00:00"} {"global_step": 7045, "acc_step": 0, "speed/wps": 12966.419286575228, "speed/FLOPS": 203655432622928.1, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.323689579963684, "optim/lr": 0.002989788169668072, "optim/total_tokens": 1846804480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338354, "loss/out": 5.619777679443359, "created_at": "2025-01-14T22:55:53.793707+00:00"} {"global_step": 7046, "acc_step": 0, "speed/wps": 12963.906411881702, "speed/FLOPS": 203615964472814.6, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8779743909835815, "optim/lr": 0.0029897781865844196, "optim/total_tokens": 1847066624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464820, "loss/out": 5.796308517456055, "created_at": "2025-01-14T22:55:58.853015+00:00"} {"global_step": 7047, "acc_step": 0, "speed/wps": 12966.43684036728, "speed/FLOPS": 203655708329353.7, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.2493503093719482, "optim/lr": 0.0029897681986401047, "optim/total_tokens": 1847328768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464820, "loss/out": 5.714824199676514, "created_at": "2025-01-14T22:56:03.913446+00:00"} {"global_step": 7048, "acc_step": 0, "speed/wps": 12962.273799201736, "speed/FLOPS": 203590322047230.78, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.8693766593933105, "optim/lr": 0.0029897582058351587, "optim/total_tokens": 1847590912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444881, "loss/out": 5.746367931365967, "created_at": "2025-01-14T22:56:08.975053+00:00"} {"global_step": 7049, "acc_step": 0, "speed/wps": 12964.05404737963, "speed/FLOPS": 203618283291181.97, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.5051584243774414, "optim/lr": 0.0029897482081696143, "optim/total_tokens": 1847853056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468777, "loss/out": 5.740202903747559, "created_at": "2025-01-14T22:56:14.038116+00:00"} {"global_step": 7050, "acc_step": 0, "speed/wps": 12969.198464214553, "speed/FLOPS": 203699083426743.97, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.564624071121216, "optim/lr": 0.002989738205643505, "optim/total_tokens": 1848115200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 5.582694053649902, "created_at": "2025-01-14T22:56:19.095757+00:00"} {"global_step": 7051, "acc_step": 0, "speed/wps": 12970.90041922245, "speed/FLOPS": 203725814968874.25, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7799793481826782, "optim/lr": 0.0029897281982568624, "optim/total_tokens": 1848377344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462435, "loss/out": 5.655317783355713, "created_at": "2025-01-14T22:56:24.155078+00:00"} {"global_step": 7052, "acc_step": 0, "speed/wps": 12968.591651543407, "speed/FLOPS": 203689552599895.62, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.587372779846191, "optim/lr": 0.0029897181860097195, "optim/total_tokens": 1848639488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421476, "loss/out": 5.656958103179932, "created_at": "2025-01-14T22:56:29.216368+00:00"} {"global_step": 7053, "acc_step": 0, "speed/wps": 12969.455991947516, "speed/FLOPS": 203703128253670.34, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5004351139068604, "optim/lr": 0.002989708168902109, "optim/total_tokens": 1848901632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298593, "loss/out": 5.767245292663574, "created_at": "2025-01-14T22:56:34.273115+00:00"} {"global_step": 7054, "acc_step": 0, "speed/wps": 12966.929964371173, "speed/FLOPS": 203663453519458.22, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.235769510269165, "optim/lr": 0.002989698146934064, "optim/total_tokens": 1849163776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367900, "loss/out": 5.681107044219971, "created_at": "2025-01-14T22:56:39.331606+00:00"} {"global_step": 7055, "acc_step": 0, "speed/wps": 12966.096882974318, "speed/FLOPS": 203650368831346.8, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0015, "optim/grad_norm": 1.2585487365722656, "optim/lr": 0.002989688120105617, "optim/total_tokens": 1849425920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447791, "loss/out": 5.741946220397949, "created_at": "2025-01-14T22:56:44.390104+00:00"} {"global_step": 7056, "acc_step": 0, "speed/wps": 12973.767644778818, "speed/FLOPS": 203770848686217.8, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.2060627937316895, "optim/lr": 0.0029896780884168, "optim/total_tokens": 1849688064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 5.699103355407715, "created_at": "2025-01-14T22:56:49.446620+00:00"} {"global_step": 7057, "acc_step": 0, "speed/wps": 12967.220229412824, "speed/FLOPS": 203668012530803.3, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.375153064727783, "optim/lr": 0.002989668051867647, "optim/total_tokens": 1849950208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470641, "loss/out": 5.639452934265137, "created_at": "2025-01-14T22:56:54.504173+00:00"} {"global_step": 7058, "acc_step": 0, "speed/wps": 12964.549920912927, "speed/FLOPS": 203626071666424.03, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3446155786514282, "optim/lr": 0.002989658010458189, "optim/total_tokens": 1850212352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445404, "loss/out": 5.674038887023926, "created_at": "2025-01-14T22:56:59.567816+00:00"} {"global_step": 7059, "acc_step": 0, "speed/wps": 12965.533401466239, "speed/FLOPS": 203641518579957.7, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.29119074344635, "optim/lr": 0.0029896479641884606, "optim/total_tokens": 1850474496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466369, "loss/out": 5.707120895385742, "created_at": "2025-01-14T22:57:04.632250+00:00"} {"global_step": 7060, "acc_step": 0, "speed/wps": 12973.424927797265, "speed/FLOPS": 203765465845080.03, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8593919277191162, "optim/lr": 0.0029896379130584932, "optim/total_tokens": 1850736640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428118, "loss/out": 5.654003620147705, "created_at": "2025-01-14T22:57:09.690773+00:00"} {"global_step": 7061, "acc_step": 0, "speed/wps": 12970.216033035393, "speed/FLOPS": 203715065743361.34, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.001, "optim/grad_norm": 2.324180841445923, "optim/lr": 0.0029896278570683207, "optim/total_tokens": 1850998784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298942, "loss/out": 5.660549163818359, "created_at": "2025-01-14T22:57:14.748374+00:00"} {"global_step": 7062, "acc_step": 0, "speed/wps": 12963.003378085781, "speed/FLOPS": 203601781086151.16, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.318986654281616, "optim/lr": 0.0029896177962179748, "optim/total_tokens": 1851260928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 414480, "loss/out": 5.704236030578613, "created_at": "2025-01-14T22:57:19.807584+00:00"} {"global_step": 7063, "acc_step": 0, "speed/wps": 12967.368294057347, "speed/FLOPS": 203670338089507.72, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.210245132446289, "optim/lr": 0.002989607730507489, "optim/total_tokens": 1851523072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438240, "loss/out": 5.661686897277832, "created_at": "2025-01-14T22:57:24.867833+00:00"} {"global_step": 7064, "acc_step": 0, "speed/wps": 12963.661439669122, "speed/FLOPS": 203612116847588.9, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.255033254623413, "optim/lr": 0.002989597659936896, "optim/total_tokens": 1851785216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 5.673935890197754, "created_at": "2025-01-14T22:57:29.932673+00:00"} {"global_step": 7065, "acc_step": 0, "speed/wps": 12972.877667731147, "speed/FLOPS": 203756870373727.5, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.603940725326538, "optim/lr": 0.002989587584506229, "optim/total_tokens": 1852047360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469875, "loss/out": 5.757246971130371, "created_at": "2025-01-14T22:57:34.989221+00:00"} {"global_step": 7066, "acc_step": 0, "speed/wps": 12960.80949036987, "speed/FLOPS": 203567323064855.12, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.5803415775299072, "optim/lr": 0.0029895775042155207, "optim/total_tokens": 1852309504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458173, "loss/out": 5.611417770385742, "created_at": "2025-01-14T22:57:40.050269+00:00"} {"global_step": 7067, "acc_step": 0, "speed/wps": 12965.791091847424, "speed/FLOPS": 203645565961497.9, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.7293426990509033, "optim/lr": 0.0029895674190648035, "optim/total_tokens": 1852571648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423627, "loss/out": 5.684372901916504, "created_at": "2025-01-14T22:57:45.114582+00:00"} {"global_step": 7068, "acc_step": 0, "speed/wps": 12968.618518079615, "speed/FLOPS": 203689974575765.25, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.368072748184204, "optim/lr": 0.0029895573290541107, "optim/total_tokens": 1852833792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 5.708924293518066, "created_at": "2025-01-14T22:57:50.173798+00:00"} {"global_step": 7069, "acc_step": 0, "speed/wps": 12967.660096183097, "speed/FLOPS": 203674921242870.75, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.274343013763428, "optim/lr": 0.002989547234183475, "optim/total_tokens": 1853095936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446742, "loss/out": 5.711474895477295, "created_at": "2025-01-14T22:57:55.231967+00:00"} {"global_step": 7070, "acc_step": 0, "speed/wps": 12973.315062616693, "speed/FLOPS": 203763740261448.56, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3060053586959839, "optim/lr": 0.0029895371344529297, "optim/total_tokens": 1853358080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335427, "loss/out": 5.695474147796631, "created_at": "2025-01-14T22:58:00.292306+00:00"} {"global_step": 7071, "acc_step": 0, "speed/wps": 12965.57053728011, "speed/FLOPS": 203642101848944.5, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3030215501785278, "optim/lr": 0.0029895270298625075, "optim/total_tokens": 1853620224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367376, "loss/out": 5.648306369781494, "created_at": "2025-01-14T22:58:05.350568+00:00"} {"global_step": 7072, "acc_step": 0, "speed/wps": 12965.045163535597, "speed/FLOPS": 203633850132347.75, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 22.220090866088867, "optim/lr": 0.0029895169204122414, "optim/total_tokens": 1853882368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450227, "loss/out": 5.747130393981934, "created_at": "2025-01-14T22:58:10.408961+00:00"} {"global_step": 7073, "acc_step": 0, "speed/wps": 12965.40097323084, "speed/FLOPS": 203639438612545.3, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.630321025848389, "optim/lr": 0.0029895068061021645, "optim/total_tokens": 1854144512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 5.796148777008057, "created_at": "2025-01-14T22:58:15.469749+00:00"} {"global_step": 7074, "acc_step": 0, "speed/wps": 12963.2557523443, "speed/FLOPS": 203605744970679.78, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.774923324584961, "optim/lr": 0.0029894966869323096, "optim/total_tokens": 1854406656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420954, "loss/out": 5.713562965393066, "created_at": "2025-01-14T22:58:20.531164+00:00"} {"global_step": 7075, "acc_step": 0, "speed/wps": 12958.77528158702, "speed/FLOPS": 203535373020624.7, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.470777988433838, "optim/lr": 0.0029894865629027097, "optim/total_tokens": 1854668800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 5.834747791290283, "created_at": "2025-01-14T22:58:25.595885+00:00"} {"global_step": 7076, "acc_step": 0, "speed/wps": 12963.037106063844, "speed/FLOPS": 203602310830393.75, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5031437873840332, "optim/lr": 0.0029894764340133986, "optim/total_tokens": 1854930944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424963, "loss/out": 5.742476940155029, "created_at": "2025-01-14T22:58:30.661280+00:00"} {"global_step": 7077, "acc_step": 0, "speed/wps": 12968.006769712987, "speed/FLOPS": 203680366227037.03, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2672878503799438, "optim/lr": 0.002989466300264408, "optim/total_tokens": 1855193088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306907, "loss/out": 5.656658172607422, "created_at": "2025-01-14T22:58:35.718946+00:00"} {"global_step": 7078, "acc_step": 0, "speed/wps": 12961.000555598113, "speed/FLOPS": 203570324006815.84, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.001, "optim/grad_norm": 2.167008399963379, "optim/lr": 0.002989456161655772, "optim/total_tokens": 1855455232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335714, "loss/out": 5.674355506896973, "created_at": "2025-01-14T22:58:40.780052+00:00"} {"global_step": 7079, "acc_step": 0, "speed/wps": 12958.390104659571, "speed/FLOPS": 203529323287690.53, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3309706449508667, "optim/lr": 0.0029894460181875233, "optim/total_tokens": 1855717376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448897, "loss/out": 5.65763521194458, "created_at": "2025-01-14T22:58:45.841720+00:00"} {"global_step": 7080, "acc_step": 0, "speed/wps": 12957.871137106538, "speed/FLOPS": 203521172189134.0, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.0105090141296387, "optim/lr": 0.0029894358698596946, "optim/total_tokens": 1855979520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449411, "loss/out": 5.654119491577148, "created_at": "2025-01-14T22:58:50.903982+00:00"} {"global_step": 7081, "acc_step": 0, "speed/wps": 12960.73239936419, "speed/FLOPS": 203566112244676.78, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.0211031436920166, "optim/lr": 0.00298942571667232, "optim/total_tokens": 1856241664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456330, "loss/out": 5.643551826477051, "created_at": "2025-01-14T22:58:55.969089+00:00"} {"global_step": 7082, "acc_step": 0, "speed/wps": 12964.831544728193, "speed/FLOPS": 203630494955434.47, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1470210552215576, "optim/lr": 0.002989415558625432, "optim/total_tokens": 1856503808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441450, "loss/out": 5.742399215698242, "created_at": "2025-01-14T22:59:01.027806+00:00"} {"global_step": 7083, "acc_step": 0, "speed/wps": 12964.356204901724, "speed/FLOPS": 203623029090274.4, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.134597897529602, "optim/lr": 0.002989405395719064, "optim/total_tokens": 1856765952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 5.700124740600586, "created_at": "2025-01-14T22:59:06.087906+00:00"} {"global_step": 7084, "acc_step": 0, "speed/wps": 12966.36844368365, "speed/FLOPS": 203654634065450.4, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2952769994735718, "optim/lr": 0.0029893952279532485, "optim/total_tokens": 1857028096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443016, "loss/out": 5.565552711486816, "created_at": "2025-01-14T22:59:11.145768+00:00"} {"global_step": 7085, "acc_step": 0, "speed/wps": 12961.810025934448, "speed/FLOPS": 203583037850776.78, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7152988910675049, "optim/lr": 0.0029893850553280197, "optim/total_tokens": 1857290240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316820, "loss/out": 5.669321060180664, "created_at": "2025-01-14T22:59:16.207983+00:00"} {"global_step": 7086, "acc_step": 0, "speed/wps": 12957.657733163607, "speed/FLOPS": 203517820386962.53, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2656303644180298, "optim/lr": 0.0029893748778434094, "optim/total_tokens": 1857552384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439583, "loss/out": 5.658266067504883, "created_at": "2025-01-14T22:59:21.274393+00:00"} {"global_step": 7087, "acc_step": 0, "speed/wps": 12963.50359174342, "speed/FLOPS": 203609637629009.97, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.4993467330932617, "optim/lr": 0.0029893646954994527, "optim/total_tokens": 1857814528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 5.583651542663574, "created_at": "2025-01-14T22:59:26.333400+00:00"} {"global_step": 7088, "acc_step": 0, "speed/wps": 12966.17498630342, "speed/FLOPS": 203651595551455.6, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.4359349012374878, "optim/lr": 0.0029893545082961812, "optim/total_tokens": 1858076672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445166, "loss/out": 5.6394572257995605, "created_at": "2025-01-14T22:59:31.394699+00:00"} {"global_step": 7089, "acc_step": 0, "speed/wps": 12968.767902926833, "speed/FLOPS": 203692320870067.12, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6817524433135986, "optim/lr": 0.002989344316233629, "optim/total_tokens": 1858338816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 5.578405857086182, "created_at": "2025-01-14T22:59:36.455926+00:00"} {"global_step": 7090, "acc_step": 0, "speed/wps": 12965.634556148185, "speed/FLOPS": 203643107353238.78, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.463489055633545, "optim/lr": 0.0029893341193118284, "optim/total_tokens": 1858600960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442256, "loss/out": 5.68892240524292, "created_at": "2025-01-14T22:59:41.515004+00:00"} {"global_step": 7091, "acc_step": 0, "speed/wps": 12968.367557754007, "speed/FLOPS": 203686032898995.97, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3849265575408936, "optim/lr": 0.0029893239175308134, "optim/total_tokens": 1858863104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432365, "loss/out": 5.68596887588501, "created_at": "2025-01-14T22:59:46.572125+00:00"} {"global_step": 7092, "acc_step": 0, "speed/wps": 12969.083165915385, "speed/FLOPS": 203697272508519.06, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.951586365699768, "optim/lr": 0.002989313710890618, "optim/total_tokens": 1859125248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460308, "loss/out": 5.545657157897949, "created_at": "2025-01-14T22:59:51.632233+00:00"} {"global_step": 7093, "acc_step": 0, "speed/wps": 12962.563289907863, "speed/FLOPS": 203594868896572.44, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.8700668811798096, "optim/lr": 0.0029893034993912744, "optim/total_tokens": 1859387392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358144, "loss/out": 5.631949424743652, "created_at": "2025-01-14T22:59:56.695974+00:00"} {"global_step": 7094, "acc_step": 0, "speed/wps": 12966.751703307302, "speed/FLOPS": 203660653684493.75, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4379535913467407, "optim/lr": 0.0029892932830328156, "optim/total_tokens": 1859649536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462435, "loss/out": 5.637522220611572, "created_at": "2025-01-14T23:00:01.754999+00:00"} {"global_step": 7095, "acc_step": 0, "speed/wps": 12956.360095139007, "speed/FLOPS": 203497439198644.7, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.6442949771881104, "optim/lr": 0.0029892830618152766, "optim/total_tokens": 1859911680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 5.583690166473389, "created_at": "2025-01-14T23:00:06.816965+00:00"} {"global_step": 7096, "acc_step": 0, "speed/wps": 12962.539655287172, "speed/FLOPS": 203594497682377.1, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.4422764778137207, "optim/lr": 0.002989272835738689, "optim/total_tokens": 1860173824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431333, "loss/out": 5.6589860916137695, "created_at": "2025-01-14T23:00:11.882888+00:00"} {"global_step": 7097, "acc_step": 0, "speed/wps": 12967.74611401561, "speed/FLOPS": 203676272271131.25, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.461897850036621, "optim/lr": 0.0029892626048030873, "optim/total_tokens": 1860435968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 5.6370954513549805, "created_at": "2025-01-14T23:00:16.944988+00:00"} {"global_step": 7098, "acc_step": 0, "speed/wps": 12967.828096679661, "speed/FLOPS": 203677559921526.16, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2210729122161865, "optim/lr": 0.002989252369008504, "optim/total_tokens": 1860698112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 5.513707637786865, "created_at": "2025-01-14T23:00:22.004340+00:00"} {"global_step": 7099, "acc_step": 0, "speed/wps": 12961.187568022353, "speed/FLOPS": 203573261293920.53, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3139662742614746, "optim/lr": 0.002989242128354973, "optim/total_tokens": 1860960256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288730, "loss/out": 5.595266819000244, "created_at": "2025-01-14T23:00:27.064516+00:00"} {"global_step": 7100, "acc_step": 0, "speed/wps": 12970.265000748683, "speed/FLOPS": 203715834848587.25, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0009, "optim/grad_norm": 3.932420015335083, "optim/lr": 0.002989231882842528, "optim/total_tokens": 1861222400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370586, "loss/out": 5.617886543273926, "created_at": "2025-01-14T23:00:32.120904+00:00"} {"global_step": 7101, "acc_step": 0, "speed/wps": 12965.494187519947, "speed/FLOPS": 203640902671045.6, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4911285638809204, "optim/lr": 0.0029892216324712015, "optim/total_tokens": 1861484544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429977, "loss/out": 5.733842849731445, "created_at": "2025-01-14T23:00:37.184119+00:00"} {"global_step": 7102, "acc_step": 0, "speed/wps": 12964.091456549104, "speed/FLOPS": 203618870853594.16, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0014, "optim/grad_norm": 1.8590259552001953, "optim/lr": 0.002989211377241028, "optim/total_tokens": 1861746688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482522, "loss/out": 5.596785545349121, "created_at": "2025-01-14T23:00:42.242808+00:00"} {"global_step": 7103, "acc_step": 0, "speed/wps": 12963.427779774524, "speed/FLOPS": 203608446897861.9, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1183090209960938, "optim/lr": 0.0029892011171520406, "optim/total_tokens": 1862008832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 5.644539833068848, "created_at": "2025-01-14T23:00:47.302324+00:00"} {"global_step": 7104, "acc_step": 0, "speed/wps": 12930.942035941454, "speed/FLOPS": 203098213650873.56, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.9486234188079834, "optim/lr": 0.0029891908522042725, "optim/total_tokens": 1862270976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 5.803753852844238, "created_at": "2025-01-14T23:00:52.379471+00:00"} {"global_step": 7105, "acc_step": 0, "speed/wps": 12969.123766401104, "speed/FLOPS": 203697910194938.2, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3769137859344482, "optim/lr": 0.0029891805823977573, "optim/total_tokens": 1862533120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 5.529531002044678, "created_at": "2025-01-14T23:00:57.442352+00:00"} {"global_step": 7106, "acc_step": 0, "speed/wps": 12968.394963495184, "speed/FLOPS": 203686463343820.6, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0011, "optim/grad_norm": 59.11275863647461, "optim/lr": 0.0029891703077325293, "optim/total_tokens": 1862795264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428403, "loss/out": 5.599395275115967, "created_at": "2025-01-14T23:01:02.500690+00:00"} {"global_step": 7107, "acc_step": 0, "speed/wps": 12965.040868588576, "speed/FLOPS": 203633782674302.9, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9386249780654907, "optim/lr": 0.0029891600282086206, "optim/total_tokens": 1863057408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 285748, "loss/out": 5.715521812438965, "created_at": "2025-01-14T23:01:07.564542+00:00"} {"global_step": 7108, "acc_step": 0, "speed/wps": 12966.23821315579, "speed/FLOPS": 203652588616054.84, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2989383935928345, "optim/lr": 0.0029891497438260654, "optim/total_tokens": 1863319552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 397433, "loss/out": 5.602514266967773, "created_at": "2025-01-14T23:01:12.623575+00:00"} {"global_step": 7109, "acc_step": 0, "speed/wps": 12952.739317695372, "speed/FLOPS": 203440569913422.78, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5517023801803589, "optim/lr": 0.0029891394545848975, "optim/total_tokens": 1863581696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 5.739691734313965, "created_at": "2025-01-14T23:01:17.687626+00:00"} {"global_step": 7110, "acc_step": 0, "speed/wps": 12960.856434357447, "speed/FLOPS": 203568060384687.84, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7411224842071533, "optim/lr": 0.00298912916048515, "optim/total_tokens": 1863843840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 5.524743556976318, "created_at": "2025-01-14T23:01:22.752830+00:00"} {"global_step": 7111, "acc_step": 0, "speed/wps": 12964.96259976886, "speed/FLOPS": 203632553354936.66, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9612619876861572, "optim/lr": 0.002989118861526857, "optim/total_tokens": 1864105984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 5.59177827835083, "created_at": "2025-01-14T23:01:27.811194+00:00"} {"global_step": 7112, "acc_step": 0, "speed/wps": 12948.61251924727, "speed/FLOPS": 203375752872980.03, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8102129697799683, "optim/lr": 0.002989108557710052, "optim/total_tokens": 1864368128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 5.733781814575195, "created_at": "2025-01-14T23:01:32.879354+00:00"} {"global_step": 7113, "acc_step": 0, "speed/wps": 12955.385161225788, "speed/FLOPS": 203482126521837.97, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.256014108657837, "optim/lr": 0.002989098249034768, "optim/total_tokens": 1864630272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320435, "loss/out": 5.576408386230469, "created_at": "2025-01-14T23:01:37.947464+00:00"} {"global_step": 7114, "acc_step": 0, "speed/wps": 12964.56381992432, "speed/FLOPS": 203626289969497.22, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3073700666427612, "optim/lr": 0.0029890879355010396, "optim/total_tokens": 1864892416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 412610, "loss/out": 5.6178107261657715, "created_at": "2025-01-14T23:01:43.007526+00:00"} {"global_step": 7115, "acc_step": 0, "speed/wps": 12958.893940586746, "speed/FLOPS": 203537236723269.94, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.001, "optim/grad_norm": 2.902716636657715, "optim/lr": 0.0029890776171088993, "optim/total_tokens": 1865154560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464820, "loss/out": 5.691906452178955, "created_at": "2025-01-14T23:01:48.069438+00:00"} {"global_step": 7116, "acc_step": 0, "speed/wps": 12972.280231776063, "speed/FLOPS": 203747486821089.88, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.477421760559082, "optim/lr": 0.002989067293858382, "optim/total_tokens": 1865416704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 5.661863327026367, "created_at": "2025-01-14T23:01:53.134779+00:00"} {"global_step": 7117, "acc_step": 0, "speed/wps": 12973.759641312909, "speed/FLOPS": 203770722980787.78, "speed/curr_iter_time": 1.2723, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0745415687561035, "optim/lr": 0.0029890569657495206, "optim/total_tokens": 1865678848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433708, "loss/out": 5.589308261871338, "created_at": "2025-01-14T23:01:58.189984+00:00"} {"global_step": 7118, "acc_step": 0, "speed/wps": 12963.632687870924, "speed/FLOPS": 203611665261089.1, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.275765061378479, "optim/lr": 0.002989046632782349, "optim/total_tokens": 1865940992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443307, "loss/out": 5.603494644165039, "created_at": "2025-01-14T23:02:03.250113+00:00"} {"global_step": 7119, "acc_step": 0, "speed/wps": 12964.753760938878, "speed/FLOPS": 203629273254137.16, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.001, "optim/grad_norm": 1.3238059282302856, "optim/lr": 0.0029890362949569005, "optim/total_tokens": 1866203136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 5.524200439453125, "created_at": "2025-01-14T23:02:08.314038+00:00"} {"global_step": 7120, "acc_step": 0, "speed/wps": 12960.449772016536, "speed/FLOPS": 203561673193811.22, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3061274290084839, "optim/lr": 0.00298902595227321, "optim/total_tokens": 1866465280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311555, "loss/out": 5.5395731925964355, "created_at": "2025-01-14T23:02:13.376733+00:00"} {"global_step": 7121, "acc_step": 0, "speed/wps": 12967.042139578598, "speed/FLOPS": 203665215385235.72, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4319493770599365, "optim/lr": 0.00298901560473131, "optim/total_tokens": 1866727424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 395284, "loss/out": 5.566757678985596, "created_at": "2025-01-14T23:02:18.447847+00:00"} {"global_step": 7122, "acc_step": 0, "speed/wps": 12971.542843800355, "speed/FLOPS": 203735905129654.12, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.8926661014556885, "optim/lr": 0.0029890052523312348, "optim/total_tokens": 1866989568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435329, "loss/out": 5.5139875411987305, "created_at": "2025-01-14T23:02:23.505762+00:00"} {"global_step": 7123, "acc_step": 0, "speed/wps": 12968.13310681449, "speed/FLOPS": 203682350524821.5, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.325688362121582, "optim/lr": 0.0029889948950730183, "optim/total_tokens": 1867251712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461622, "loss/out": 5.5927276611328125, "created_at": "2025-01-14T23:02:28.567343+00:00"} {"global_step": 7124, "acc_step": 0, "speed/wps": 12969.370336272836, "speed/FLOPS": 203701782913596.25, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.370806932449341, "optim/lr": 0.002988984532956694, "optim/total_tokens": 1867513856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445691, "loss/out": 5.656611442565918, "created_at": "2025-01-14T23:02:33.626307+00:00"} {"global_step": 7125, "acc_step": 0, "speed/wps": 12965.889113730054, "speed/FLOPS": 203647105529860.8, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.337435007095337, "optim/lr": 0.0029889741659822958, "optim/total_tokens": 1867776000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457380, "loss/out": 5.526118278503418, "created_at": "2025-01-14T23:02:38.686206+00:00"} {"global_step": 7126, "acc_step": 0, "speed/wps": 12966.132887175718, "speed/FLOPS": 203650934326804.75, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2762218713760376, "optim/lr": 0.0029889637941498577, "optim/total_tokens": 1868038144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457635, "loss/out": 5.548733711242676, "created_at": "2025-01-14T23:02:43.744504+00:00"} {"global_step": 7127, "acc_step": 0, "speed/wps": 12964.544114340433, "speed/FLOPS": 203625980466223.97, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3224807977676392, "optim/lr": 0.002988953417459413, "optim/total_tokens": 1868300288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317346, "loss/out": 5.562391757965088, "created_at": "2025-01-14T23:02:48.804441+00:00"} {"global_step": 7128, "acc_step": 0, "speed/wps": 12963.194381436639, "speed/FLOPS": 203604781056242.5, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.284949541091919, "optim/lr": 0.002988943035910996, "optim/total_tokens": 1868562432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 381609, "loss/out": 5.4720354080200195, "created_at": "2025-01-14T23:02:53.869815+00:00"} {"global_step": 7129, "acc_step": 0, "speed/wps": 12966.11579530893, "speed/FLOPS": 203650665875550.16, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.6768271923065186, "optim/lr": 0.0029889326495046407, "optim/total_tokens": 1868824576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442784, "loss/out": 5.6117377281188965, "created_at": "2025-01-14T23:02:58.931073+00:00"} {"global_step": 7130, "acc_step": 0, "speed/wps": 12964.122620274136, "speed/FLOPS": 203619360322719.22, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4820623397827148, "optim/lr": 0.0029889222582403807, "optim/total_tokens": 1869086720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 5.567406177520752, "created_at": "2025-01-14T23:03:03.991114+00:00"} {"global_step": 7131, "acc_step": 0, "speed/wps": 12968.552285254122, "speed/FLOPS": 203688934298227.44, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.8102123737335205, "optim/lr": 0.0029889118621182504, "optim/total_tokens": 1869348864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 5.503665924072266, "created_at": "2025-01-14T23:03:09.049084+00:00"} {"global_step": 7132, "acc_step": 0, "speed/wps": 12973.218368255182, "speed/FLOPS": 203762221543629.06, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.4723682403564453, "optim/lr": 0.002988901461138283, "optim/total_tokens": 1869611008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456851, "loss/out": 5.551352500915527, "created_at": "2025-01-14T23:03:14.105767+00:00"} {"global_step": 7133, "acc_step": 0, "speed/wps": 12964.487840299762, "speed/FLOPS": 203625096605086.6, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.9525364637374878, "optim/lr": 0.002988891055300512, "optim/total_tokens": 1869873152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 5.578015327453613, "created_at": "2025-01-14T23:03:19.164610+00:00"} {"global_step": 7134, "acc_step": 0, "speed/wps": 12971.689232916508, "speed/FLOPS": 203738204371883.03, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2734119892120361, "optim/lr": 0.0029888806446049733, "optim/total_tokens": 1870135296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 5.4253339767456055, "created_at": "2025-01-14T23:03:24.221626+00:00"} {"global_step": 7135, "acc_step": 0, "speed/wps": 12970.495976131331, "speed/FLOPS": 203719462634365.2, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.93440580368042, "optim/lr": 0.002988870229051699, "optim/total_tokens": 1870397440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449135, "loss/out": 5.637479782104492, "created_at": "2025-01-14T23:03:29.278697+00:00"} {"global_step": 7136, "acc_step": 0, "speed/wps": 12976.223098753102, "speed/FLOPS": 203809414965031.28, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2534815073013306, "optim/lr": 0.0029888598086407242, "optim/total_tokens": 1870659584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429739, "loss/out": 5.52903938293457, "created_at": "2025-01-14T23:03:34.335583+00:00"} {"global_step": 7137, "acc_step": 0, "speed/wps": 12971.190371895415, "speed/FLOPS": 203730369074039.88, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.18593692779541, "optim/lr": 0.0029888493833720824, "optim/total_tokens": 1870921728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292724, "loss/out": 5.648025035858154, "created_at": "2025-01-14T23:03:39.396835+00:00"} {"global_step": 7138, "acc_step": 0, "speed/wps": 12966.435036484185, "speed/FLOPS": 203655679996890.9, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3710362911224365, "optim/lr": 0.0029888389532458075, "optim/total_tokens": 1871183872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 412125, "loss/out": 5.462273120880127, "created_at": "2025-01-14T23:03:44.454713+00:00"} {"global_step": 7139, "acc_step": 0, "speed/wps": 12965.373620752998, "speed/FLOPS": 203639009004294.56, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.305596113204956, "optim/lr": 0.002988828518261934, "optim/total_tokens": 1871446016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447554, "loss/out": 5.586511611938477, "created_at": "2025-01-14T23:03:49.513621+00:00"} {"global_step": 7140, "acc_step": 0, "speed/wps": 12967.587385180072, "speed/FLOPS": 203673779216652.84, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7862097024917603, "optim/lr": 0.0029888180784204948, "optim/total_tokens": 1871708160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 5.580569267272949, "created_at": "2025-01-14T23:03:54.574064+00:00"} {"global_step": 7141, "acc_step": 0, "speed/wps": 12970.614916815863, "speed/FLOPS": 203721330761257.38, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0014, "optim/grad_norm": 5.406080722808838, "optim/lr": 0.0029888076337215254, "optim/total_tokens": 1871970304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440628, "loss/out": 5.594448089599609, "created_at": "2025-01-14T23:03:59.632968+00:00"} {"global_step": 7142, "acc_step": 0, "speed/wps": 12969.609139483691, "speed/FLOPS": 203705533646172.75, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2372074127197266, "optim/lr": 0.002988797184165059, "optim/total_tokens": 1872232448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459237, "loss/out": 5.604651927947998, "created_at": "2025-01-14T23:04:04.724422+00:00"} {"global_step": 7143, "acc_step": 0, "speed/wps": 12965.289256405045, "speed/FLOPS": 203637683946281.2, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2423477172851562, "optim/lr": 0.0029887867297511304, "optim/total_tokens": 1872494592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356961, "loss/out": 5.565342903137207, "created_at": "2025-01-14T23:04:09.787035+00:00"} {"global_step": 7144, "acc_step": 0, "speed/wps": 12969.556911436068, "speed/FLOPS": 203704713332915.97, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.001, "optim/grad_norm": 1.1838089227676392, "optim/lr": 0.002988776270479773, "optim/total_tokens": 1872756736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433179, "loss/out": 5.516989707946777, "created_at": "2025-01-14T23:04:14.843874+00:00"} {"global_step": 7145, "acc_step": 0, "speed/wps": 12973.737704786769, "speed/FLOPS": 203770378437500.8, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1779520511627197, "optim/lr": 0.0029887658063510215, "optim/total_tokens": 1873018880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376530, "loss/out": 5.721333980560303, "created_at": "2025-01-14T23:04:19.903509+00:00"} {"global_step": 7146, "acc_step": 0, "speed/wps": 12970.566025355865, "speed/FLOPS": 203720562853694.22, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.6043992042541504, "optim/lr": 0.0029887553373649095, "optim/total_tokens": 1873281024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426252, "loss/out": 5.429810523986816, "created_at": "2025-01-14T23:04:24.960477+00:00"} {"global_step": 7147, "acc_step": 0, "speed/wps": 12973.112251990113, "speed/FLOPS": 203760554841867.12, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.5049885511398315, "optim/lr": 0.0029887448635214716, "optim/total_tokens": 1873543168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 5.566577911376953, "created_at": "2025-01-14T23:04:30.018948+00:00"} {"global_step": 7148, "acc_step": 0, "speed/wps": 12973.665217425922, "speed/FLOPS": 203769239923890.62, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.621225595474243, "optim/lr": 0.0029887343848207415, "optim/total_tokens": 1873805312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 5.527960300445557, "created_at": "2025-01-14T23:04:35.074118+00:00"} {"global_step": 7149, "acc_step": 0, "speed/wps": 12972.613226508007, "speed/FLOPS": 203752716960937.78, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.510849118232727, "optim/lr": 0.002988723901262754, "optim/total_tokens": 1874067456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425730, "loss/out": 5.572236061096191, "created_at": "2025-01-14T23:04:40.133974+00:00"} {"global_step": 7150, "acc_step": 0, "speed/wps": 12976.657184773623, "speed/FLOPS": 203816232882480.25, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1959024667739868, "optim/lr": 0.0029887134128475427, "optim/total_tokens": 1874329600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 5.451259136199951, "created_at": "2025-01-14T23:04:45.188118+00:00"} {"global_step": 7151, "acc_step": 0, "speed/wps": 12973.93342111911, "speed/FLOPS": 203773452431442.66, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.4073611497879028, "optim/lr": 0.002988702919575142, "optim/total_tokens": 1874591744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466369, "loss/out": 5.506750583648682, "created_at": "2025-01-14T23:04:50.243608+00:00"} {"global_step": 7152, "acc_step": 0, "speed/wps": 12966.932317961331, "speed/FLOPS": 203663490485825.84, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.858096718788147, "optim/lr": 0.0029886924214455866, "optim/total_tokens": 1874853888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444068, "loss/out": 5.5754313468933105, "created_at": "2025-01-14T23:04:55.306264+00:00"} {"global_step": 7153, "acc_step": 0, "speed/wps": 12968.790338915114, "speed/FLOPS": 203692673258093.34, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.625234842300415, "optim/lr": 0.00298868191845891, "optim/total_tokens": 1875116032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317586, "loss/out": 5.5168609619140625, "created_at": "2025-01-14T23:05:00.368950+00:00"} {"global_step": 7154, "acc_step": 0, "speed/wps": 12969.756771031984, "speed/FLOPS": 203707852402505.7, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.3809971809387207, "optim/lr": 0.002988671410615147, "optim/total_tokens": 1875378176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452615, "loss/out": 5.46000862121582, "created_at": "2025-01-14T23:05:05.425698+00:00"} {"global_step": 7155, "acc_step": 0, "speed/wps": 12965.39081338433, "speed/FLOPS": 203639279038194.78, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3285306692123413, "optim/lr": 0.002988660897914332, "optim/total_tokens": 1875640320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479707, "loss/out": 5.562402725219727, "created_at": "2025-01-14T23:05:10.485159+00:00"} {"global_step": 7156, "acc_step": 0, "speed/wps": 12973.90465244297, "speed/FLOPS": 203773000579851.56, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5671664476394653, "optim/lr": 0.002988650380356499, "optim/total_tokens": 1875902464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 5.500586986541748, "created_at": "2025-01-14T23:05:15.541304+00:00"} {"global_step": 7157, "acc_step": 0, "speed/wps": 12976.148384605753, "speed/FLOPS": 203808241476679.16, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.275478482246399, "optim/lr": 0.0029886398579416825, "optim/total_tokens": 1876164608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445976, "loss/out": 5.453908920288086, "created_at": "2025-01-14T23:05:20.609412+00:00"} {"global_step": 7158, "acc_step": 0, "speed/wps": 12976.001768010225, "speed/FLOPS": 203805938661574.1, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.298839807510376, "optim/lr": 0.002988629330669916, "optim/total_tokens": 1876426752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424963, "loss/out": 5.551332473754883, "created_at": "2025-01-14T23:05:25.664620+00:00"} {"global_step": 7159, "acc_step": 0, "speed/wps": 12969.722185759261, "speed/FLOPS": 203707309193272.44, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.80320405960083, "optim/lr": 0.0029886187985412355, "optim/total_tokens": 1876688896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 5.564135551452637, "created_at": "2025-01-14T23:05:30.729903+00:00"} {"global_step": 7160, "acc_step": 0, "speed/wps": 12958.169505917953, "speed/FLOPS": 203525858481319.0, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.001, "optim/grad_norm": 1.420320749282837, "optim/lr": 0.002988608261555673, "optim/total_tokens": 1876951040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330091, "loss/out": 5.4183197021484375, "created_at": "2025-01-14T23:05:35.796714+00:00"} {"global_step": 7161, "acc_step": 0, "speed/wps": 12968.383534663251, "speed/FLOPS": 203686283838310.22, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1616051197052002, "optim/lr": 0.002988597719713265, "optim/total_tokens": 1877213184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377497, "loss/out": 5.66171932220459, "created_at": "2025-01-14T23:05:40.854276+00:00"} {"global_step": 7162, "acc_step": 0, "speed/wps": 12962.128663539397, "speed/FLOPS": 203588042492216.47, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3561431169509888, "optim/lr": 0.0029885871730140454, "optim/total_tokens": 1877475328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442208, "loss/out": 5.5272064208984375, "created_at": "2025-01-14T23:05:45.914948+00:00"} {"global_step": 7163, "acc_step": 0, "speed/wps": 12960.758013509405, "speed/FLOPS": 203566514550024.9, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1349495649337769, "optim/lr": 0.0029885766214580484, "optim/total_tokens": 1877737472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 5.415843486785889, "created_at": "2025-01-14T23:05:50.977980+00:00"} {"global_step": 7164, "acc_step": 0, "speed/wps": 12970.110348377026, "speed/FLOPS": 203713405820575.75, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0410420894622803, "optim/lr": 0.0029885660650453083, "optim/total_tokens": 1877999616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441679, "loss/out": 5.509177207946777, "created_at": "2025-01-14T23:05:56.038802+00:00"} {"global_step": 7165, "acc_step": 0, "speed/wps": 12972.410241725125, "speed/FLOPS": 203749528805992.28, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2838116884231567, "optim/lr": 0.0029885555037758595, "optim/total_tokens": 1878261760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 5.56407356262207, "created_at": "2025-01-14T23:06:01.094381+00:00"} {"global_step": 7166, "acc_step": 0, "speed/wps": 12963.463993640795, "speed/FLOPS": 203609015686394.75, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1698228120803833, "optim/lr": 0.0029885449376497366, "optim/total_tokens": 1878523904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427351, "loss/out": 5.507027626037598, "created_at": "2025-01-14T23:06:06.156667+00:00"} {"global_step": 7167, "acc_step": 0, "speed/wps": 12964.120993776198, "speed/FLOPS": 203619334776334.06, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3695268630981445, "optim/lr": 0.002988534366666974, "optim/total_tokens": 1878786048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289551, "loss/out": 5.355626583099365, "created_at": "2025-01-14T23:06:11.221576+00:00"} {"global_step": 7168, "acc_step": 0, "speed/wps": 12966.41981220871, "speed/FLOPS": 203655440878724.22, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1669225692749023, "optim/lr": 0.0029885237908276064, "optim/total_tokens": 1879048192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447791, "loss/out": 5.4632158279418945, "created_at": "2025-01-14T23:06:16.283264+00:00"} {"global_step": 7169, "acc_step": 0, "speed/wps": 12962.332792757385, "speed/FLOPS": 203591248622087.9, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.9460344314575195, "optim/lr": 0.0029885132101316685, "optim/total_tokens": 1879310336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 5.561367988586426, "created_at": "2025-01-14T23:06:21.342726+00:00"} {"global_step": 7170, "acc_step": 0, "speed/wps": 12965.552278712168, "speed/FLOPS": 203641815073045.1, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1059646606445312, "optim/lr": 0.002988502624579194, "optim/total_tokens": 1879572480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459528, "loss/out": 5.5352301597595215, "created_at": "2025-01-14T23:06:26.400940+00:00"} {"global_step": 7171, "acc_step": 0, "speed/wps": 12964.07754660861, "speed/FLOPS": 203618652378864.44, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7734003067016602, "optim/lr": 0.002988492034170218, "optim/total_tokens": 1879834624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435866, "loss/out": 5.529821872711182, "created_at": "2025-01-14T23:06:31.469604+00:00"} {"global_step": 7172, "acc_step": 0, "speed/wps": 12960.958327733073, "speed/FLOPS": 203569660760168.22, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2905969619750977, "optim/lr": 0.0029884814389047752, "optim/total_tokens": 1880096768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 5.484334945678711, "created_at": "2025-01-14T23:06:36.531816+00:00"} {"global_step": 7173, "acc_step": 0, "speed/wps": 12966.7738288696, "speed/FLOPS": 203661001196853.38, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6547518968582153, "optim/lr": 0.0029884708387829, "optim/total_tokens": 1880358912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419618, "loss/out": 5.609044075012207, "created_at": "2025-01-14T23:06:41.589436+00:00"} {"global_step": 7174, "acc_step": 0, "speed/wps": 12968.090213486552, "speed/FLOPS": 203681676826164.62, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4258334636688232, "optim/lr": 0.0029884602338046268, "optim/total_tokens": 1880621056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310904, "loss/out": 5.519279479980469, "created_at": "2025-01-14T23:06:46.652910+00:00"} {"global_step": 7175, "acc_step": 0, "speed/wps": 12964.301363181226, "speed/FLOPS": 203622167725693.56, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.638505935668945, "optim/lr": 0.00298844962396999, "optim/total_tokens": 1880883200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451523, "loss/out": 5.627704620361328, "created_at": "2025-01-14T23:06:51.715644+00:00"} {"global_step": 7176, "acc_step": 0, "speed/wps": 12965.115462046742, "speed/FLOPS": 203634954267065.56, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.084803581237793, "optim/lr": 0.002988439009279025, "optim/total_tokens": 1881145344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 5.562430381774902, "created_at": "2025-01-14T23:06:56.775239+00:00"} {"global_step": 7177, "acc_step": 0, "speed/wps": 12966.586394773189, "speed/FLOPS": 203658057286807.78, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2795535326004028, "optim/lr": 0.002988428389731766, "optim/total_tokens": 1881407488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 5.554576873779297, "created_at": "2025-01-14T23:07:01.833108+00:00"} {"global_step": 7178, "acc_step": 0, "speed/wps": 12966.814156843586, "speed/FLOPS": 203661634603101.25, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4685449600219727, "optim/lr": 0.0029884177653282475, "optim/total_tokens": 1881669632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438240, "loss/out": 5.515321731567383, "created_at": "2025-01-14T23:07:06.894947+00:00"} {"global_step": 7179, "acc_step": 0, "speed/wps": 12962.255344574816, "speed/FLOPS": 203590032191955.72, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9390161037445068, "optim/lr": 0.0029884071360685043, "optim/total_tokens": 1881931776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455808, "loss/out": 5.5665059089660645, "created_at": "2025-01-14T23:07:11.956280+00:00"} {"global_step": 7180, "acc_step": 0, "speed/wps": 12957.048624044659, "speed/FLOPS": 203508253491245.8, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.185092449188232, "optim/lr": 0.002988396501952571, "optim/total_tokens": 1882193920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320721, "loss/out": 5.506223678588867, "created_at": "2025-01-14T23:07:17.018970+00:00"} {"global_step": 7181, "acc_step": 0, "speed/wps": 12964.761411722857, "speed/FLOPS": 203629393420212.8, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.967989444732666, "optim/lr": 0.0029883858629804823, "optim/total_tokens": 1882456064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363341, "loss/out": 5.617919921875, "created_at": "2025-01-14T23:07:22.087559+00:00"} {"global_step": 7182, "acc_step": 0, "speed/wps": 12963.080694314727, "speed/FLOPS": 203602995443770.4, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.243143916130066, "optim/lr": 0.0029883752191522733, "optim/total_tokens": 1882718208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461100, "loss/out": 5.523999214172363, "created_at": "2025-01-14T23:07:27.146598+00:00"} {"global_step": 7183, "acc_step": 0, "speed/wps": 12967.087269703521, "speed/FLOPS": 203665924215862.94, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1212531328201294, "optim/lr": 0.0029883645704679787, "optim/total_tokens": 1882980352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 5.509479522705078, "created_at": "2025-01-14T23:07:32.204525+00:00"} {"global_step": 7184, "acc_step": 0, "speed/wps": 12956.686082247415, "speed/FLOPS": 203502559274135.47, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2693438529968262, "optim/lr": 0.002988353916927632, "optim/total_tokens": 1883242496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461385, "loss/out": 5.5421552658081055, "created_at": "2025-01-14T23:07:37.275218+00:00"} {"global_step": 7185, "acc_step": 0, "speed/wps": 12966.11190580458, "speed/FLOPS": 203650604785539.62, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3535003662109375, "optim/lr": 0.0029883432585312694, "optim/total_tokens": 1883504640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419618, "loss/out": 5.541325569152832, "created_at": "2025-01-14T23:07:42.334563+00:00"} {"global_step": 7186, "acc_step": 0, "speed/wps": 12966.391693266596, "speed/FLOPS": 203654999232098.84, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1321744918823242, "optim/lr": 0.0029883325952789253, "optim/total_tokens": 1883766784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 5.453032970428467, "created_at": "2025-01-14T23:07:47.394353+00:00"} {"global_step": 7187, "acc_step": 0, "speed/wps": 12967.772683110346, "speed/FLOPS": 203676689575273.84, "speed/curr_iter_time": 1.2722, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.093995451927185, "optim/lr": 0.002988321927170634, "optim/total_tokens": 1884028928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429454, "loss/out": 5.5448832511901855, "created_at": "2025-01-14T23:07:52.457734+00:00"} {"global_step": 7188, "acc_step": 0, "speed/wps": 12964.11518519952, "speed/FLOPS": 203619243544655.5, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3989489078521729, "optim/lr": 0.002988311254206431, "optim/total_tokens": 1884291072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327688, "loss/out": 5.50894832611084, "created_at": "2025-01-14T23:07:57.520869+00:00"} {"global_step": 7189, "acc_step": 0, "speed/wps": 12957.495655563725, "speed/FLOPS": 203515274735538.44, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0959199666976929, "optim/lr": 0.002988300576386351, "optim/total_tokens": 1884553216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440919, "loss/out": 5.569884777069092, "created_at": "2025-01-14T23:08:02.582328+00:00"} {"global_step": 7190, "acc_step": 0, "speed/wps": 12964.163235087002, "speed/FLOPS": 203619998234165.88, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3618804216384888, "optim/lr": 0.0029882898937104285, "optim/total_tokens": 1884815360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 5.424412727355957, "created_at": "2025-01-14T23:08:07.646599+00:00"} {"global_step": 7191, "acc_step": 0, "speed/wps": 12960.971245052368, "speed/FLOPS": 203569863644418.0, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4019687175750732, "optim/lr": 0.0029882792061786986, "optim/total_tokens": 1885077504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441157, "loss/out": 5.473504066467285, "created_at": "2025-01-14T23:08:12.706466+00:00"} {"global_step": 7192, "acc_step": 0, "speed/wps": 12961.428149417685, "speed/FLOPS": 203577039955327.7, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.763302206993103, "optim/lr": 0.0029882685137911963, "optim/total_tokens": 1885339648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 5.390180587768555, "created_at": "2025-01-14T23:08:17.766294+00:00"} {"global_step": 7193, "acc_step": 0, "speed/wps": 12967.23320222855, "speed/FLOPS": 203668216286700.8, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.707115650177002, "optim/lr": 0.0029882578165479554, "optim/total_tokens": 1885601792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447839, "loss/out": 5.465573310852051, "created_at": "2025-01-14T23:08:22.831873+00:00"} {"global_step": 7194, "acc_step": 0, "speed/wps": 12965.993499227912, "speed/FLOPS": 203648745047545.38, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.301658272743225, "optim/lr": 0.002988247114449012, "optim/total_tokens": 1885863936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337583, "loss/out": 5.3861494064331055, "created_at": "2025-01-14T23:08:27.895128+00:00"} {"global_step": 7195, "acc_step": 0, "speed/wps": 12966.307623876213, "speed/FLOPS": 203653678806799.44, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.029651165008545, "optim/lr": 0.002988236407494401, "optim/total_tokens": 1886126080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450756, "loss/out": 5.473532676696777, "created_at": "2025-01-14T23:08:32.954180+00:00"} {"global_step": 7196, "acc_step": 0, "speed/wps": 12965.600336864138, "speed/FLOPS": 203642569892360.6, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.21931791305542, "optim/lr": 0.0029882256956841567, "optim/total_tokens": 1886388224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 5.40777063369751, "created_at": "2025-01-14T23:08:38.012415+00:00"} {"global_step": 7197, "acc_step": 0, "speed/wps": 12965.445064712534, "speed/FLOPS": 203640131129853.34, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0849320888519287, "optim/lr": 0.0029882149790183142, "optim/total_tokens": 1886650368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 5.436701774597168, "created_at": "2025-01-14T23:08:43.071088+00:00"} {"global_step": 7198, "acc_step": 0, "speed/wps": 12964.071609098459, "speed/FLOPS": 203618559122108.5, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.261679768562317, "optim/lr": 0.002988204257496909, "optim/total_tokens": 1886912512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 5.371443271636963, "created_at": "2025-01-14T23:08:48.131979+00:00"} {"global_step": 7199, "acc_step": 0, "speed/wps": 12952.688806142045, "speed/FLOPS": 203439776560067.66, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0018, "optim/grad_norm": 1.3018938302993774, "optim/lr": 0.0029881935311199762, "optim/total_tokens": 1887174656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430791, "loss/out": 5.432765007019043, "created_at": "2025-01-14T23:08:53.222613+00:00"} {"global_step": 7200, "acc_step": 0, "speed/wps": 12959.240367518882, "speed/FLOPS": 203542677834279.62, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3328484296798706, "optim/lr": 0.0029881827998875492, "optim/total_tokens": 1887436800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 285796, "loss/out": 5.475802421569824, "created_at": "2025-01-14T23:08:58.283555+00:00"} {"global_step": 7201, "acc_step": 0, "speed/wps": 12946.177379787132, "speed/FLOPS": 203337505661526.28, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0027, "optim/grad_norm": 1.2872098684310913, "optim/lr": 0.002988172063799665, "optim/total_tokens": 1887698944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469590, "loss/out": 5.5299835205078125, "created_at": "2025-01-14T23:09:03.357518+00:00"} {"global_step": 7202, "acc_step": 0, "speed/wps": 12953.093264865654, "speed/FLOPS": 203446129140106.47, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.834371566772461, "optim/lr": 0.002988161322856357, "optim/total_tokens": 1887961088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447839, "loss/out": 5.453561782836914, "created_at": "2025-01-14T23:09:08.421849+00:00"} {"global_step": 7203, "acc_step": 0, "speed/wps": 12955.923778354318, "speed/FLOPS": 203490586243979.03, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.4897403717041016, "optim/lr": 0.0029881505770576614, "optim/total_tokens": 1888223232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 5.406217098236084, "created_at": "2025-01-14T23:09:13.484202+00:00"} {"global_step": 7204, "acc_step": 0, "speed/wps": 12959.913657047224, "speed/FLOPS": 203553252771519.0, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0011, "optim/grad_norm": 11.715352058410645, "optim/lr": 0.002988139826403613, "optim/total_tokens": 1888485376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 5.472294330596924, "created_at": "2025-01-14T23:09:18.546007+00:00"} {"global_step": 7205, "acc_step": 0, "speed/wps": 12960.049128686922, "speed/FLOPS": 203555380539778.0, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4814672470092773, "optim/lr": 0.0029881290708942466, "optim/total_tokens": 1888747520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326058, "loss/out": 5.4600419998168945, "created_at": "2025-01-14T23:09:23.607103+00:00"} {"global_step": 7206, "acc_step": 0, "speed/wps": 12962.277280491888, "speed/FLOPS": 203590376725676.4, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3158929347991943, "optim/lr": 0.0029881183105295974, "optim/total_tokens": 1889009664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357199, "loss/out": 5.377813816070557, "created_at": "2025-01-14T23:09:28.666571+00:00"} {"global_step": 7207, "acc_step": 0, "speed/wps": 12964.838242528416, "speed/FLOPS": 203630600153590.72, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.33330500125885, "optim/lr": 0.002988107545309701, "optim/total_tokens": 1889271808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433993, "loss/out": 5.454853534698486, "created_at": "2025-01-14T23:09:33.727591+00:00"} {"global_step": 7208, "acc_step": 0, "speed/wps": 12961.530935136183, "speed/FLOPS": 203578654346279.1, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3810491561889648, "optim/lr": 0.002988096775234591, "optim/total_tokens": 1889533952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 5.543233394622803, "created_at": "2025-01-14T23:09:38.789133+00:00"} {"global_step": 7209, "acc_step": 0, "speed/wps": 12952.109977969578, "speed/FLOPS": 203430685268225.78, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.4517698287963867, "optim/lr": 0.002988086000304304, "optim/total_tokens": 1889796096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426755, "loss/out": 5.508371829986572, "created_at": "2025-01-14T23:09:43.855839+00:00"} {"global_step": 7210, "acc_step": 0, "speed/wps": 12940.458542440425, "speed/FLOPS": 203247683462490.22, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 20.866985321044922, "optim/lr": 0.002988075220518875, "optim/total_tokens": 1890058240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461622, "loss/out": 5.454257011413574, "created_at": "2025-01-14T23:09:48.924941+00:00"} {"global_step": 7211, "acc_step": 0, "speed/wps": 12963.020244440864, "speed/FLOPS": 203602045995434.56, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.1138458251953125, "optim/lr": 0.002988064435878339, "optim/total_tokens": 1890320384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462435, "loss/out": 5.520493984222412, "created_at": "2025-01-14T23:09:53.986355+00:00"} {"global_step": 7212, "acc_step": 0, "speed/wps": 12954.98311373035, "speed/FLOPS": 203475811813450.97, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0011, "optim/grad_norm": 1.2956829071044922, "optim/lr": 0.0029880536463827304, "optim/total_tokens": 1890582528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315344, "loss/out": 5.498818397521973, "created_at": "2025-01-14T23:09:59.048959+00:00"} {"global_step": 7213, "acc_step": 0, "speed/wps": 12957.048163587035, "speed/FLOPS": 203508246259126.1, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.5166208744049072, "optim/lr": 0.002988042852032086, "optim/total_tokens": 1890844672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401462, "loss/out": 5.58674955368042, "created_at": "2025-01-14T23:10:04.118064+00:00"} {"global_step": 7214, "acc_step": 0, "speed/wps": 12959.404094990005, "speed/FLOPS": 203545249399198.72, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1294864416122437, "optim/lr": 0.0029880320528264397, "optim/total_tokens": 1891106816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463485, "loss/out": 5.527431488037109, "created_at": "2025-01-14T23:10:09.184332+00:00"} {"global_step": 7215, "acc_step": 0, "speed/wps": 12953.602356981692, "speed/FLOPS": 203454125131354.7, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.599924921989441, "optim/lr": 0.0029880212487658274, "optim/total_tokens": 1891368960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431299, "loss/out": 5.4639081954956055, "created_at": "2025-01-14T23:10:14.254896+00:00"} {"global_step": 7216, "acc_step": 0, "speed/wps": 12957.674659577619, "speed/FLOPS": 203518086239553.9, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.5123509168624878, "optim/lr": 0.0029880104398502835, "optim/total_tokens": 1891631104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450994, "loss/out": 5.495415210723877, "created_at": "2025-01-14T23:10:19.318726+00:00"} {"global_step": 7217, "acc_step": 0, "speed/wps": 12952.368210254155, "speed/FLOPS": 203434741161103.4, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5756114721298218, "optim/lr": 0.0029879996260798445, "optim/total_tokens": 1891893248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319751, "loss/out": 5.450000762939453, "created_at": "2025-01-14T23:10:24.384264+00:00"} {"global_step": 7218, "acc_step": 0, "speed/wps": 12963.201230381068, "speed/FLOPS": 203604888628326.2, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9220361709594727, "optim/lr": 0.002987988807454545, "optim/total_tokens": 1892155392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448368, "loss/out": 5.517208099365234, "created_at": "2025-01-14T23:10:29.445603+00:00"} {"global_step": 7219, "acc_step": 0, "speed/wps": 12965.631278504388, "speed/FLOPS": 203643055873339.0, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1864396333694458, "optim/lr": 0.00298797798397442, "optim/total_tokens": 1892417536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 5.392237186431885, "created_at": "2025-01-14T23:10:34.504795+00:00"} {"global_step": 7220, "acc_step": 0, "speed/wps": 12962.715916522051, "speed/FLOPS": 203597266107279.22, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4795758724212646, "optim/lr": 0.002987967155639505, "optim/total_tokens": 1892679680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435329, "loss/out": 5.431000709533691, "created_at": "2025-01-14T23:10:39.568209+00:00"} {"global_step": 7221, "acc_step": 0, "speed/wps": 12965.572221476132, "speed/FLOPS": 203642128301557.34, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.747562050819397, "optim/lr": 0.002987956322449836, "optim/total_tokens": 1892941824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 5.472772598266602, "created_at": "2025-01-14T23:10:44.636500+00:00"} {"global_step": 7222, "acc_step": 0, "speed/wps": 12966.06594866821, "speed/FLOPS": 203649882965561.22, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5790847539901733, "optim/lr": 0.0029879454844054477, "optim/total_tokens": 1893203968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441442, "loss/out": 5.460744380950928, "created_at": "2025-01-14T23:10:49.694526+00:00"} {"global_step": 7223, "acc_step": 0, "speed/wps": 12962.174334899877, "speed/FLOPS": 203588759823691.8, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.1081143617630005, "optim/lr": 0.0029879346415063756, "optim/total_tokens": 1893466112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291425, "loss/out": 5.36564826965332, "created_at": "2025-01-14T23:10:54.758618+00:00"} {"global_step": 7224, "acc_step": 0, "speed/wps": 12970.153888583784, "speed/FLOPS": 203714089679353.4, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4399908781051636, "optim/lr": 0.0029879237937526547, "optim/total_tokens": 1893728256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334076, "loss/out": 5.373978614807129, "created_at": "2025-01-14T23:10:59.816572+00:00"} {"global_step": 7225, "acc_step": 0, "speed/wps": 12964.832717559331, "speed/FLOPS": 203630513376359.12, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.132796287536621, "optim/lr": 0.002987912941144321, "optim/total_tokens": 1893990400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390139, "loss/out": 5.336524486541748, "created_at": "2025-01-14T23:11:04.883052+00:00"} {"global_step": 7226, "acc_step": 0, "speed/wps": 12968.156541967437, "speed/FLOPS": 203682718606101.8, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.189780592918396, "optim/lr": 0.0029879020836814096, "optim/total_tokens": 1894252544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444641, "loss/out": 5.492183208465576, "created_at": "2025-01-14T23:11:09.944659+00:00"} {"global_step": 7227, "acc_step": 0, "speed/wps": 12962.347816417549, "speed/FLOPS": 203591484589315.44, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3302339315414429, "optim/lr": 0.0029878912213639566, "optim/total_tokens": 1894514688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457124, "loss/out": 5.468387603759766, "created_at": "2025-01-14T23:11:15.005004+00:00"} {"global_step": 7228, "acc_step": 0, "speed/wps": 12961.206435003032, "speed/FLOPS": 203573557625778.03, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6735773086547852, "optim/lr": 0.0029878803541919965, "optim/total_tokens": 1894776832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467205, "loss/out": 5.444411277770996, "created_at": "2025-01-14T23:11:20.065178+00:00"} {"global_step": 7229, "acc_step": 0, "speed/wps": 12964.652641497232, "speed/FLOPS": 203627685034353.44, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7103201150894165, "optim/lr": 0.002987869482165565, "optim/total_tokens": 1895038976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459743, "loss/out": 5.350101470947266, "created_at": "2025-01-14T23:11:25.128880+00:00"} {"global_step": 7230, "acc_step": 0, "speed/wps": 12966.081012041803, "speed/FLOPS": 203650119556542.72, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.215192198753357, "optim/lr": 0.0029878586052846977, "optim/total_tokens": 1895301120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311065, "loss/out": 5.475317001342773, "created_at": "2025-01-14T23:11:30.188039+00:00"} {"global_step": 7231, "acc_step": 0, "speed/wps": 12961.734429548887, "speed/FLOPS": 203581850505661.25, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.316732406616211, "optim/lr": 0.0029878477235494296, "optim/total_tokens": 1895563264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368286, "loss/out": 5.362995147705078, "created_at": "2025-01-14T23:11:35.248143+00:00"} {"global_step": 7232, "acc_step": 0, "speed/wps": 12971.029294509166, "speed/FLOPS": 203727839132345.3, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2738360166549683, "optim/lr": 0.0029878368369597973, "optim/total_tokens": 1895825408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434515, "loss/out": 5.435450553894043, "created_at": "2025-01-14T23:11:40.307765+00:00"} {"global_step": 7233, "acc_step": 0, "speed/wps": 12963.321202708961, "speed/FLOPS": 203606772958595.0, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5063905715942383, "optim/lr": 0.0029878259455158357, "optim/total_tokens": 1896087552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 5.335796356201172, "created_at": "2025-01-14T23:11:45.366894+00:00"} {"global_step": 7234, "acc_step": 0, "speed/wps": 12969.948461122774, "speed/FLOPS": 203710863158791.34, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2535136938095093, "optim/lr": 0.00298781504921758, "optim/total_tokens": 1896349696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418566, "loss/out": 5.535739898681641, "created_at": "2025-01-14T23:11:50.424883+00:00"} {"global_step": 7235, "acc_step": 0, "speed/wps": 12960.561028298278, "speed/FLOPS": 203563420626598.12, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3499008417129517, "optim/lr": 0.002987804148065066, "optim/total_tokens": 1896611840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 5.332585334777832, "created_at": "2025-01-14T23:11:55.490173+00:00"} {"global_step": 7236, "acc_step": 0, "speed/wps": 12966.875358929958, "speed/FLOPS": 203662595865968.22, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2492059469223022, "optim/lr": 0.00298779324205833, "optim/total_tokens": 1896873984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452325, "loss/out": 5.345487117767334, "created_at": "2025-01-14T23:12:00.548144+00:00"} {"global_step": 7237, "acc_step": 0, "speed/wps": 12959.980649002762, "speed/FLOPS": 203554304972237.3, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.5609917640686035, "optim/lr": 0.002987782331197406, "optim/total_tokens": 1897136128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444881, "loss/out": 5.393049716949463, "created_at": "2025-01-14T23:12:05.614369+00:00"} {"global_step": 7238, "acc_step": 0, "speed/wps": 12962.503195866042, "speed/FLOPS": 203593925037067.78, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.811657428741455, "optim/lr": 0.002987771415482331, "optim/total_tokens": 1897398272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290611, "loss/out": 5.460131645202637, "created_at": "2025-01-14T23:12:10.678745+00:00"} {"global_step": 7239, "acc_step": 0, "speed/wps": 12963.186381458412, "speed/FLOPS": 203604655405591.3, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1033697128295898, "optim/lr": 0.0029877604949131403, "optim/total_tokens": 1897660416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 405430, "loss/out": 5.363764762878418, "created_at": "2025-01-14T23:12:15.741834+00:00"} {"global_step": 7240, "acc_step": 0, "speed/wps": 12958.441178517329, "speed/FLOPS": 203530125472809.47, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.6660828590393066, "optim/lr": 0.002987749569489869, "optim/total_tokens": 1897922560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466155, "loss/out": 5.45304012298584, "created_at": "2025-01-14T23:12:20.804728+00:00"} {"global_step": 7241, "acc_step": 0, "speed/wps": 12961.257566873785, "speed/FLOPS": 203574360722070.75, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3884438276290894, "optim/lr": 0.0029877386392125533, "optim/total_tokens": 1898184704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 5.51182746887207, "created_at": "2025-01-14T23:12:25.865857+00:00"} {"global_step": 7242, "acc_step": 0, "speed/wps": 12961.747782944145, "speed/FLOPS": 203582060239083.25, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1702901124954224, "optim/lr": 0.002987727704081229, "optim/total_tokens": 1898446848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437195, "loss/out": 5.446225166320801, "created_at": "2025-01-14T23:12:30.929218+00:00"} {"global_step": 7243, "acc_step": 0, "speed/wps": 12957.763759003035, "speed/FLOPS": 203519485668465.53, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1677272319793701, "optim/lr": 0.0029877167640959307, "optim/total_tokens": 1898708992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452325, "loss/out": 5.488491535186768, "created_at": "2025-01-14T23:12:35.995121+00:00"} {"global_step": 7244, "acc_step": 0, "speed/wps": 12965.739372343904, "speed/FLOPS": 203644753635625.06, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3664095401763916, "optim/lr": 0.0029877058192566953, "optim/total_tokens": 1898971136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294596, "loss/out": 5.400527477264404, "created_at": "2025-01-14T23:12:41.057543+00:00"} {"global_step": 7245, "acc_step": 0, "speed/wps": 12963.902523702758, "speed/FLOPS": 203615903403621.34, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3963841199874878, "optim/lr": 0.0029876948695635583, "optim/total_tokens": 1899233280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419791, "loss/out": 5.296215057373047, "created_at": "2025-01-14T23:12:46.117528+00:00"} {"global_step": 7246, "acc_step": 0, "speed/wps": 12966.04908872123, "speed/FLOPS": 203649618156925.9, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0781468152999878, "optim/lr": 0.0029876839150165544, "optim/total_tokens": 1899495424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469353, "loss/out": 5.414857387542725, "created_at": "2025-01-14T23:12:51.175491+00:00"} {"global_step": 7247, "acc_step": 0, "speed/wps": 12968.697313843537, "speed/FLOPS": 203691212171514.38, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7730003595352173, "optim/lr": 0.002987672955615721, "optim/total_tokens": 1899757568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 5.426417350769043, "created_at": "2025-01-14T23:12:56.237288+00:00"} {"global_step": 7248, "acc_step": 0, "speed/wps": 12971.07741718163, "speed/FLOPS": 203728594965043.84, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0929945707321167, "optim/lr": 0.002987661991361092, "optim/total_tokens": 1900019712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446220, "loss/out": 5.337163925170898, "created_at": "2025-01-14T23:13:01.294481+00:00"} {"global_step": 7249, "acc_step": 0, "speed/wps": 12970.875869562107, "speed/FLOPS": 203725429382723.8, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.895373821258545, "optim/lr": 0.0029876510222527053, "optim/total_tokens": 1900281856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464820, "loss/out": 5.415677070617676, "created_at": "2025-01-14T23:13:06.355689+00:00"} {"global_step": 7250, "acc_step": 0, "speed/wps": 12956.345405016713, "speed/FLOPS": 203497208470087.72, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1371582746505737, "optim/lr": 0.0029876400482905944, "optim/total_tokens": 1900544000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 5.351179122924805, "created_at": "2025-01-14T23:13:11.418094+00:00"} {"global_step": 7251, "acc_step": 0, "speed/wps": 12965.312459119004, "speed/FLOPS": 203638048376787.78, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.2475183010101318, "optim/lr": 0.002987629069474797, "optim/total_tokens": 1900806144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 5.198922157287598, "created_at": "2025-01-14T23:13:16.481617+00:00"} {"global_step": 7252, "acc_step": 0, "speed/wps": 12964.629281109741, "speed/FLOPS": 203627318127367.34, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2370179891586304, "optim/lr": 0.0029876180858053477, "optim/total_tokens": 1901068288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453945, "loss/out": 5.461220741271973, "created_at": "2025-01-14T23:13:21.543571+00:00"} {"global_step": 7253, "acc_step": 0, "speed/wps": 12964.90204396093, "speed/FLOPS": 203631602242760.72, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0327074527740479, "optim/lr": 0.002987607097282283, "optim/total_tokens": 1901330432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305274, "loss/out": 5.356864929199219, "created_at": "2025-01-14T23:13:26.602021+00:00"} {"global_step": 7254, "acc_step": 0, "speed/wps": 12965.284086986418, "speed/FLOPS": 203637602753458.12, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1957097053527832, "optim/lr": 0.0029875961039056388, "optim/total_tokens": 1901592576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364976, "loss/out": 5.385214805603027, "created_at": "2025-01-14T23:13:31.665406+00:00"} {"global_step": 7255, "acc_step": 0, "speed/wps": 12974.679654874959, "speed/FLOPS": 203785173058010.9, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.144190788269043, "optim/lr": 0.0029875851056754507, "optim/total_tokens": 1901854720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451797, "loss/out": 5.376928329467773, "created_at": "2025-01-14T23:13:36.723252+00:00"} {"global_step": 7256, "acc_step": 0, "speed/wps": 12969.414196598496, "speed/FLOPS": 203702471800281.16, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4088228940963745, "optim/lr": 0.0029875741025917548, "optim/total_tokens": 1902116864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 5.353456020355225, "created_at": "2025-01-14T23:13:41.784853+00:00"} {"global_step": 7257, "acc_step": 0, "speed/wps": 12961.876872754483, "speed/FLOPS": 203584087771942.6, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.154752254486084, "optim/lr": 0.0029875630946545858, "optim/total_tokens": 1902379008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429470, "loss/out": 5.50943660736084, "created_at": "2025-01-14T23:13:46.844766+00:00"} {"global_step": 7258, "acc_step": 0, "speed/wps": 12967.013033478928, "speed/FLOPS": 203664758233944.44, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7048084735870361, "optim/lr": 0.0029875520818639814, "optim/total_tokens": 1902641152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459765, "loss/out": 5.4403157234191895, "created_at": "2025-01-14T23:13:51.906708+00:00"} {"global_step": 7259, "acc_step": 0, "speed/wps": 12962.5247481849, "speed/FLOPS": 203594263545851.44, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.727174758911133, "optim/lr": 0.0029875410642199764, "optim/total_tokens": 1902903296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447555, "loss/out": 5.431421279907227, "created_at": "2025-01-14T23:13:56.971664+00:00"} {"global_step": 7260, "acc_step": 0, "speed/wps": 12972.632617031255, "speed/FLOPS": 203753021515750.75, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3762470483779907, "optim/lr": 0.0029875300417226075, "optim/total_tokens": 1903165440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297001, "loss/out": 5.360841274261475, "created_at": "2025-01-14T23:14:02.028923+00:00"} {"global_step": 7261, "acc_step": 0, "speed/wps": 12967.777672893115, "speed/FLOPS": 203676767946668.9, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8821061849594116, "optim/lr": 0.0029875190143719096, "optim/total_tokens": 1903427584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366026, "loss/out": 5.411224842071533, "created_at": "2025-01-14T23:14:07.087821+00:00"} {"global_step": 7262, "acc_step": 0, "speed/wps": 12962.160949585961, "speed/FLOPS": 203588549588943.44, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.5500404834747314, "optim/lr": 0.00298750798216792, "optim/total_tokens": 1903689728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434807, "loss/out": 5.409559726715088, "created_at": "2025-01-14T23:14:12.151257+00:00"} {"global_step": 7263, "acc_step": 0, "speed/wps": 12956.759558525897, "speed/FLOPS": 203503713320056.0, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.9567155838012695, "optim/lr": 0.0029874969451106734, "optim/total_tokens": 1903951872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 5.317777633666992, "created_at": "2025-01-14T23:14:17.216536+00:00"} {"global_step": 7264, "acc_step": 0, "speed/wps": 12969.049394516987, "speed/FLOPS": 203696742082300.47, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.2881755828857422, "optim/lr": 0.002987485903200207, "optim/total_tokens": 1904214016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 5.432241439819336, "created_at": "2025-01-14T23:14:22.276537+00:00"} {"global_step": 7265, "acc_step": 0, "speed/wps": 12962.403326899439, "speed/FLOPS": 203592356457712.94, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.853105306625366, "optim/lr": 0.002987474856436556, "optim/total_tokens": 1904476160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457380, "loss/out": 5.548856258392334, "created_at": "2025-01-14T23:14:27.337374+00:00"} {"global_step": 7266, "acc_step": 0, "speed/wps": 12967.442577362906, "speed/FLOPS": 203671504810897.47, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2802764177322388, "optim/lr": 0.002987463804819757, "optim/total_tokens": 1904738304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455517, "loss/out": 5.347443580627441, "created_at": "2025-01-14T23:14:32.395900+00:00"} {"global_step": 7267, "acc_step": 0, "speed/wps": 12964.253209566708, "speed/FLOPS": 203621411407007.56, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3312872648239136, "optim/lr": 0.0029874527483498454, "optim/total_tokens": 1905000448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 5.388426780700684, "created_at": "2025-01-14T23:14:37.458952+00:00"} {"global_step": 7268, "acc_step": 0, "speed/wps": 12963.994763495384, "speed/FLOPS": 203617352156315.3, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.9521355628967285, "optim/lr": 0.0029874416870268577, "optim/total_tokens": 1905262592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321540, "loss/out": 5.437107086181641, "created_at": "2025-01-14T23:14:42.521610+00:00"} {"global_step": 7269, "acc_step": 0, "speed/wps": 12962.439105242458, "speed/FLOPS": 203592918405754.62, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.6952877044677734, "optim/lr": 0.00298743062085083, "optim/total_tokens": 1905524736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361228, "loss/out": 5.52362060546875, "created_at": "2025-01-14T23:14:47.582995+00:00"} {"global_step": 7270, "acc_step": 0, "speed/wps": 12961.956828167826, "speed/FLOPS": 203585343581579.38, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3169019222259521, "optim/lr": 0.002987419549821798, "optim/total_tokens": 1905786880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451808, "loss/out": 5.362710475921631, "created_at": "2025-01-14T23:14:52.650865+00:00"} {"global_step": 7271, "acc_step": 0, "speed/wps": 12965.712723188068, "speed/FLOPS": 203644335074012.9, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4273475408554077, "optim/lr": 0.002987408473939799, "optim/total_tokens": 1906049024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 5.351127624511719, "created_at": "2025-01-14T23:14:57.711636+00:00"} {"global_step": 7272, "acc_step": 0, "speed/wps": 12964.580537472091, "speed/FLOPS": 203626552541557.28, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.5932984352111816, "optim/lr": 0.0029873973932048677, "optim/total_tokens": 1906311168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443307, "loss/out": 5.413628578186035, "created_at": "2025-01-14T23:15:02.770268+00:00"} {"global_step": 7273, "acc_step": 0, "speed/wps": 12968.414750096463, "speed/FLOPS": 203686774119583.3, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3999451398849487, "optim/lr": 0.002987386307617041, "optim/total_tokens": 1906573312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502126, "loss/out": 5.362174034118652, "created_at": "2025-01-14T23:15:07.832585+00:00"} {"global_step": 7274, "acc_step": 0, "speed/wps": 12965.940246717826, "speed/FLOPS": 203647908643697.84, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6749709844589233, "optim/lr": 0.002987375217176355, "optim/total_tokens": 1906835456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442256, "loss/out": 5.426146507263184, "created_at": "2025-01-14T23:15:12.893813+00:00"} {"global_step": 7275, "acc_step": 0, "speed/wps": 12962.836830260328, "speed/FLOPS": 203599165223691.75, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2980867624282837, "optim/lr": 0.0029873641218828456, "optim/total_tokens": 1907097600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352115, "loss/out": 5.3645339012146, "created_at": "2025-01-14T23:15:17.963363+00:00"} {"global_step": 7276, "acc_step": 0, "speed/wps": 12967.530905493097, "speed/FLOPS": 203672892125557.7, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3740954399108887, "optim/lr": 0.002987353021736549, "optim/total_tokens": 1907359744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437432, "loss/out": 5.379066467285156, "created_at": "2025-01-14T23:15:23.025300+00:00"} {"global_step": 7277, "acc_step": 0, "speed/wps": 12969.622179230753, "speed/FLOPS": 203705738453318.84, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.512369990348816, "optim/lr": 0.002987341916737502, "optim/total_tokens": 1907621888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 5.381120681762695, "created_at": "2025-01-14T23:15:28.086503+00:00"} {"global_step": 7278, "acc_step": 0, "speed/wps": 12965.318779979803, "speed/FLOPS": 203638147654592.4, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5633753538131714, "optim/lr": 0.0029873308068857404, "optim/total_tokens": 1907884032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450225, "loss/out": 5.463040351867676, "created_at": "2025-01-14T23:15:33.148591+00:00"} {"global_step": 7279, "acc_step": 0, "speed/wps": 12957.599199702885, "speed/FLOPS": 203516901038528.5, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2239611148834229, "optim/lr": 0.0029873196921813006, "optim/total_tokens": 1908146176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 5.272077560424805, "created_at": "2025-01-14T23:15:38.210204+00:00"} {"global_step": 7280, "acc_step": 0, "speed/wps": 12963.357912707665, "speed/FLOPS": 203607349539569.47, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.026618242263794, "optim/lr": 0.0029873085726242186, "optim/total_tokens": 1908408320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439871, "loss/out": 5.509817600250244, "created_at": "2025-01-14T23:15:43.274462+00:00"} {"global_step": 7281, "acc_step": 0, "speed/wps": 12965.449976259859, "speed/FLOPS": 203640208272453.25, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.957566261291504, "optim/lr": 0.0029872974482145305, "optim/total_tokens": 1908670464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 5.348294258117676, "created_at": "2025-01-14T23:15:48.336132+00:00"} {"global_step": 7282, "acc_step": 0, "speed/wps": 12965.79791964587, "speed/FLOPS": 203645673201454.88, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.6858490705490112, "optim/lr": 0.0029872863189522735, "optim/total_tokens": 1908932608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329371, "loss/out": 5.347341537475586, "created_at": "2025-01-14T23:15:53.399246+00:00"} {"global_step": 7283, "acc_step": 0, "speed/wps": 12961.931118555605, "speed/FLOPS": 203584939776790.75, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.001, "optim/grad_norm": 1.431720495223999, "optim/lr": 0.002987275184837483, "optim/total_tokens": 1909194752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 346260, "loss/out": 5.4051289558410645, "created_at": "2025-01-14T23:15:58.458894+00:00"} {"global_step": 7284, "acc_step": 0, "speed/wps": 12960.774067160883, "speed/FLOPS": 203566766694681.4, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.74149227142334, "optim/lr": 0.002987264045870196, "optim/total_tokens": 1909456896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435567, "loss/out": 5.358041763305664, "created_at": "2025-01-14T23:16:03.523023+00:00"} {"global_step": 7285, "acc_step": 0, "speed/wps": 12962.802945722342, "speed/FLOPS": 203598633020460.44, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.5582873821258545, "optim/lr": 0.0029872529020504484, "optim/total_tokens": 1909719040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 5.435300350189209, "created_at": "2025-01-14T23:16:08.589253+00:00"} {"global_step": 7286, "acc_step": 0, "speed/wps": 12959.250429424268, "speed/FLOPS": 203542835870330.22, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.9433090686798096, "optim/lr": 0.0029872417533782767, "optim/total_tokens": 1909981184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461622, "loss/out": 5.4210124015808105, "created_at": "2025-01-14T23:16:13.655392+00:00"} {"global_step": 7287, "acc_step": 0, "speed/wps": 12965.296396611584, "speed/FLOPS": 203637796093036.56, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5331233739852905, "optim/lr": 0.0029872305998537166, "optim/total_tokens": 1910243328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454466, "loss/out": 5.3530378341674805, "created_at": "2025-01-14T23:16:18.715828+00:00"} {"global_step": 7288, "acc_step": 0, "speed/wps": 12962.90916801371, "speed/FLOPS": 203600301387511.1, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6348780393600464, "optim/lr": 0.002987219441476806, "optim/total_tokens": 1910505472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 5.486350059509277, "created_at": "2025-01-14T23:16:23.775312+00:00"} {"global_step": 7289, "acc_step": 0, "speed/wps": 12963.045628829777, "speed/FLOPS": 203602444692144.25, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.449175834655762, "optim/lr": 0.00298720827824758, "optim/total_tokens": 1910767616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365889, "loss/out": 5.462472915649414, "created_at": "2025-01-14T23:16:28.836831+00:00"} {"global_step": 7290, "acc_step": 0, "speed/wps": 12961.967001734034, "speed/FLOPS": 203585503371416.62, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.001, "optim/grad_norm": 2.2451846599578857, "optim/lr": 0.0029871971101660755, "optim/total_tokens": 1911029760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438536, "loss/out": 5.507914066314697, "created_at": "2025-01-14T23:16:33.898640+00:00"} {"global_step": 7291, "acc_step": 0, "speed/wps": 12969.428775340202, "speed/FLOPS": 203702700779452.94, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.209399700164795, "optim/lr": 0.002987185937232329, "optim/total_tokens": 1911291904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441158, "loss/out": 5.389034748077393, "created_at": "2025-01-14T23:16:38.957183+00:00"} {"global_step": 7292, "acc_step": 0, "speed/wps": 12967.30661418283, "speed/FLOPS": 203669369322321.38, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.243079662322998, "optim/lr": 0.002987174759446377, "optim/total_tokens": 1911554048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457665, "loss/out": 5.39549446105957, "created_at": "2025-01-14T23:16:44.020078+00:00"} {"global_step": 7293, "acc_step": 0, "speed/wps": 12964.06171644479, "speed/FLOPS": 203618403744388.72, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.344585657119751, "optim/lr": 0.0029871635768082554, "optim/total_tokens": 1911816192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443545, "loss/out": 5.308840274810791, "created_at": "2025-01-14T23:16:49.080342+00:00"} {"global_step": 7294, "acc_step": 0, "speed/wps": 12954.441126343747, "speed/FLOPS": 203467299156775.66, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.4136943817138672, "optim/lr": 0.0029871523893180017, "optim/total_tokens": 1912078336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 5.435118675231934, "created_at": "2025-01-14T23:16:54.145981+00:00"} {"global_step": 7295, "acc_step": 0, "speed/wps": 12962.37576500706, "speed/FLOPS": 203591923560319.12, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.725292444229126, "optim/lr": 0.0029871411969756515, "optim/total_tokens": 1912340480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450746, "loss/out": 5.313562870025635, "created_at": "2025-01-14T23:16:59.213557+00:00"} {"global_step": 7296, "acc_step": 0, "speed/wps": 12963.379752055465, "speed/FLOPS": 203607692556537.1, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.269812822341919, "optim/lr": 0.002987129999781241, "optim/total_tokens": 1912602624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305105, "loss/out": 5.377199649810791, "created_at": "2025-01-14T23:17:04.281824+00:00"} {"global_step": 7297, "acc_step": 0, "speed/wps": 12961.971211078586, "speed/FLOPS": 203585569484957.06, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.539604663848877, "optim/lr": 0.0029871187977348087, "optim/total_tokens": 1912864768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448605, "loss/out": 5.3792619705200195, "created_at": "2025-01-14T23:17:09.342381+00:00"} {"global_step": 7298, "acc_step": 0, "speed/wps": 12956.805731227147, "speed/FLOPS": 203504438525777.06, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0006, "optim/grad_norm": 11.624746322631836, "optim/lr": 0.002987107590836389, "optim/total_tokens": 1913126912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 5.268982410430908, "created_at": "2025-01-14T23:17:14.408134+00:00"} {"global_step": 7299, "acc_step": 0, "speed/wps": 12956.279655319266, "speed/FLOPS": 203496175780739.75, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3913042545318604, "optim/lr": 0.0029870963790860194, "optim/total_tokens": 1913389056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445404, "loss/out": 5.458294868469238, "created_at": "2025-01-14T23:17:19.475401+00:00"} {"global_step": 7300, "acc_step": 0, "speed/wps": 12966.389072278716, "speed/FLOPS": 203654958065882.6, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2416167259216309, "optim/lr": 0.0029870851624837365, "optim/total_tokens": 1913651200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460840, "loss/out": 5.357506275177002, "created_at": "2025-01-14T23:17:24.535055+00:00"} {"global_step": 7301, "acc_step": 0, "speed/wps": 12965.396183552804, "speed/FLOPS": 203639363384070.03, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2397316694259644, "optim/lr": 0.002987073941029576, "optim/total_tokens": 1913913344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322066, "loss/out": 5.422533988952637, "created_at": "2025-01-14T23:17:29.594354+00:00"} {"global_step": 7302, "acc_step": 0, "speed/wps": 12962.465977333917, "speed/FLOPS": 203593340468877.12, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6414759159088135, "optim/lr": 0.0029870627147235763, "optim/total_tokens": 1914175488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 344264, "loss/out": 5.392087936401367, "created_at": "2025-01-14T23:17:34.655513+00:00"} {"global_step": 7303, "acc_step": 0, "speed/wps": 12966.277061278253, "speed/FLOPS": 203653198779201.06, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7135871648788452, "optim/lr": 0.002987051483565772, "optim/total_tokens": 1914437632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422788, "loss/out": 5.311691761016846, "created_at": "2025-01-14T23:17:39.716965+00:00"} {"global_step": 7304, "acc_step": 0, "speed/wps": 12969.009229512856, "speed/FLOPS": 203696111235718.1, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6454519033432007, "optim/lr": 0.0029870402475562015, "optim/total_tokens": 1914699776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444067, "loss/out": 5.415414810180664, "created_at": "2025-01-14T23:17:44.779655+00:00"} {"global_step": 7305, "acc_step": 0, "speed/wps": 12965.440490001325, "speed/FLOPS": 203640059277727.5, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.718672275543213, "optim/lr": 0.0029870290066949007, "optim/total_tokens": 1914961920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 5.379012107849121, "created_at": "2025-01-14T23:17:49.842065+00:00"} {"global_step": 7306, "acc_step": 0, "speed/wps": 12958.146970649505, "speed/FLOPS": 203525504533961.3, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2917144298553467, "optim/lr": 0.002987017760981906, "optim/total_tokens": 1915224064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434807, "loss/out": 5.4867119789123535, "created_at": "2025-01-14T23:17:54.908632+00:00"} {"global_step": 7307, "acc_step": 0, "speed/wps": 12968.303810636507, "speed/FLOPS": 203685031662917.9, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5518152713775635, "optim/lr": 0.0029870065104172544, "optim/total_tokens": 1915486208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 5.335127830505371, "created_at": "2025-01-14T23:17:59.967053+00:00"} {"global_step": 7308, "acc_step": 0, "speed/wps": 12961.936022665815, "speed/FLOPS": 203585016802580.56, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.605217218399048, "optim/lr": 0.0029869952550009827, "optim/total_tokens": 1915748352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442784, "loss/out": 5.545159339904785, "created_at": "2025-01-14T23:18:05.030619+00:00"} {"global_step": 7309, "acc_step": 0, "speed/wps": 12960.652194085667, "speed/FLOPS": 203564852510564.0, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.68670654296875, "optim/lr": 0.002986983994733128, "optim/total_tokens": 1916010496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291617, "loss/out": 5.505169868469238, "created_at": "2025-01-14T23:18:10.120225+00:00"} {"global_step": 7310, "acc_step": 0, "speed/wps": 12965.675352440749, "speed/FLOPS": 203643748115073.47, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4106719493865967, "optim/lr": 0.002986972729613725, "optim/total_tokens": 1916272640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 5.429234504699707, "created_at": "2025-01-14T23:18:15.179652+00:00"} {"global_step": 7311, "acc_step": 0, "speed/wps": 12961.71467765244, "speed/FLOPS": 203581540274985.6, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.0072994232177734, "optim/lr": 0.002986961459642813, "optim/total_tokens": 1916534784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453660, "loss/out": 5.347455024719238, "created_at": "2025-01-14T23:18:20.243835+00:00"} {"global_step": 7312, "acc_step": 0, "speed/wps": 12963.82719427851, "speed/FLOPS": 203614720251500.06, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.425863742828369, "optim/lr": 0.002986950184820428, "optim/total_tokens": 1916796928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468018, "loss/out": 5.481111526489258, "created_at": "2025-01-14T23:18:25.304923+00:00"} {"global_step": 7313, "acc_step": 0, "speed/wps": 12971.433562836171, "speed/FLOPS": 203734188721943.84, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.6049957275390625, "optim/lr": 0.0029869389051466063, "optim/total_tokens": 1917059072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460578, "loss/out": 5.375959396362305, "created_at": "2025-01-14T23:18:30.360896+00:00"} {"global_step": 7314, "acc_step": 0, "speed/wps": 12961.770614854633, "speed/FLOPS": 203582418845611.78, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5874828100204468, "optim/lr": 0.002986927620621385, "optim/total_tokens": 1917321216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437432, "loss/out": 5.453564643859863, "created_at": "2025-01-14T23:18:35.424389+00:00"} {"global_step": 7315, "acc_step": 0, "speed/wps": 12966.353704575955, "speed/FLOPS": 203654402567510.4, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1830748319625854, "optim/lr": 0.0029869163312448003, "optim/total_tokens": 1917583360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306907, "loss/out": 5.326276779174805, "created_at": "2025-01-14T23:18:40.482115+00:00"} {"global_step": 7316, "acc_step": 0, "speed/wps": 12975.627610890151, "speed/FLOPS": 203800062009857.56, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8247835636138916, "optim/lr": 0.00298690503701689, "optim/total_tokens": 1917845504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353744, "loss/out": 5.409732341766357, "created_at": "2025-01-14T23:18:45.542392+00:00"} {"global_step": 7317, "acc_step": 0, "speed/wps": 12971.049533866584, "speed/FLOPS": 203728157019265.38, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4673222303390503, "optim/lr": 0.002986893737937691, "optim/total_tokens": 1918107648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 405960, "loss/out": 5.40082311630249, "created_at": "2025-01-14T23:18:50.599294+00:00"} {"global_step": 7318, "acc_step": 0, "speed/wps": 12970.896832881386, "speed/FLOPS": 203725758640459.7, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.336620330810547, "optim/lr": 0.002986882434007239, "optim/total_tokens": 1918369792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441147, "loss/out": 5.470193386077881, "created_at": "2025-01-14T23:18:55.656682+00:00"} {"global_step": 7319, "acc_step": 0, "speed/wps": 12970.299220752764, "speed/FLOPS": 203716372320774.56, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.3018670082092285, "optim/lr": 0.002986871125225572, "optim/total_tokens": 1918631936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422290, "loss/out": 5.506631851196289, "created_at": "2025-01-14T23:19:00.716839+00:00"} {"global_step": 7320, "acc_step": 0, "speed/wps": 12970.255397472733, "speed/FLOPS": 203715684015942.16, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.319810152053833, "optim/lr": 0.0029868598115927263, "optim/total_tokens": 1918894080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 5.369786262512207, "created_at": "2025-01-14T23:19:05.773325+00:00"} {"global_step": 7321, "acc_step": 0, "speed/wps": 12973.803038586708, "speed/FLOPS": 203771404594606.9, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.757640838623047, "optim/lr": 0.0029868484931087395, "optim/total_tokens": 1919156224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456330, "loss/out": 5.39647912979126, "created_at": "2025-01-14T23:19:10.829569+00:00"} {"global_step": 7322, "acc_step": 0, "speed/wps": 12958.961399299555, "speed/FLOPS": 203538296255052.6, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.417057752609253, "optim/lr": 0.002986837169773647, "optim/total_tokens": 1919418368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470403, "loss/out": 5.452578544616699, "created_at": "2025-01-14T23:19:15.893146+00:00"} {"global_step": 7323, "acc_step": 0, "speed/wps": 12970.299655855277, "speed/FLOPS": 203716379154657.4, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.397523880004883, "optim/lr": 0.002986825841587488, "optim/total_tokens": 1919680512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 5.512103080749512, "created_at": "2025-01-14T23:19:20.950857+00:00"} {"global_step": 7324, "acc_step": 0, "speed/wps": 12967.864951697195, "speed/FLOPS": 203678138780220.62, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.3128653764724731, "optim/lr": 0.0029868145085502973, "optim/total_tokens": 1919942656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449411, "loss/out": 5.5066399574279785, "created_at": "2025-01-14T23:19:26.011230+00:00"} {"global_step": 7325, "acc_step": 0, "speed/wps": 12965.282701468263, "speed/FLOPS": 203637580991991.6, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5230587720870972, "optim/lr": 0.0029868031706621135, "optim/total_tokens": 1920204800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 282328, "loss/out": 5.416294574737549, "created_at": "2025-01-14T23:19:31.075184+00:00"} {"global_step": 7326, "acc_step": 0, "speed/wps": 12967.421242869565, "speed/FLOPS": 203671169723362.88, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.8844172954559326, "optim/lr": 0.0029867918279229723, "optim/total_tokens": 1920466944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444357, "loss/out": 5.402256011962891, "created_at": "2025-01-14T23:19:36.134050+00:00"} {"global_step": 7327, "acc_step": 0, "speed/wps": 12959.803092119002, "speed/FLOPS": 203551516197389.16, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5981923341751099, "optim/lr": 0.0029867804803329115, "optim/total_tokens": 1920729088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 5.376399993896484, "created_at": "2025-01-14T23:19:41.197006+00:00"} {"global_step": 7328, "acc_step": 0, "speed/wps": 12960.780720187657, "speed/FLOPS": 203566871189609.12, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7443199157714844, "optim/lr": 0.002986769127891968, "optim/total_tokens": 1920991232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429454, "loss/out": 5.488397598266602, "created_at": "2025-01-14T23:19:46.259262+00:00"} {"global_step": 7329, "acc_step": 0, "speed/wps": 12963.762963983772, "speed/FLOPS": 203613711426459.53, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.7853267192840576, "optim/lr": 0.0029867577706001783, "optim/total_tokens": 1921253376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471454, "loss/out": 5.369353771209717, "created_at": "2025-01-14T23:19:51.319644+00:00"} {"global_step": 7330, "acc_step": 0, "speed/wps": 12961.79311259017, "speed/FLOPS": 203582772203463.53, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.05036735534668, "optim/lr": 0.002986746408457581, "optim/total_tokens": 1921515520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 5.4988813400268555, "created_at": "2025-01-14T23:19:56.381425+00:00"} {"global_step": 7331, "acc_step": 0, "speed/wps": 12962.105032891237, "speed/FLOPS": 203587671340415.2, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9454801082611084, "optim/lr": 0.0029867350414642113, "optim/total_tokens": 1921777664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 5.471714973449707, "created_at": "2025-01-14T23:20:01.441540+00:00"} {"global_step": 7332, "acc_step": 0, "speed/wps": 12967.572602676706, "speed/FLOPS": 203673547037124.2, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 136.95228576660156, "optim/lr": 0.002986723669620107, "optim/total_tokens": 1922039808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461622, "loss/out": 5.323751449584961, "created_at": "2025-01-14T23:20:06.499994+00:00"} {"global_step": 7333, "acc_step": 0, "speed/wps": 12961.515650751537, "speed/FLOPS": 203578414284015.28, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.0256128311157227, "optim/lr": 0.002986712292925306, "optim/total_tokens": 1922301952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322066, "loss/out": 5.39422607421875, "created_at": "2025-01-14T23:20:11.559932+00:00"} {"global_step": 7334, "acc_step": 0, "speed/wps": 12966.796575746706, "speed/FLOPS": 203661358467815.8, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.192768573760986, "optim/lr": 0.0029867009113798437, "optim/total_tokens": 1922564096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360307, "loss/out": 5.385455131530762, "created_at": "2025-01-14T23:20:16.617580+00:00"} {"global_step": 7335, "acc_step": 0, "speed/wps": 12967.39713172671, "speed/FLOPS": 203670791024732.16, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.8831708431243896, "optim/lr": 0.002986689524983759, "optim/total_tokens": 1922826240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451559, "loss/out": 5.403919696807861, "created_at": "2025-01-14T23:20:21.678236+00:00"} {"global_step": 7336, "acc_step": 0, "speed/wps": 12952.491291297069, "speed/FLOPS": 203436674318013.44, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.0551517009735107, "optim/lr": 0.0029866781337370882, "optim/total_tokens": 1923088384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480143, "loss/out": 5.5191779136657715, "created_at": "2025-01-14T23:20:26.745133+00:00"} {"global_step": 7337, "acc_step": 0, "speed/wps": 12966.978152209218, "speed/FLOPS": 203664210375671.75, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.663019895553589, "optim/lr": 0.0029866667376398687, "optim/total_tokens": 1923350528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433975, "loss/out": 5.367874622344971, "created_at": "2025-01-14T23:20:31.803937+00:00"} {"global_step": 7338, "acc_step": 0, "speed/wps": 12969.563341613743, "speed/FLOPS": 203704814327692.3, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.5920238494873047, "optim/lr": 0.0029866553366921372, "optim/total_tokens": 1923612672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494531, "loss/out": 5.401561737060547, "created_at": "2025-01-14T23:20:36.864474+00:00"} {"global_step": 7339, "acc_step": 0, "speed/wps": 12962.5936941188, "speed/FLOPS": 203595346436484.97, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.2878196239471436, "optim/lr": 0.002986643930893931, "optim/total_tokens": 1923874816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429454, "loss/out": 5.368775844573975, "created_at": "2025-01-14T23:20:41.923882+00:00"} {"global_step": 7340, "acc_step": 0, "speed/wps": 12963.966884350753, "speed/FLOPS": 203616914276038.9, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0163185596466064, "optim/lr": 0.002986632520245288, "optim/total_tokens": 1924136960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297882, "loss/out": 5.467898368835449, "created_at": "2025-01-14T23:20:46.984005+00:00"} {"global_step": 7341, "acc_step": 0, "speed/wps": 12968.133661265427, "speed/FLOPS": 203682359233235.34, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.344994306564331, "optim/lr": 0.002986621104746245, "optim/total_tokens": 1924399104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440089, "loss/out": 5.406775951385498, "created_at": "2025-01-14T23:20:52.042598+00:00"} {"global_step": 7342, "acc_step": 0, "speed/wps": 12961.423832992732, "speed/FLOPS": 203576972159942.4, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.042487144470215, "optim/lr": 0.002986609684396839, "optim/total_tokens": 1924661248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466392, "loss/out": 5.387287616729736, "created_at": "2025-01-14T23:20:57.107163+00:00"} {"global_step": 7343, "acc_step": 0, "speed/wps": 12961.924418983586, "speed/FLOPS": 203584834550806.03, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0013, "optim/grad_norm": 3.07383131980896, "optim/lr": 0.0029865982591971074, "optim/total_tokens": 1924923392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455517, "loss/out": 5.344220161437988, "created_at": "2025-01-14T23:21:02.166877+00:00"} {"global_step": 7344, "acc_step": 0, "speed/wps": 12962.974858451984, "speed/FLOPS": 203601333146112.3, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.3805317878723145, "optim/lr": 0.002986586829147088, "optim/total_tokens": 1925185536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433718, "loss/out": 5.488454818725586, "created_at": "2025-01-14T23:21:07.226557+00:00"} {"global_step": 7345, "acc_step": 0, "speed/wps": 12958.77422201361, "speed/FLOPS": 203535356378568.28, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.783374309539795, "optim/lr": 0.0029865753942468175, "optim/total_tokens": 1925447680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450756, "loss/out": 5.467511177062988, "created_at": "2025-01-14T23:21:12.287517+00:00"} {"global_step": 7346, "acc_step": 0, "speed/wps": 12961.859851795418, "speed/FLOPS": 203583820434391.22, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8115205764770508, "optim/lr": 0.0029865639544963333, "optim/total_tokens": 1925709824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 5.369933605194092, "created_at": "2025-01-14T23:21:17.347714+00:00"} {"global_step": 7347, "acc_step": 0, "speed/wps": 12959.394650989592, "speed/FLOPS": 203545101068194.84, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.354817271232605, "optim/lr": 0.002986552509895673, "optim/total_tokens": 1925971968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438531, "loss/out": 5.412689208984375, "created_at": "2025-01-14T23:21:22.434599+00:00"} {"global_step": 7348, "acc_step": 0, "speed/wps": 12962.543941413744, "speed/FLOPS": 203594565001884.7, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.7297698259353638, "optim/lr": 0.0029865410604448736, "optim/total_tokens": 1926234112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469875, "loss/out": 5.3065619468688965, "created_at": "2025-01-14T23:21:27.494608+00:00"} {"global_step": 7349, "acc_step": 0, "speed/wps": 12960.579573584699, "speed/FLOPS": 203563711905805.25, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.836961269378662, "optim/lr": 0.002986529606143972, "optim/total_tokens": 1926496256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469353, "loss/out": 5.393802642822266, "created_at": "2025-01-14T23:21:32.558800+00:00"} {"global_step": 7350, "acc_step": 0, "speed/wps": 12960.71549360638, "speed/FLOPS": 203565846716519.44, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.166463017463684, "optim/lr": 0.002986518146993007, "optim/total_tokens": 1926758400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445404, "loss/out": 5.460574150085449, "created_at": "2025-01-14T23:21:37.623001+00:00"} {"global_step": 7351, "acc_step": 0, "speed/wps": 12963.798888352921, "speed/FLOPS": 203614275668042.28, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6551593542099, "optim/lr": 0.002986506682992015, "optim/total_tokens": 1927020544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 5.544224739074707, "created_at": "2025-01-14T23:21:42.681831+00:00"} {"global_step": 7352, "acc_step": 0, "speed/wps": 12961.811200606526, "speed/FLOPS": 203583056300615.94, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.4549193382263184, "optim/lr": 0.002986495214141033, "optim/total_tokens": 1927282688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325239, "loss/out": 5.446378231048584, "created_at": "2025-01-14T23:21:47.741599+00:00"} {"global_step": 7353, "acc_step": 0, "speed/wps": 12952.773423826342, "speed/FLOPS": 203441105597076.4, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.2057788372039795, "optim/lr": 0.0029864837404400993, "optim/total_tokens": 1927544832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 406246, "loss/out": 5.439464569091797, "created_at": "2025-01-14T23:21:52.805162+00:00"} {"global_step": 7354, "acc_step": 0, "speed/wps": 12960.663623617373, "speed/FLOPS": 203565032027065.25, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7983901500701904, "optim/lr": 0.0029864722618892507, "optim/total_tokens": 1927806976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465870, "loss/out": 5.286473274230957, "created_at": "2025-01-14T23:21:57.869332+00:00"} {"global_step": 7355, "acc_step": 0, "speed/wps": 12959.70465385301, "speed/FLOPS": 203549970089153.4, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 77.44064331054688, "optim/lr": 0.0029864607784885256, "optim/total_tokens": 1928069120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444878, "loss/out": 5.37592077255249, "created_at": "2025-01-14T23:22:02.932140+00:00"} {"global_step": 7356, "acc_step": 0, "speed/wps": 12961.432289177503, "speed/FLOPS": 203577104975944.25, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.9309279918670654, "optim/lr": 0.00298644929023796, "optim/total_tokens": 1928331264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471976, "loss/out": 5.484566688537598, "created_at": "2025-01-14T23:22:07.995158+00:00"} {"global_step": 7357, "acc_step": 0, "speed/wps": 12959.061451502947, "speed/FLOPS": 203539867712393.0, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0389821529388428, "optim/lr": 0.002986437797137593, "optim/total_tokens": 1928593408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 5.293607234954834, "created_at": "2025-01-14T23:22:13.059677+00:00"} {"global_step": 7358, "acc_step": 0, "speed/wps": 12941.027143256602, "speed/FLOPS": 203256614119647.7, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 22.764188766479492, "optim/lr": 0.002986426299187461, "optim/total_tokens": 1928855552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328221, "loss/out": 5.349545955657959, "created_at": "2025-01-14T23:22:18.132123+00:00"} {"global_step": 7359, "acc_step": 0, "speed/wps": 12965.386221989369, "speed/FLOPS": 203639206924027.66, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.9184153079986572, "optim/lr": 0.002986414796387602, "optim/total_tokens": 1929117696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358542, "loss/out": 5.705007553100586, "created_at": "2025-01-14T23:22:23.190447+00:00"} {"global_step": 7360, "acc_step": 0, "speed/wps": 12968.207891202595, "speed/FLOPS": 203683525116401.22, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.8516688346862793, "optim/lr": 0.0029864032887380533, "optim/total_tokens": 1929379840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438536, "loss/out": 5.496340274810791, "created_at": "2025-01-14T23:22:28.252267+00:00"} {"global_step": 7361, "acc_step": 0, "speed/wps": 12956.615412304152, "speed/FLOPS": 203501449305565.84, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.39881706237793, "optim/lr": 0.002986391776238852, "optim/total_tokens": 1929641984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 5.431609153747559, "created_at": "2025-01-14T23:22:33.315036+00:00"} {"global_step": 7362, "acc_step": 0, "speed/wps": 12957.415220730518, "speed/FLOPS": 203514011395953.84, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.877526760101318, "optim/lr": 0.0029863802588900364, "optim/total_tokens": 1929904128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434515, "loss/out": 5.444324970245361, "created_at": "2025-01-14T23:22:38.384503+00:00"} {"global_step": 7363, "acc_step": 0, "speed/wps": 12962.097535664094, "speed/FLOPS": 203587553586160.66, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.706486940383911, "optim/lr": 0.002986368736691644, "optim/total_tokens": 1930166272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448890, "loss/out": 5.5075297355651855, "created_at": "2025-01-14T23:22:43.449496+00:00"} {"global_step": 7364, "acc_step": 0, "speed/wps": 12957.445557849349, "speed/FLOPS": 203514487882092.66, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0013, "optim/grad_norm": 3.3935446739196777, "optim/lr": 0.0029863572096437124, "optim/total_tokens": 1930428416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 283675, "loss/out": 5.347832202911377, "created_at": "2025-01-14T23:22:48.516084+00:00"} {"global_step": 7365, "acc_step": 0, "speed/wps": 12959.078085653717, "speed/FLOPS": 203540128974588.56, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.368288993835449, "optim/lr": 0.0029863456777462784, "optim/total_tokens": 1930690560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431855, "loss/out": 5.427527904510498, "created_at": "2025-01-14T23:22:53.581281+00:00"} {"global_step": 7366, "acc_step": 0, "speed/wps": 12962.829282040302, "speed/FLOPS": 203599046668523.88, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.001, "optim/grad_norm": 1.6983031034469604, "optim/lr": 0.002986334140999381, "optim/total_tokens": 1930952704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 5.358465194702148, "created_at": "2025-01-14T23:22:58.640903+00:00"} {"global_step": 7367, "acc_step": 0, "speed/wps": 12955.146917297869, "speed/FLOPS": 203478384573568.56, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.8726816177368164, "optim/lr": 0.0029863225994030563, "optim/total_tokens": 1931214848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455808, "loss/out": 5.244224548339844, "created_at": "2025-01-14T23:23:03.705516+00:00"} {"global_step": 7368, "acc_step": 0, "speed/wps": 12960.782780311793, "speed/FLOPS": 203566903546689.56, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.7308156490325928, "optim/lr": 0.002986311052957343, "optim/total_tokens": 1931476992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476746, "loss/out": 5.4062275886535645, "created_at": "2025-01-14T23:23:08.771099+00:00"} {"global_step": 7369, "acc_step": 0, "speed/wps": 12959.176905799673, "speed/FLOPS": 203541681080773.9, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.246127128601074, "optim/lr": 0.002986299501662279, "optim/total_tokens": 1931739136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445933, "loss/out": 5.400371551513672, "created_at": "2025-01-14T23:23:13.832932+00:00"} {"global_step": 7370, "acc_step": 0, "speed/wps": 12958.511577246567, "speed/FLOPS": 203531231181590.12, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.120079278945923, "optim/lr": 0.002986287945517901, "optim/total_tokens": 1932001280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337010, "loss/out": 5.41521692276001, "created_at": "2025-01-14T23:23:18.895033+00:00"} {"global_step": 7371, "acc_step": 0, "speed/wps": 12960.177776497196, "speed/FLOPS": 203557401130418.97, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.682956218719482, "optim/lr": 0.0029862763845242475, "optim/total_tokens": 1932263424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294654, "loss/out": 5.4887824058532715, "created_at": "2025-01-14T23:23:23.962995+00:00"} {"global_step": 7372, "acc_step": 0, "speed/wps": 12961.18227295907, "speed/FLOPS": 203573178127675.47, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.48031759262085, "optim/lr": 0.002986264818681356, "optim/total_tokens": 1932525568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457920, "loss/out": 5.372808456420898, "created_at": "2025-01-14T23:23:29.027411+00:00"} {"global_step": 7373, "acc_step": 0, "speed/wps": 12962.486760659307, "speed/FLOPS": 203593666899561.66, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.156399965286255, "optim/lr": 0.002986253247989264, "optim/total_tokens": 1932787712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 5.4588236808776855, "created_at": "2025-01-14T23:23:34.095452+00:00"} {"global_step": 7374, "acc_step": 0, "speed/wps": 12962.7189515162, "speed/FLOPS": 203597313776032.84, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.4690580368042, "optim/lr": 0.0029862416724480098, "optim/total_tokens": 1933049856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427351, "loss/out": 5.388510704040527, "created_at": "2025-01-14T23:23:39.156038+00:00"} {"global_step": 7375, "acc_step": 0, "speed/wps": 12962.215742159091, "speed/FLOPS": 203589410181597.3, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.48592209815979, "optim/lr": 0.0029862300920576305, "optim/total_tokens": 1933312000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 5.332876682281494, "created_at": "2025-01-14T23:23:44.219579+00:00"} {"global_step": 7376, "acc_step": 0, "speed/wps": 12960.604814738144, "speed/FLOPS": 203564108352805.1, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.9395523071289062, "optim/lr": 0.002986218506818164, "optim/total_tokens": 1933574144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456567, "loss/out": 5.40633487701416, "created_at": "2025-01-14T23:23:49.280918+00:00"} {"global_step": 7377, "acc_step": 0, "speed/wps": 12959.220912000406, "speed/FLOPS": 203542372258627.28, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 52.478572845458984, "optim/lr": 0.0029862069167296483, "optim/total_tokens": 1933836288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290025, "loss/out": 5.6109466552734375, "created_at": "2025-01-14T23:23:54.342234+00:00"} {"global_step": 7378, "acc_step": 0, "speed/wps": 12961.865885169958, "speed/FLOPS": 203583915196829.1, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 17.3173770904541, "optim/lr": 0.0029861953217921212, "optim/total_tokens": 1934098432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366844, "loss/out": 5.3992767333984375, "created_at": "2025-01-14T23:23:59.406506+00:00"} {"global_step": 7379, "acc_step": 0, "speed/wps": 12957.52361186488, "speed/FLOPS": 203515713827664.06, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.6907694339752197, "optim/lr": 0.002986183722005621, "optim/total_tokens": 1934360576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449942, "loss/out": 5.466418743133545, "created_at": "2025-01-14T23:24:04.493575+00:00"} {"global_step": 7380, "acc_step": 0, "speed/wps": 12963.279719132264, "speed/FLOPS": 203606121402018.56, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.319155693054199, "optim/lr": 0.0029861721173701844, "optim/total_tokens": 1934622720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 5.399044990539551, "created_at": "2025-01-14T23:24:09.557261+00:00"} {"global_step": 7381, "acc_step": 0, "speed/wps": 12954.624131301356, "speed/FLOPS": 203470173501110.1, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0013, "optim/grad_norm": 7.445300102233887, "optim/lr": 0.0029861605078858504, "optim/total_tokens": 1934884864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 5.458286285400391, "created_at": "2025-01-14T23:24:14.623744+00:00"} {"global_step": 7382, "acc_step": 0, "speed/wps": 12959.794599871077, "speed/FLOPS": 203551382814965.9, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.2704122066497803, "optim/lr": 0.002986148893552656, "optim/total_tokens": 1935147008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334076, "loss/out": 5.534417152404785, "created_at": "2025-01-14T23:24:19.685252+00:00"} {"global_step": 7383, "acc_step": 0, "speed/wps": 12962.805443477164, "speed/FLOPS": 203598672251132.2, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 43.62034225463867, "optim/lr": 0.00298613727437064, "optim/total_tokens": 1935409152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 391242, "loss/out": 5.371455669403076, "created_at": "2025-01-14T23:24:24.744596+00:00"} {"global_step": 7384, "acc_step": 0, "speed/wps": 12961.085375019895, "speed/FLOPS": 203571656212389.03, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 30.95120620727539, "optim/lr": 0.0029861256503398393, "optim/total_tokens": 1935671296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424102, "loss/out": 5.452703952789307, "created_at": "2025-01-14T23:24:29.808840+00:00"} {"global_step": 7385, "acc_step": 0, "speed/wps": 12965.856701602073, "speed/FLOPS": 203646596452851.94, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.964690923690796, "optim/lr": 0.0029861140214602925, "optim/total_tokens": 1935933440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432941, "loss/out": 5.562403202056885, "created_at": "2025-01-14T23:24:34.872581+00:00"} {"global_step": 7386, "acc_step": 0, "speed/wps": 12955.666792510468, "speed/FLOPS": 203486549928165.84, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.087217330932617, "optim/lr": 0.0029861023877320375, "optim/total_tokens": 1936195584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 5.384699821472168, "created_at": "2025-01-14T23:24:39.934784+00:00"} {"global_step": 7387, "acc_step": 0, "speed/wps": 12964.362598873206, "speed/FLOPS": 203623129516382.75, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.696624279022217, "optim/lr": 0.0029860907491551123, "optim/total_tokens": 1936457728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425492, "loss/out": 5.444614410400391, "created_at": "2025-01-14T23:24:44.999990+00:00"} {"global_step": 7388, "acc_step": 0, "speed/wps": 12961.690821701117, "speed/FLOPS": 203581165584488.53, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3498449325561523, "optim/lr": 0.002986079105729554, "optim/total_tokens": 1936719872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 5.4039812088012695, "created_at": "2025-01-14T23:24:50.063578+00:00"} {"global_step": 7389, "acc_step": 0, "speed/wps": 12967.678934217849, "speed/FLOPS": 203675217120092.7, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.851333737373352, "optim/lr": 0.002986067457455402, "optim/total_tokens": 1936982016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306620, "loss/out": 5.48696231842041, "created_at": "2025-01-14T23:24:55.123991+00:00"} {"global_step": 7390, "acc_step": 0, "speed/wps": 12965.17171764827, "speed/FLOPS": 203635837838590.88, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.116925001144409, "optim/lr": 0.0029860558043326934, "optim/total_tokens": 1937244160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359599, "loss/out": 5.306657791137695, "created_at": "2025-01-14T23:25:00.183150+00:00"} {"global_step": 7391, "acc_step": 0, "speed/wps": 12966.74054736937, "speed/FLOPS": 203660478465158.7, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.127647876739502, "optim/lr": 0.0029860441463614665, "optim/total_tokens": 1937506304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442208, "loss/out": 5.358368873596191, "created_at": "2025-01-14T23:25:05.251167+00:00"} {"global_step": 7392, "acc_step": 0, "speed/wps": 12970.147158993388, "speed/FLOPS": 203713983981888.8, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.551545143127441, "optim/lr": 0.002986032483541759, "optim/total_tokens": 1937768448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457095, "loss/out": 5.287550926208496, "created_at": "2025-01-14T23:25:10.307693+00:00"} {"global_step": 7393, "acc_step": 0, "speed/wps": 12968.943822086594, "speed/FLOPS": 203695083922209.4, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.014327526092529, "optim/lr": 0.0029860208158736094, "optim/total_tokens": 1938030592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430246, "loss/out": 5.357381820678711, "created_at": "2025-01-14T23:25:15.366961+00:00"} {"global_step": 7394, "acc_step": 0, "speed/wps": 12969.231144031552, "speed/FLOPS": 203699596708176.7, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 11.568202018737793, "optim/lr": 0.002986009143357056, "optim/total_tokens": 1938292736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 5.4033613204956055, "created_at": "2025-01-14T23:25:20.424377+00:00"} {"global_step": 7395, "acc_step": 0, "speed/wps": 12969.410008712202, "speed/FLOPS": 203702406023772.16, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 15.853814125061035, "optim/lr": 0.0029859974659921357, "optim/total_tokens": 1938554880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452081, "loss/out": 5.510375022888184, "created_at": "2025-01-14T23:25:25.481648+00:00"} {"global_step": 7396, "acc_step": 0, "speed/wps": 12969.583530978836, "speed/FLOPS": 203705131429414.22, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.501768589019775, "optim/lr": 0.0029859857837788877, "optim/total_tokens": 1938817024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 5.467944145202637, "created_at": "2025-01-14T23:25:30.542979+00:00"} {"global_step": 7397, "acc_step": 0, "speed/wps": 12971.535763663725, "speed/FLOPS": 203735793926379.22, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.7801268100738525, "optim/lr": 0.0029859740967173496, "optim/total_tokens": 1939079168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294941, "loss/out": 5.474876880645752, "created_at": "2025-01-14T23:25:35.605055+00:00"} {"global_step": 7398, "acc_step": 0, "speed/wps": 12970.212425557003, "speed/FLOPS": 203715009082956.03, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.024057149887085, "optim/lr": 0.00298596240480756, "optim/total_tokens": 1939341312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314655, "loss/out": 5.385068893432617, "created_at": "2025-01-14T23:25:40.661806+00:00"} {"global_step": 7399, "acc_step": 0, "speed/wps": 12972.22797982214, "speed/FLOPS": 203746666132351.8, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7477961778640747, "optim/lr": 0.0029859507080495567, "optim/total_tokens": 1939603456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355325, "loss/out": 5.3017497062683105, "created_at": "2025-01-14T23:25:45.722708+00:00"} {"global_step": 7400, "acc_step": 0, "speed/wps": 12966.251778912816, "speed/FLOPS": 203652801684910.25, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0009, "optim/grad_norm": 11.952603340148926, "optim/lr": 0.0029859390064433775, "optim/total_tokens": 1939865600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435338, "loss/out": 5.452315330505371, "created_at": "2025-01-14T23:25:50.781043+00:00"} {"global_step": 7401, "acc_step": 0, "speed/wps": 12968.065153475412, "speed/FLOPS": 203681283224253.62, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.371059894561768, "optim/lr": 0.002985927299989061, "optim/total_tokens": 1940127744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 5.371294975280762, "created_at": "2025-01-14T23:25:55.858018+00:00"} {"global_step": 7402, "acc_step": 0, "speed/wps": 12974.024624430578, "speed/FLOPS": 203774884904775.75, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0006, "optim/grad_norm": 116.44271087646484, "optim/lr": 0.0029859155886866463, "optim/total_tokens": 1940389888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 5.541915416717529, "created_at": "2025-01-14T23:26:00.915886+00:00"} {"global_step": 7403, "acc_step": 0, "speed/wps": 12969.40404241311, "speed/FLOPS": 203702312314846.38, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.839601516723633, "optim/lr": 0.00298590387253617, "optim/total_tokens": 1940652032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443306, "loss/out": 5.501791000366211, "created_at": "2025-01-14T23:26:05.973804+00:00"} {"global_step": 7404, "acc_step": 0, "speed/wps": 12971.39153715247, "speed/FLOPS": 203733528650831.97, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.2731571197509766, "optim/lr": 0.002985892151537671, "optim/total_tokens": 1940914176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321779, "loss/out": 5.498569488525391, "created_at": "2025-01-14T23:26:11.030248+00:00"} {"global_step": 7405, "acc_step": 0, "speed/wps": 12970.031071670053, "speed/FLOPS": 203712160670955.12, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 19.166954040527344, "optim/lr": 0.0029858804256911874, "optim/total_tokens": 1941176320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 346260, "loss/out": 5.446682929992676, "created_at": "2025-01-14T23:26:16.092009+00:00"} {"global_step": 7406, "acc_step": 0, "speed/wps": 12955.114017441412, "speed/FLOPS": 203477867836113.75, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.869268417358398, "optim/lr": 0.002985868694996758, "optim/total_tokens": 1941438464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461100, "loss/out": 5.510772228240967, "created_at": "2025-01-14T23:26:21.157745+00:00"} {"global_step": 7407, "acc_step": 0, "speed/wps": 12968.794820385801, "speed/FLOPS": 203692743645748.7, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.03045129776001, "optim/lr": 0.0029858569594544207, "optim/total_tokens": 1941700608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446220, "loss/out": 5.627346992492676, "created_at": "2025-01-14T23:26:26.216495+00:00"} {"global_step": 7408, "acc_step": 0, "speed/wps": 12952.964559767694, "speed/FLOPS": 203444107649683.66, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.860747814178467, "optim/lr": 0.002985845219064214, "optim/total_tokens": 1941962752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 5.623744964599609, "created_at": "2025-01-14T23:26:31.285498+00:00"} {"global_step": 7409, "acc_step": 0, "speed/wps": 12956.659756571658, "speed/FLOPS": 203502145793222.44, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 20.02647590637207, "optim/lr": 0.0029858334738261756, "optim/total_tokens": 1942224896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438773, "loss/out": 5.517734050750732, "created_at": "2025-01-14T23:26:36.349163+00:00"} {"global_step": 7410, "acc_step": 0, "speed/wps": 12950.709273832415, "speed/FLOPS": 203408685284982.47, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.299695014953613, "optim/lr": 0.002985821723740344, "optim/total_tokens": 1942487040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322311, "loss/out": 5.541585445404053, "created_at": "2025-01-14T23:26:41.413825+00:00"} {"global_step": 7411, "acc_step": 0, "speed/wps": 12961.304982029555, "speed/FLOPS": 203575105442247.38, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.983015537261963, "optim/lr": 0.002985809968806758, "optim/total_tokens": 1942749184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303402, "loss/out": 5.451406002044678, "created_at": "2025-01-14T23:26:46.479736+00:00"} {"global_step": 7412, "acc_step": 0, "speed/wps": 12954.970878536433, "speed/FLOPS": 203475619642917.6, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0234949588775635, "optim/lr": 0.002985798209025456, "optim/total_tokens": 1943011328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419297, "loss/out": 5.477736949920654, "created_at": "2025-01-14T23:26:51.545714+00:00"} {"global_step": 7413, "acc_step": 0, "speed/wps": 12967.670195064387, "speed/FLOPS": 203675079859678.72, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.108990669250488, "optim/lr": 0.0029857864443964756, "optim/total_tokens": 1943273472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451559, "loss/out": 5.472397327423096, "created_at": "2025-01-14T23:26:56.603491+00:00"} {"global_step": 7414, "acc_step": 0, "speed/wps": 12968.609925842824, "speed/FLOPS": 203689839622879.5, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.239057540893555, "optim/lr": 0.002985774674919856, "optim/total_tokens": 1943535616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455270, "loss/out": 5.461009979248047, "created_at": "2025-01-14T23:27:01.661478+00:00"} {"global_step": 7415, "acc_step": 0, "speed/wps": 12957.265197832932, "speed/FLOPS": 203511655080194.22, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.362704277038574, "optim/lr": 0.002985762900595635, "optim/total_tokens": 1943797760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460815, "loss/out": 5.4223713874816895, "created_at": "2025-01-14T23:27:06.724162+00:00"} {"global_step": 7416, "acc_step": 0, "speed/wps": 12968.978972072578, "speed/FLOPS": 203695636001040.44, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.210545539855957, "optim/lr": 0.0029857511214238515, "optim/total_tokens": 1944059904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 5.7273759841918945, "created_at": "2025-01-14T23:27:11.786262+00:00"} {"global_step": 7417, "acc_step": 0, "speed/wps": 12964.675901647683, "speed/FLOPS": 203628050366979.06, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.958425045013428, "optim/lr": 0.002985739337404544, "optim/total_tokens": 1944322048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322976, "loss/out": 5.53877067565918, "created_at": "2025-01-14T23:27:16.847681+00:00"} {"global_step": 7418, "acc_step": 0, "speed/wps": 12965.65513701792, "speed/FLOPS": 203643430604079.03, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.993860721588135, "optim/lr": 0.00298572754853775, "optim/total_tokens": 1944584192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 400880, "loss/out": 5.539461612701416, "created_at": "2025-01-14T23:27:21.906163+00:00"} {"global_step": 7419, "acc_step": 0, "speed/wps": 12975.991782964868, "speed/FLOPS": 203805781832715.84, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.290193796157837, "optim/lr": 0.0029857157548235087, "optim/total_tokens": 1944846336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432941, "loss/out": 5.346620082855225, "created_at": "2025-01-14T23:27:26.965747+00:00"} {"global_step": 7420, "acc_step": 0, "speed/wps": 12973.918557167575, "speed/FLOPS": 203773218972658.6, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.1912245750427246, "optim/lr": 0.002985703956261859, "optim/total_tokens": 1945108480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 5.443910121917725, "created_at": "2025-01-14T23:27:32.021380+00:00"} {"global_step": 7421, "acc_step": 0, "speed/wps": 12965.660041557516, "speed/FLOPS": 203643507636612.94, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 47.814998626708984, "optim/lr": 0.0029856921528528383, "optim/total_tokens": 1945370624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 5.558246612548828, "created_at": "2025-01-14T23:27:37.084389+00:00"} {"global_step": 7422, "acc_step": 0, "speed/wps": 12971.02312588119, "speed/FLOPS": 203727742245566.38, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0011, "optim/grad_norm": 9.224117279052734, "optim/lr": 0.002985680344596486, "optim/total_tokens": 1945632768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 5.686197280883789, "created_at": "2025-01-14T23:27:42.144748+00:00"} {"global_step": 7423, "acc_step": 0, "speed/wps": 12978.209431283532, "speed/FLOPS": 203840613046928.62, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 32.97087860107422, "optim/lr": 0.0029856685314928405, "optim/total_tokens": 1945894912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478574, "loss/out": 5.5617756843566895, "created_at": "2025-01-14T23:27:47.198184+00:00"} {"global_step": 7424, "acc_step": 0, "speed/wps": 12970.931070564968, "speed/FLOPS": 203726296390327.84, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.715388298034668, "optim/lr": 0.00298565671354194, "optim/total_tokens": 1946157056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484902, "loss/out": 5.667401313781738, "created_at": "2025-01-14T23:27:52.259731+00:00"} {"global_step": 7425, "acc_step": 0, "speed/wps": 12973.267944624185, "speed/FLOPS": 203763000208629.34, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0281565189361572, "optim/lr": 0.0029856448907438233, "optim/total_tokens": 1946419200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449411, "loss/out": 5.620444297790527, "created_at": "2025-01-14T23:27:57.322276+00:00"} {"global_step": 7426, "acc_step": 0, "speed/wps": 12974.13489153966, "speed/FLOPS": 203776616801246.75, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6798930168151855, "optim/lr": 0.0029856330630985283, "optim/total_tokens": 1946681344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322976, "loss/out": 5.540046691894531, "created_at": "2025-01-14T23:28:02.382396+00:00"} {"global_step": 7427, "acc_step": 0, "speed/wps": 12982.819078123346, "speed/FLOPS": 203913013884866.44, "speed/curr_iter_time": 1.2714, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.393572807312012, "optim/lr": 0.002985621230606095, "optim/total_tokens": 1946943488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356847, "loss/out": 5.509327411651611, "created_at": "2025-01-14T23:28:07.434161+00:00"} {"global_step": 7428, "acc_step": 0, "speed/wps": 12978.503648457674, "speed/FLOPS": 203845234131945.28, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.218503713607788, "optim/lr": 0.0029856093932665606, "optim/total_tokens": 1947205632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 404138, "loss/out": 5.592516899108887, "created_at": "2025-01-14T23:28:12.487786+00:00"} {"global_step": 7429, "acc_step": 0, "speed/wps": 12971.665074814275, "speed/FLOPS": 203737824935690.75, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.329986572265625, "optim/lr": 0.002985597551079964, "optim/total_tokens": 1947467776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440105, "loss/out": 5.600816249847412, "created_at": "2025-01-14T23:28:17.544103+00:00"} {"global_step": 7430, "acc_step": 0, "speed/wps": 12972.686531982694, "speed/FLOPS": 203753868324150.0, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.81587028503418, "optim/lr": 0.0029855857040463453, "optim/total_tokens": 1947729920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 5.762581825256348, "created_at": "2025-01-14T23:28:22.604942+00:00"} {"global_step": 7431, "acc_step": 0, "speed/wps": 12959.126071957497, "speed/FLOPS": 203540882665429.5, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.951469898223877, "optim/lr": 0.0029855738521657412, "optim/total_tokens": 1947992064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431605, "loss/out": 5.792361259460449, "created_at": "2025-01-14T23:28:27.667175+00:00"} {"global_step": 7432, "acc_step": 0, "speed/wps": 12976.703343851117, "speed/FLOPS": 203816957874221.53, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 17.37419319152832, "optim/lr": 0.002985561995438192, "optim/total_tokens": 1948254208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 5.706894397735596, "created_at": "2025-01-14T23:28:32.723127+00:00"} {"global_step": 7433, "acc_step": 0, "speed/wps": 12969.514412663735, "speed/FLOPS": 203704045831297.03, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0011, "optim/grad_norm": 2.7624404430389404, "optim/lr": 0.0029855501338637348, "optim/total_tokens": 1948516352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449135, "loss/out": 5.762584686279297, "created_at": "2025-01-14T23:28:37.779905+00:00"} {"global_step": 7434, "acc_step": 0, "speed/wps": 12974.674743509266, "speed/FLOPS": 203785095918263.75, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.234883308410645, "optim/lr": 0.002985538267442409, "optim/total_tokens": 1948778496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323417, "loss/out": 5.676836967468262, "created_at": "2025-01-14T23:28:42.838539+00:00"} {"global_step": 7435, "acc_step": 0, "speed/wps": 12971.094809047658, "speed/FLOPS": 203728868128198.75, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 141.4965057373047, "optim/lr": 0.0029855263961742536, "optim/total_tokens": 1949040640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353219, "loss/out": 5.643858909606934, "created_at": "2025-01-14T23:28:47.898599+00:00"} {"global_step": 7436, "acc_step": 0, "speed/wps": 12975.527106528401, "speed/FLOPS": 203798483450748.8, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 75.57241821289062, "optim/lr": 0.002985514520059307, "optim/total_tokens": 1949302784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425994, "loss/out": 5.9550347328186035, "created_at": "2025-01-14T23:28:52.953484+00:00"} {"global_step": 7437, "acc_step": 0, "speed/wps": 12971.910555258857, "speed/FLOPS": 203741680543398.4, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.9693284034729004, "optim/lr": 0.0029855026390976083, "optim/total_tokens": 1949564928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423841, "loss/out": 6.123045921325684, "created_at": "2025-01-14T23:28:58.010590+00:00"} {"global_step": 7438, "acc_step": 0, "speed/wps": 12975.398208840968, "speed/FLOPS": 203796458935443.7, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 98.84365844726562, "optim/lr": 0.002985490753289196, "optim/total_tokens": 1949827072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447263, "loss/out": 5.753554344177246, "created_at": "2025-01-14T23:29:03.066238+00:00"} {"global_step": 7439, "acc_step": 0, "speed/wps": 12966.497673333293, "speed/FLOPS": 203656663794678.2, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.787864685058594, "optim/lr": 0.002985478862634109, "optim/total_tokens": 1950089216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457095, "loss/out": 5.949311256408691, "created_at": "2025-01-14T23:29:08.129340+00:00"} {"global_step": 7440, "acc_step": 0, "speed/wps": 12972.25095389026, "speed/FLOPS": 203747026971661.3, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.8079833984375, "optim/lr": 0.002985466967132386, "optim/total_tokens": 1950351360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450225, "loss/out": 6.0141496658325195, "created_at": "2025-01-14T23:29:13.191781+00:00"} {"global_step": 7441, "acc_step": 0, "speed/wps": 12977.499146006763, "speed/FLOPS": 203829457040622.72, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0009, "optim/grad_norm": 5.773390293121338, "optim/lr": 0.0029854550667840654, "optim/total_tokens": 1950613504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428093, "loss/out": 5.80673885345459, "created_at": "2025-01-14T23:29:18.275025+00:00"} {"global_step": 7442, "acc_step": 0, "speed/wps": 12976.724840515275, "speed/FLOPS": 203817295508871.62, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.134782791137695, "optim/lr": 0.0029854431615891864, "optim/total_tokens": 1950875648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330862, "loss/out": 5.85575008392334, "created_at": "2025-01-14T23:29:23.332642+00:00"} {"global_step": 7443, "acc_step": 0, "speed/wps": 12974.366906623756, "speed/FLOPS": 203780260916964.62, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.973118782043457, "optim/lr": 0.002985431251547788, "optim/total_tokens": 1951137792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 344678, "loss/out": 5.602435111999512, "created_at": "2025-01-14T23:29:28.388588+00:00"} {"global_step": 7444, "acc_step": 0, "speed/wps": 12963.300010731766, "speed/FLOPS": 203606440109472.3, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0009, "optim/grad_norm": 6.743805408477783, "optim/lr": 0.002985419336659909, "optim/total_tokens": 1951399936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435338, "loss/out": 5.728074073791504, "created_at": "2025-01-14T23:29:33.452201+00:00"} {"global_step": 7445, "acc_step": 0, "speed/wps": 12969.44724561799, "speed/FLOPS": 203702990880546.4, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.559069633483887, "optim/lr": 0.0029854074169255885, "optim/total_tokens": 1951662080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 5.735568046569824, "created_at": "2025-01-14T23:29:38.509438+00:00"} {"global_step": 7446, "acc_step": 0, "speed/wps": 12971.501759508332, "speed/FLOPS": 203735259844392.16, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.901594161987305, "optim/lr": 0.0029853954923448646, "optim/total_tokens": 1951924224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427589, "loss/out": 5.610686302185059, "created_at": "2025-01-14T23:29:43.566921+00:00"} {"global_step": 7447, "acc_step": 0, "speed/wps": 12968.17161497871, "speed/FLOPS": 203682955348456.34, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0009, "optim/grad_norm": 11.742870330810547, "optim/lr": 0.002985383562917777, "optim/total_tokens": 1952186368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 5.725394248962402, "created_at": "2025-01-14T23:29:48.628350+00:00"} {"global_step": 7448, "acc_step": 0, "speed/wps": 12967.008667933665, "speed/FLOPS": 203664689667057.22, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 30.711467742919922, "optim/lr": 0.002985371628644364, "optim/total_tokens": 1952448512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445407, "loss/out": 5.653071403503418, "created_at": "2025-01-14T23:29:53.694230+00:00"} {"global_step": 7449, "acc_step": 0, "speed/wps": 12968.98699354311, "speed/FLOPS": 203695761989257.8, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.7643866539001465, "optim/lr": 0.002985359689524665, "optim/total_tokens": 1952710656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303156, "loss/out": 5.68101692199707, "created_at": "2025-01-14T23:29:58.753570+00:00"} {"global_step": 7450, "acc_step": 0, "speed/wps": 12971.377647408259, "speed/FLOPS": 203733310493312.62, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 33.98570251464844, "optim/lr": 0.0029853477455587195, "optim/total_tokens": 1952972800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322156, "loss/out": 5.663825988769531, "created_at": "2025-01-14T23:30:03.809819+00:00"} {"global_step": 7451, "acc_step": 0, "speed/wps": 12970.567882985531, "speed/FLOPS": 203720592030320.78, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 42.45341110229492, "optim/lr": 0.0029853357967465654, "optim/total_tokens": 1953234944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359599, "loss/out": 5.686408996582031, "created_at": "2025-01-14T23:30:08.867182+00:00"} {"global_step": 7452, "acc_step": 0, "speed/wps": 12970.665974716383, "speed/FLOPS": 203722132695745.97, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 26.20671844482422, "optim/lr": 0.002985323843088242, "optim/total_tokens": 1953497088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451523, "loss/out": 5.596199989318848, "created_at": "2025-01-14T23:30:13.923447+00:00"} {"global_step": 7453, "acc_step": 0, "speed/wps": 12964.832106062826, "speed/FLOPS": 203630503771966.25, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 8.57413387298584, "optim/lr": 0.002985311884583788, "optim/total_tokens": 1953759232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 5.668376922607422, "created_at": "2025-01-14T23:30:18.991018+00:00"} {"global_step": 7454, "acc_step": 0, "speed/wps": 12968.291731903402, "speed/FLOPS": 203684841949816.56, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.802019119262695, "optim/lr": 0.002985299921233243, "optim/total_tokens": 1954021376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440628, "loss/out": 5.623923301696777, "created_at": "2025-01-14T23:30:24.048874+00:00"} {"global_step": 7455, "acc_step": 0, "speed/wps": 12966.672171375027, "speed/FLOPS": 203659404526209.06, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.422577857971191, "optim/lr": 0.0029852879530366462, "optim/total_tokens": 1954283520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 5.571192264556885, "created_at": "2025-01-14T23:30:29.109167+00:00"} {"global_step": 7456, "acc_step": 0, "speed/wps": 12972.77495787373, "speed/FLOPS": 203755257174278.62, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.134782075881958, "optim/lr": 0.002985275979994036, "optim/total_tokens": 1954545664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 5.591486930847168, "created_at": "2025-01-14T23:30:34.167159+00:00"} {"global_step": 7457, "acc_step": 0, "speed/wps": 12960.971476614232, "speed/FLOPS": 203569867281415.28, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.314786911010742, "optim/lr": 0.0029852640021054517, "optim/total_tokens": 1954807808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326583, "loss/out": 5.537095069885254, "created_at": "2025-01-14T23:30:39.228219+00:00"} {"global_step": 7458, "acc_step": 0, "speed/wps": 12968.019519458237, "speed/FLOPS": 203680566479306.16, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.489231109619141, "optim/lr": 0.0029852520193709327, "optim/total_tokens": 1955069952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324848, "loss/out": 5.652744293212891, "created_at": "2025-01-14T23:30:44.288338+00:00"} {"global_step": 7459, "acc_step": 0, "speed/wps": 12965.609878302308, "speed/FLOPS": 203642719753758.34, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.324440002441406, "optim/lr": 0.0029852400317905174, "optim/total_tokens": 1955332096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 5.636384010314941, "created_at": "2025-01-14T23:30:49.351129+00:00"} {"global_step": 7460, "acc_step": 0, "speed/wps": 12966.15421504082, "speed/FLOPS": 203651269310233.75, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.001, "optim/grad_norm": 17.76806640625, "optim/lr": 0.0029852280393642453, "optim/total_tokens": 1955594240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 5.5154571533203125, "created_at": "2025-01-14T23:30:54.409027+00:00"} {"global_step": 7461, "acc_step": 0, "speed/wps": 12960.110837402817, "speed/FLOPS": 203556349759957.78, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.599226951599121, "optim/lr": 0.002985216042092156, "optim/total_tokens": 1955856384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434515, "loss/out": 5.453352928161621, "created_at": "2025-01-14T23:30:59.470040+00:00"} {"global_step": 7462, "acc_step": 0, "speed/wps": 12965.412946246975, "speed/FLOPS": 203639626665216.34, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.449875593185425, "optim/lr": 0.0029852040399742878, "optim/total_tokens": 1956118528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466392, "loss/out": 5.541454315185547, "created_at": "2025-01-14T23:31:04.531926+00:00"} {"global_step": 7463, "acc_step": 0, "speed/wps": 12965.890958075075, "speed/FLOPS": 203647134497833.75, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.523784160614014, "optim/lr": 0.0029851920330106804, "optim/total_tokens": 1956380672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460840, "loss/out": 5.514321804046631, "created_at": "2025-01-14T23:31:09.590006+00:00"} {"global_step": 7464, "acc_step": 0, "speed/wps": 12967.608856431945, "speed/FLOPS": 203674116452168.12, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0009, "optim/grad_norm": 5.625888824462891, "optim/lr": 0.0029851800212013727, "optim/total_tokens": 1956642816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 5.54653263092041, "created_at": "2025-01-14T23:31:14.648005+00:00"} {"global_step": 7465, "acc_step": 0, "speed/wps": 12967.451634029492, "speed/FLOPS": 203671647058291.38, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.8086671829223633, "optim/lr": 0.0029851680045464043, "optim/total_tokens": 1956904960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423342, "loss/out": 5.612151145935059, "created_at": "2025-01-14T23:31:19.709787+00:00"} {"global_step": 7466, "acc_step": 0, "speed/wps": 12973.435396825778, "speed/FLOPS": 203765630275558.78, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.752406597137451, "optim/lr": 0.0029851559830458138, "optim/total_tokens": 1957167104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326058, "loss/out": 5.456830978393555, "created_at": "2025-01-14T23:31:24.766283+00:00"} {"global_step": 7467, "acc_step": 0, "speed/wps": 12976.20327396034, "speed/FLOPS": 203809103589418.84, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.011488914489746, "optim/lr": 0.002985143956699641, "optim/total_tokens": 1957429248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376727, "loss/out": 5.525973320007324, "created_at": "2025-01-14T23:31:29.820490+00:00"} {"global_step": 7468, "acc_step": 0, "speed/wps": 12968.77790075883, "speed/FLOPS": 203692477899757.1, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.206763744354248, "optim/lr": 0.002985131925507925, "optim/total_tokens": 1957691392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451797, "loss/out": 5.579329490661621, "created_at": "2025-01-14T23:31:34.878085+00:00"} {"global_step": 7469, "acc_step": 0, "speed/wps": 12975.00823440284, "speed/FLOPS": 203790333851011.84, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0008, "optim/grad_norm": 785.5844116210938, "optim/lr": 0.0029851198894707046, "optim/total_tokens": 1957953536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463485, "loss/out": 5.455691814422607, "created_at": "2025-01-14T23:31:39.933254+00:00"} {"global_step": 7470, "acc_step": 0, "speed/wps": 12963.863993121893, "speed/FLOPS": 203615298227902.28, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.532998085021973, "optim/lr": 0.002985107848588019, "optim/total_tokens": 1958215680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 5.604801177978516, "created_at": "2025-01-14T23:31:44.999649+00:00"} {"global_step": 7471, "acc_step": 0, "speed/wps": 12967.550218694782, "speed/FLOPS": 203673195465929.25, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 28.79227066040039, "optim/lr": 0.0029850958028599093, "optim/total_tokens": 1958477824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435320, "loss/out": 5.444853782653809, "created_at": "2025-01-14T23:31:50.058742+00:00"} {"global_step": 7472, "acc_step": 0, "speed/wps": 12980.523169414677, "speed/FLOPS": 203876953483687.62, "speed/curr_iter_time": 1.2717, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.515989065170288, "optim/lr": 0.0029850837522864123, "optim/total_tokens": 1958739968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455517, "loss/out": 5.510678768157959, "created_at": "2025-01-14T23:31:55.116404+00:00"} {"global_step": 7473, "acc_step": 0, "speed/wps": 12971.498208701458, "speed/FLOPS": 203735204074090.9, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 19.458419799804688, "optim/lr": 0.0029850716968675686, "optim/total_tokens": 1959002112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 5.494311809539795, "created_at": "2025-01-14T23:32:00.174692+00:00"} {"global_step": 7474, "acc_step": 0, "speed/wps": 12973.245868646962, "speed/FLOPS": 203762653475071.47, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.0040552616119385, "optim/lr": 0.002985059636603417, "optim/total_tokens": 1959264256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330623, "loss/out": 5.538898468017578, "created_at": "2025-01-14T23:32:05.233061+00:00"} {"global_step": 7475, "acc_step": 0, "speed/wps": 12967.64338264081, "speed/FLOPS": 203674658733722.3, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.3023200035095215, "optim/lr": 0.002985047571493998, "optim/total_tokens": 1959526400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290898, "loss/out": 5.504419803619385, "created_at": "2025-01-14T23:32:10.292189+00:00"} {"global_step": 7476, "acc_step": 0, "speed/wps": 12970.568243992131, "speed/FLOPS": 203720597700425.5, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.646695137023926, "optim/lr": 0.0029850355015393496, "optim/total_tokens": 1959788544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348659, "loss/out": 5.448212146759033, "created_at": "2025-01-14T23:32:15.355203+00:00"} {"global_step": 7477, "acc_step": 0, "speed/wps": 12969.94767224338, "speed/FLOPS": 203710850768356.4, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.641188144683838, "optim/lr": 0.0029850234267395125, "optim/total_tokens": 1960050688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428385, "loss/out": 5.529265880584717, "created_at": "2025-01-14T23:32:20.412342+00:00"} {"global_step": 7478, "acc_step": 0, "speed/wps": 12954.678111310552, "speed/FLOPS": 203471021331330.8, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.996370553970337, "optim/lr": 0.0029850113470945247, "optim/total_tokens": 1960312832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460578, "loss/out": 5.550981521606445, "created_at": "2025-01-14T23:32:25.474758+00:00"} {"global_step": 7479, "acc_step": 0, "speed/wps": 12979.563401370426, "speed/FLOPS": 203861879007693.16, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.582408428192139, "optim/lr": 0.002984999262604426, "optim/total_tokens": 1960574976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466683, "loss/out": 5.364144325256348, "created_at": "2025-01-14T23:32:30.531438+00:00"} {"global_step": 7480, "acc_step": 0, "speed/wps": 12972.01482012539, "speed/FLOPS": 203743318166402.25, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 104.15425872802734, "optim/lr": 0.0029849871732692563, "optim/total_tokens": 1960837120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443543, "loss/out": 5.493770599365234, "created_at": "2025-01-14T23:32:35.588685+00:00"} {"global_step": 7481, "acc_step": 0, "speed/wps": 12981.941234736321, "speed/FLOPS": 203899226148190.0, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0008, "optim/grad_norm": 23.853050231933594, "optim/lr": 0.0029849750790890547, "optim/total_tokens": 1961099264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424963, "loss/out": 5.576844215393066, "created_at": "2025-01-14T23:32:40.641152+00:00"} {"global_step": 7482, "acc_step": 0, "speed/wps": 12970.61468969087, "speed/FLOPS": 203721327193947.28, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.281546592712402, "optim/lr": 0.002984962980063861, "optim/total_tokens": 1961361408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441679, "loss/out": 5.643003940582275, "created_at": "2025-01-14T23:32:45.698167+00:00"} {"global_step": 7483, "acc_step": 0, "speed/wps": 12967.585074366218, "speed/FLOPS": 203673742922145.88, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 58.67364501953125, "optim/lr": 0.0029849508761937145, "optim/total_tokens": 1961623552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 5.684915065765381, "created_at": "2025-01-14T23:32:50.761444+00:00"} {"global_step": 7484, "acc_step": 0, "speed/wps": 12975.877703775805, "speed/FLOPS": 203803990062290.38, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0007, "optim/grad_norm": 15.58897876739502, "optim/lr": 0.0029849387674786548, "optim/total_tokens": 1961885696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429470, "loss/out": 5.626534461975098, "created_at": "2025-01-14T23:32:55.815974+00:00"} {"global_step": 7485, "acc_step": 0, "speed/wps": 12970.921222430621, "speed/FLOPS": 203726141711845.2, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.9221103191375732, "optim/lr": 0.0029849266539187214, "optim/total_tokens": 1962147840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318851, "loss/out": 5.639591217041016, "created_at": "2025-01-14T23:33:00.876986+00:00"} {"global_step": 7486, "acc_step": 0, "speed/wps": 12972.865953568935, "speed/FLOPS": 203756686386713.03, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 22.138809204101562, "optim/lr": 0.002984914535513953, "optim/total_tokens": 1962409984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357486, "loss/out": 5.540266990661621, "created_at": "2025-01-14T23:33:05.932742+00:00"} {"global_step": 7487, "acc_step": 0, "speed/wps": 12966.429072921925, "speed/FLOPS": 203655586330950.8, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 21.740571975708008, "optim/lr": 0.00298490241226439, "optim/total_tokens": 1962672128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456312, "loss/out": 5.673334121704102, "created_at": "2025-01-14T23:33:10.991585+00:00"} {"global_step": 7488, "acc_step": 0, "speed/wps": 12966.86477623695, "speed/FLOPS": 203662429650232.38, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0012, "optim/grad_norm": 6.403814315795898, "optim/lr": 0.0029848902841700716, "optim/total_tokens": 1962934272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 5.469770908355713, "created_at": "2025-01-14T23:33:16.050507+00:00"} {"global_step": 7489, "acc_step": 0, "speed/wps": 12964.476759912734, "speed/FLOPS": 203624922572382.25, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.805716514587402, "optim/lr": 0.002984878151231038, "optim/total_tokens": 1963196416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 5.667588233947754, "created_at": "2025-01-14T23:33:21.110578+00:00"} {"global_step": 7490, "acc_step": 0, "speed/wps": 12964.789165298189, "speed/FLOPS": 203629829328250.8, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.39249038696289, "optim/lr": 0.0029848660134473284, "optim/total_tokens": 1963458560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452325, "loss/out": 5.596470355987549, "created_at": "2025-01-14T23:33:26.178122+00:00"} {"global_step": 7491, "acc_step": 0, "speed/wps": 12972.14426756916, "speed/FLOPS": 203745351316386.06, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.597395896911621, "optim/lr": 0.002984853870818982, "optim/total_tokens": 1963720704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452615, "loss/out": 5.593001365661621, "created_at": "2025-01-14T23:33:31.239336+00:00"} {"global_step": 7492, "acc_step": 0, "speed/wps": 12970.752056339585, "speed/FLOPS": 203723484725926.88, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.748077392578125, "optim/lr": 0.0029848417233460387, "optim/total_tokens": 1963982848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 5.49644136428833, "created_at": "2025-01-14T23:33:36.295572+00:00"} {"global_step": 7493, "acc_step": 0, "speed/wps": 12963.972955512103, "speed/FLOPS": 203617009631970.56, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.280803918838501, "optim/lr": 0.0029848295710285383, "optim/total_tokens": 1964244992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426829, "loss/out": 5.587669849395752, "created_at": "2025-01-14T23:33:41.356804+00:00"} {"global_step": 7494, "acc_step": 0, "speed/wps": 12963.963676811865, "speed/FLOPS": 203616863897233.06, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.62122631072998, "optim/lr": 0.0029848174138665203, "optim/total_tokens": 1964507136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326870, "loss/out": 5.6872992515563965, "created_at": "2025-01-14T23:33:46.422981+00:00"} {"global_step": 7495, "acc_step": 0, "speed/wps": 12963.664826175553, "speed/FLOPS": 203612170037325.97, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 11.614392280578613, "optim/lr": 0.0029848052518600247, "optim/total_tokens": 1964769280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433481, "loss/out": 5.529726028442383, "created_at": "2025-01-14T23:33:51.481955+00:00"} {"global_step": 7496, "acc_step": 0, "speed/wps": 12966.188372741302, "speed/FLOPS": 203651805803857.44, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.3571364879608154, "optim/lr": 0.0029847930850090904, "optim/total_tokens": 1965031424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481758, "loss/out": 5.547733306884766, "created_at": "2025-01-14T23:33:56.544731+00:00"} {"global_step": 7497, "acc_step": 0, "speed/wps": 12967.667518596445, "speed/FLOPS": 203675037822071.84, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 9.802895545959473, "optim/lr": 0.0029847809133137578, "optim/total_tokens": 1965293568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429739, "loss/out": 5.506391525268555, "created_at": "2025-01-14T23:34:01.604102+00:00"} {"global_step": 7498, "acc_step": 0, "speed/wps": 12968.144284267757, "speed/FLOPS": 203682526082084.5, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.300594329833984, "optim/lr": 0.0029847687367740663, "optim/total_tokens": 1965555712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476431, "loss/out": 5.736550807952881, "created_at": "2025-01-14T23:34:06.665603+00:00"} {"global_step": 7499, "acc_step": 0, "speed/wps": 12971.982303234034, "speed/FLOPS": 203742807443940.6, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.555534362792969, "optim/lr": 0.0029847565553900554, "optim/total_tokens": 1965817856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418037, "loss/out": 5.45380973815918, "created_at": "2025-01-14T23:34:11.722665+00:00"} {"global_step": 7500, "acc_step": 0, "speed/wps": 12964.291861884185, "speed/FLOPS": 203622018494767.38, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 39.35743713378906, "optim/lr": 0.002984744369161766, "optim/total_tokens": 1966080000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302164, "loss/out": 5.54454231262207, "created_at": "2025-01-14T23:34:16.781852+00:00"} {"global_step": 7501, "acc_step": 0, "speed/wps": 3409.816323123529, "speed/FLOPS": 53555850933296.55, "speed/curr_iter_time": 1.2642, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.246819019317627, "optim/lr": 0.0029847321780892364, "optim/total_tokens": 1966342144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433179, "loss/out": 5.7792487144470215, "created_at": "2025-01-14T23:34:36.007332+00:00"} {"global_step": 7502, "acc_step": 0, "speed/wps": 13036.430526315848, "speed/FLOPS": 204755055348579.53, "speed/curr_iter_time": 1.2693, "speed/data_load_time": 0.0007, "optim/grad_norm": 221.11813354492188, "optim/lr": 0.002984719982172507, "optim/total_tokens": 1966604288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462963, "loss/out": 5.705776691436768, "created_at": "2025-01-14T23:34:41.044734+00:00"} {"global_step": 7503, "acc_step": 0, "speed/wps": 13015.696314413746, "speed/FLOPS": 204429396058864.16, "speed/curr_iter_time": 1.2702, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.845515251159668, "optim/lr": 0.0029847077814116178, "optim/total_tokens": 1966866432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462150, "loss/out": 5.674249649047852, "created_at": "2025-01-14T23:34:46.086158+00:00"} {"global_step": 7504, "acc_step": 0, "speed/wps": 13002.182833410749, "speed/FLOPS": 204217148270243.25, "speed/curr_iter_time": 1.2709, "speed/data_load_time": 0.0009, "optim/grad_norm": 4.848862648010254, "optim/lr": 0.002984695575806608, "optim/total_tokens": 1967128576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468018, "loss/out": 5.720255374908447, "created_at": "2025-01-14T23:34:51.130980+00:00"} {"global_step": 7505, "acc_step": 0, "speed/wps": 12995.608655173757, "speed/FLOPS": 204113891767163.72, "speed/curr_iter_time": 1.2713, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.095121383666992, "optim/lr": 0.0029846833653575182, "optim/total_tokens": 1967390720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471454, "loss/out": 5.633574962615967, "created_at": "2025-01-14T23:34:56.182217+00:00"} {"global_step": 7506, "acc_step": 0, "speed/wps": 12973.897170263914, "speed/FLOPS": 203772883061947.9, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.290081977844238, "optim/lr": 0.0029846711500643875, "optim/total_tokens": 1967652864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457665, "loss/out": 5.609485626220703, "created_at": "2025-01-14T23:35:01.241122+00:00"} {"global_step": 7507, "acc_step": 0, "speed/wps": 12979.967940874067, "speed/FLOPS": 203868232856493.38, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 239.47320556640625, "optim/lr": 0.0029846589299272566, "optim/total_tokens": 1967915008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432637, "loss/out": 5.655610084533691, "created_at": "2025-01-14T23:35:06.303044+00:00"} {"global_step": 7508, "acc_step": 0, "speed/wps": 12979.451059696645, "speed/FLOPS": 203860114527332.78, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.600776672363281, "optim/lr": 0.0029846467049461645, "optim/total_tokens": 1968177152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330623, "loss/out": 5.549918174743652, "created_at": "2025-01-14T23:35:11.355809+00:00"} {"global_step": 7509, "acc_step": 0, "speed/wps": 12978.289511723588, "speed/FLOPS": 203841870820281.75, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 54.006107330322266, "optim/lr": 0.0029846344751211515, "optim/total_tokens": 1968439296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376530, "loss/out": 5.6132612228393555, "created_at": "2025-01-14T23:35:16.411580+00:00"} {"global_step": 7510, "acc_step": 0, "speed/wps": 12979.80908029482, "speed/FLOPS": 203865737732799.0, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0006, "optim/grad_norm": 559.9179077148438, "optim/lr": 0.0029846222404522576, "optim/total_tokens": 1968701440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426829, "loss/out": 5.641876220703125, "created_at": "2025-01-14T23:35:21.464322+00:00"} {"global_step": 7511, "acc_step": 0, "speed/wps": 12977.995253994302, "speed/FLOPS": 203837249098290.8, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.38524341583252, "optim/lr": 0.002984610000939523, "optim/total_tokens": 1968963584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466155, "loss/out": 5.715541839599609, "created_at": "2025-01-14T23:35:26.517684+00:00"} {"global_step": 7512, "acc_step": 0, "speed/wps": 12978.337819932263, "speed/FLOPS": 203842629567081.44, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.160389423370361, "optim/lr": 0.0029845977565829862, "optim/total_tokens": 1969225728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 5.859755992889404, "created_at": "2025-01-14T23:35:31.572015+00:00"} {"global_step": 7513, "acc_step": 0, "speed/wps": 12976.934342793355, "speed/FLOPS": 203820586030035.2, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.203065395355225, "optim/lr": 0.0029845855073826885, "optim/total_tokens": 1969487872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446742, "loss/out": 5.583514213562012, "created_at": "2025-01-14T23:35:36.625732+00:00"} {"global_step": 7514, "acc_step": 0, "speed/wps": 12975.007849224952, "speed/FLOPS": 203790327801263.84, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.8713884353637695, "optim/lr": 0.00298457325333867, "optim/total_tokens": 1969750016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 5.512199401855469, "created_at": "2025-01-14T23:35:41.680171+00:00"} {"global_step": 7515, "acc_step": 0, "speed/wps": 12976.903990927734, "speed/FLOPS": 203820109312277.78, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 20.547204971313477, "optim/lr": 0.00298456099445097, "optim/total_tokens": 1970012160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437201, "loss/out": 5.546676158905029, "created_at": "2025-01-14T23:35:46.734494+00:00"} {"global_step": 7516, "acc_step": 0, "speed/wps": 12977.256802141554, "speed/FLOPS": 203825650697204.22, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.7194149494171143, "optim/lr": 0.002984548730719628, "optim/total_tokens": 1970274304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417192, "loss/out": 5.524652481079102, "created_at": "2025-01-14T23:35:51.793388+00:00"} {"global_step": 7517, "acc_step": 0, "speed/wps": 12978.196326391697, "speed/FLOPS": 203840407216594.34, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.0565690994262695, "optim/lr": 0.002984536462144686, "optim/total_tokens": 1970536448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 5.626579284667969, "created_at": "2025-01-14T23:35:56.847799+00:00"} {"global_step": 7518, "acc_step": 0, "speed/wps": 12974.890438022698, "speed/FLOPS": 203788483696992.53, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 15.658385276794434, "optim/lr": 0.002984524188726182, "optim/total_tokens": 1970798592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444119, "loss/out": 5.537931442260742, "created_at": "2025-01-14T23:36:01.907800+00:00"} {"global_step": 7519, "acc_step": 0, "speed/wps": 12970.242155637583, "speed/FLOPS": 203715476034724.97, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.155061721801758, "optim/lr": 0.002984511910464157, "optim/total_tokens": 1971060736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287143, "loss/out": 5.624456405639648, "created_at": "2025-01-14T23:36:06.966965+00:00"} {"global_step": 7520, "acc_step": 0, "speed/wps": 12972.609632045154, "speed/FLOPS": 203752660504959.38, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.769336700439453, "optim/lr": 0.0029844996273586505, "optim/total_tokens": 1971322880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 388103, "loss/out": 5.574064254760742, "created_at": "2025-01-14T23:36:12.023856+00:00"} {"global_step": 7521, "acc_step": 0, "speed/wps": 12971.087762873605, "speed/FLOPS": 203728757458352.78, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.804041862487793, "optim/lr": 0.002984487339409703, "optim/total_tokens": 1971585024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450471, "loss/out": 5.580124855041504, "created_at": "2025-01-14T23:36:17.083846+00:00"} {"global_step": 7522, "acc_step": 0, "speed/wps": 12974.631924117573, "speed/FLOPS": 203784423380877.22, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.983626842498779, "optim/lr": 0.002984475046617355, "optim/total_tokens": 1971847168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 5.553928852081299, "created_at": "2025-01-14T23:36:22.139797+00:00"} {"global_step": 7523, "acc_step": 0, "speed/wps": 12970.261039425513, "speed/FLOPS": 203715772630563.44, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.001, "optim/grad_norm": 3.366396427154541, "optim/lr": 0.0029844627489816456, "optim/total_tokens": 1972109312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418037, "loss/out": 5.559676647186279, "created_at": "2025-01-14T23:36:27.201129+00:00"} {"global_step": 7524, "acc_step": 0, "speed/wps": 12972.91535265496, "speed/FLOPS": 203757462267240.5, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5792663097381592, "optim/lr": 0.0029844504465026157, "optim/total_tokens": 1972371456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 5.573680877685547, "created_at": "2025-01-14T23:36:32.256762+00:00"} {"global_step": 7525, "acc_step": 0, "speed/wps": 12975.459929827486, "speed/FLOPS": 203797428348350.47, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.464013576507568, "optim/lr": 0.002984438139180305, "optim/total_tokens": 1972633600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453660, "loss/out": 5.544079303741455, "created_at": "2025-01-14T23:36:37.311076+00:00"} {"global_step": 7526, "acc_step": 0, "speed/wps": 12972.584982645318, "speed/FLOPS": 203752273352263.1, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 18.16199493408203, "optim/lr": 0.002984425827014754, "optim/total_tokens": 1972895744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434505, "loss/out": 5.585090637207031, "created_at": "2025-01-14T23:36:42.369470+00:00"} {"global_step": 7527, "acc_step": 0, "speed/wps": 12972.787791135885, "speed/FLOPS": 203755458738295.53, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.239761352539062, "optim/lr": 0.0029844135100060027, "optim/total_tokens": 1973157888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331967, "loss/out": 5.589919090270996, "created_at": "2025-01-14T23:36:47.425554+00:00"} {"global_step": 7528, "acc_step": 0, "speed/wps": 12974.3144058195, "speed/FLOPS": 203779436319690.1, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.626623153686523, "optim/lr": 0.0029844011881540915, "optim/total_tokens": 1973420032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329658, "loss/out": 5.572234630584717, "created_at": "2025-01-14T23:36:52.486532+00:00"} {"global_step": 7529, "acc_step": 0, "speed/wps": 12975.693406368546, "speed/FLOPS": 203801095418258.22, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8022656440734863, "optim/lr": 0.00298438886145906, "optim/total_tokens": 1973682176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322156, "loss/out": 5.708963394165039, "created_at": "2025-01-14T23:36:57.545228+00:00"} {"global_step": 7530, "acc_step": 0, "speed/wps": 12970.895783279277, "speed/FLOPS": 203725742155016.25, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 83.43011474609375, "optim/lr": 0.002984376529920949, "optim/total_tokens": 1973944320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 5.653761863708496, "created_at": "2025-01-14T23:37:02.605010+00:00"} {"global_step": 7531, "acc_step": 0, "speed/wps": 12969.690791919338, "speed/FLOPS": 203706816109876.78, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 32.24034881591797, "optim/lr": 0.002984364193539798, "optim/total_tokens": 1974206464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 5.612405300140381, "created_at": "2025-01-14T23:37:07.666325+00:00"} {"global_step": 7532, "acc_step": 0, "speed/wps": 12975.003731890081, "speed/FLOPS": 203790263132861.88, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 15.914765357971191, "optim/lr": 0.002984351852315649, "optim/total_tokens": 1974468608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444881, "loss/out": 5.873259544372559, "created_at": "2025-01-14T23:37:12.752668+00:00"} {"global_step": 7533, "acc_step": 0, "speed/wps": 12970.895962596254, "speed/FLOPS": 203725744971435.8, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.3580121994018555, "optim/lr": 0.0029843395062485404, "optim/total_tokens": 1974730752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427066, "loss/out": 5.90136194229126, "created_at": "2025-01-14T23:37:17.808816+00:00"} {"global_step": 7534, "acc_step": 0, "speed/wps": 12969.95440640786, "speed/FLOPS": 203710956537663.28, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.108454704284668, "optim/lr": 0.0029843271553385135, "optim/total_tokens": 1974992896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298942, "loss/out": 5.74948787689209, "created_at": "2025-01-14T23:37:22.866699+00:00"} {"global_step": 7535, "acc_step": 0, "speed/wps": 12974.37694418864, "speed/FLOPS": 203780418570714.2, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.142051935195923, "optim/lr": 0.002984314799585608, "optim/total_tokens": 1975255040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351583, "loss/out": 5.74980354309082, "created_at": "2025-01-14T23:37:27.931681+00:00"} {"global_step": 7536, "acc_step": 0, "speed/wps": 12970.451532322824, "speed/FLOPS": 203718764583280.38, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.675969123840332, "optim/lr": 0.002984302438989864, "optim/total_tokens": 1975517184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446218, "loss/out": 5.650959014892578, "created_at": "2025-01-14T23:37:32.988819+00:00"} {"global_step": 7537, "acc_step": 0, "speed/wps": 12963.375362686926, "speed/FLOPS": 203607623615472.56, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.436107158660889, "optim/lr": 0.0029842900735513227, "optim/total_tokens": 1975779328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 5.566037654876709, "created_at": "2025-01-14T23:37:38.051461+00:00"} {"global_step": 7538, "acc_step": 0, "speed/wps": 12969.669239668878, "speed/FLOPS": 203706477602167.4, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.896183967590332, "optim/lr": 0.002984277703270024, "optim/total_tokens": 1976041472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422813, "loss/out": 5.628684997558594, "created_at": "2025-01-14T23:37:43.109768+00:00"} {"global_step": 7539, "acc_step": 0, "speed/wps": 12963.737020322047, "speed/FLOPS": 203613303945601.8, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 10.33055591583252, "optim/lr": 0.0029842653281460085, "optim/total_tokens": 1976303616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481190, "loss/out": 5.675915241241455, "created_at": "2025-01-14T23:37:48.170353+00:00"} {"global_step": 7540, "acc_step": 0, "speed/wps": 12969.915371300314, "speed/FLOPS": 203710343437659.4, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 16.44437599182129, "optim/lr": 0.002984252948179316, "optim/total_tokens": 1976565760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461385, "loss/out": 5.674201488494873, "created_at": "2025-01-14T23:37:53.228093+00:00"} {"global_step": 7541, "acc_step": 0, "speed/wps": 12971.3369013015, "speed/FLOPS": 203732670519714.97, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.760040283203125, "optim/lr": 0.0029842405633699874, "optim/total_tokens": 1976827904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460556, "loss/out": 5.659204006195068, "created_at": "2025-01-14T23:37:58.288643+00:00"} {"global_step": 7542, "acc_step": 0, "speed/wps": 12970.23058489205, "speed/FLOPS": 203715294300266.5, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 28.476778030395508, "optim/lr": 0.002984228173718063, "optim/total_tokens": 1977090048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324468, "loss/out": 5.604137420654297, "created_at": "2025-01-14T23:38:03.350622+00:00"} {"global_step": 7543, "acc_step": 0, "speed/wps": 12971.860446450153, "speed/FLOPS": 203740893515700.97, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0006, "optim/grad_norm": 25.40932273864746, "optim/lr": 0.0029842157792235834, "optim/total_tokens": 1977352192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356961, "loss/out": 5.600200653076172, "created_at": "2025-01-14T23:38:08.406689+00:00"} {"global_step": 7544, "acc_step": 0, "speed/wps": 12973.814854904016, "speed/FLOPS": 203771590186107.56, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 34.4198112487793, "optim/lr": 0.0029842033798865883, "optim/total_tokens": 1977614336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465057, "loss/out": 5.626954555511475, "created_at": "2025-01-14T23:38:13.461801+00:00"} {"global_step": 7545, "acc_step": 0, "speed/wps": 12965.620669174567, "speed/FLOPS": 203642889239235.38, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0012, "optim/grad_norm": 12.007943153381348, "optim/lr": 0.002984190975707119, "optim/total_tokens": 1977876480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 5.623270034790039, "created_at": "2025-01-14T23:38:18.520368+00:00"} {"global_step": 7546, "acc_step": 0, "speed/wps": 12965.76450459303, "speed/FLOPS": 203645148372132.94, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.570028305053711, "optim/lr": 0.0029841785666852158, "optim/total_tokens": 1978138624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 5.553414344787598, "created_at": "2025-01-14T23:38:23.578474+00:00"} {"global_step": 7547, "acc_step": 0, "speed/wps": 12975.13342560511, "speed/FLOPS": 203792300150874.25, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.033189058303833, "optim/lr": 0.002984166152820919, "optim/total_tokens": 1978400768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435044, "loss/out": 5.593769073486328, "created_at": "2025-01-14T23:38:28.640822+00:00"} {"global_step": 7548, "acc_step": 0, "speed/wps": 12971.466913862716, "speed/FLOPS": 203734712545644.84, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.23595142364502, "optim/lr": 0.0029841537341142687, "optim/total_tokens": 1978662912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 5.611116409301758, "created_at": "2025-01-14T23:38:33.697165+00:00"} {"global_step": 7549, "acc_step": 0, "speed/wps": 12971.07414156732, "speed/FLOPS": 203728543517019.94, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 21.156599044799805, "optim/lr": 0.0029841413105653066, "optim/total_tokens": 1978925056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432127, "loss/out": 5.646923065185547, "created_at": "2025-01-14T23:38:38.757747+00:00"} {"global_step": 7550, "acc_step": 0, "speed/wps": 12973.867564552447, "speed/FLOPS": 203772418063566.8, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 564.330078125, "optim/lr": 0.002984128882174072, "optim/total_tokens": 1979187200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296475, "loss/out": 5.632758140563965, "created_at": "2025-01-14T23:38:43.812691+00:00"} {"global_step": 7551, "acc_step": 0, "speed/wps": 12968.44158004989, "speed/FLOPS": 203687195520870.1, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0008, "optim/grad_norm": 63.303443908691406, "optim/lr": 0.0029841164489406062, "optim/total_tokens": 1979449344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358542, "loss/out": 5.736384391784668, "created_at": "2025-01-14T23:38:48.879870+00:00"} {"global_step": 7552, "acc_step": 0, "speed/wps": 12970.275500537915, "speed/FLOPS": 203715999762205.38, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 77.42623138427734, "optim/lr": 0.0029841040108649496, "optim/total_tokens": 1979711488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426252, "loss/out": 5.889147758483887, "created_at": "2025-01-14T23:38:53.940882+00:00"} {"global_step": 7553, "acc_step": 0, "speed/wps": 12969.294458060203, "speed/FLOPS": 203700591141999.28, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.500465393066406, "optim/lr": 0.0029840915679471424, "optim/total_tokens": 1979973632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 5.676937103271484, "created_at": "2025-01-14T23:38:59.002911+00:00"} {"global_step": 7554, "acc_step": 0, "speed/wps": 12974.703398378202, "speed/FLOPS": 203785545982356.3, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 700.3818359375, "optim/lr": 0.0029840791201872253, "optim/total_tokens": 1980235776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 5.678540229797363, "created_at": "2025-01-14T23:39:04.058800+00:00"} {"global_step": 7555, "acc_step": 0, "speed/wps": 12966.083675909293, "speed/FLOPS": 203650161396242.2, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1801.397216796875, "optim/lr": 0.0029840666675852398, "optim/total_tokens": 1980497920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453375, "loss/out": 5.75865364074707, "created_at": "2025-01-14T23:39:09.121565+00:00"} {"global_step": 7556, "acc_step": 0, "speed/wps": 12969.095666803189, "speed/FLOPS": 203697468852140.03, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 85.59649658203125, "optim/lr": 0.002984054210141225, "optim/total_tokens": 1980760064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465870, "loss/out": 5.814873695373535, "created_at": "2025-01-14T23:39:14.179568+00:00"} {"global_step": 7557, "acc_step": 0, "speed/wps": 12958.910939438423, "speed/FLOPS": 203537503713594.4, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.758731842041016, "optim/lr": 0.002984041747855223, "optim/total_tokens": 1981022208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428426, "loss/out": 5.939995288848877, "created_at": "2025-01-14T23:39:19.240691+00:00"} {"global_step": 7558, "acc_step": 0, "speed/wps": 12975.185057392217, "speed/FLOPS": 203793111099040.28, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 15.782739639282227, "optim/lr": 0.002984029280727274, "optim/total_tokens": 1981284352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316975, "loss/out": 5.73709774017334, "created_at": "2025-01-14T23:39:24.295483+00:00"} {"global_step": 7559, "acc_step": 0, "speed/wps": 12967.59822474809, "speed/FLOPS": 203673949466963.7, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.262639999389648, "optim/lr": 0.0029840168087574177, "optim/total_tokens": 1981546496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 347191, "loss/out": 5.84493350982666, "created_at": "2025-01-14T23:39:29.354663+00:00"} {"global_step": 7560, "acc_step": 0, "speed/wps": 12967.61621665485, "speed/FLOPS": 203674232054582.8, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.042123794555664, "optim/lr": 0.0029840043319456964, "optim/total_tokens": 1981808640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435338, "loss/out": 5.923250198364258, "created_at": "2025-01-14T23:39:34.415724+00:00"} {"global_step": 7561, "acc_step": 0, "speed/wps": 12969.31479956803, "speed/FLOPS": 203700910633331.84, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 102.83008575439453, "optim/lr": 0.0029839918502921497, "optim/total_tokens": 1982070784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 5.773574352264404, "created_at": "2025-01-14T23:39:39.472800+00:00"} {"global_step": 7562, "acc_step": 0, "speed/wps": 12969.29510105166, "speed/FLOPS": 203700601241063.7, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 33.72872543334961, "optim/lr": 0.002983979363796818, "optim/total_tokens": 1982332928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439586, "loss/out": 5.8067779541015625, "created_at": "2025-01-14T23:39:44.562339+00:00"} {"global_step": 7563, "acc_step": 0, "speed/wps": 12973.905724060116, "speed/FLOPS": 203773017411071.4, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 443.59130859375, "optim/lr": 0.0029839668724597436, "optim/total_tokens": 1982595072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 414278, "loss/out": 5.861395359039307, "created_at": "2025-01-14T23:39:49.622133+00:00"} {"global_step": 7564, "acc_step": 0, "speed/wps": 12968.627062526657, "speed/FLOPS": 203690108778047.34, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0011, "optim/grad_norm": 286.05029296875, "optim/lr": 0.002983954376280966, "optim/total_tokens": 1982857216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301523, "loss/out": 5.935503005981445, "created_at": "2025-01-14T23:39:54.684277+00:00"} {"global_step": 7565, "acc_step": 0, "speed/wps": 12975.948344726874, "speed/FLOPS": 203805099575497.78, "speed/curr_iter_time": 1.2723, "speed/data_load_time": 0.0006, "optim/grad_norm": 29.067827224731445, "optim/lr": 0.0029839418752605268, "optim/total_tokens": 1983119360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417484, "loss/out": 6.075290679931641, "created_at": "2025-01-14T23:39:59.788718+00:00"} {"global_step": 7566, "acc_step": 0, "speed/wps": 12967.47380021459, "speed/FLOPS": 203671995208687.84, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 45.79678726196289, "optim/lr": 0.0029839293693984656, "optim/total_tokens": 1983381504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 6.065140247344971, "created_at": "2025-01-14T23:40:04.846412+00:00"} {"global_step": 7567, "acc_step": 0, "speed/wps": 12971.063958482013, "speed/FLOPS": 203728383577672.2, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 165.5980224609375, "optim/lr": 0.0029839168586948245, "optim/total_tokens": 1983643648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 5.819584369659424, "created_at": "2025-01-14T23:40:09.903645+00:00"} {"global_step": 7568, "acc_step": 0, "speed/wps": 12967.128825442884, "speed/FLOPS": 203666576905853.2, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0008, "optim/grad_norm": 36.55744171142578, "optim/lr": 0.002983904343149643, "optim/total_tokens": 1983905792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 6.0098042488098145, "created_at": "2025-01-14T23:40:14.964502+00:00"} {"global_step": 7569, "acc_step": 0, "speed/wps": 12974.37509982081, "speed/FLOPS": 203780389602383.0, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.831483840942383, "optim/lr": 0.002983891822762963, "optim/total_tokens": 1984167936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434807, "loss/out": 6.018734455108643, "created_at": "2025-01-14T23:40:20.020158+00:00"} {"global_step": 7570, "acc_step": 0, "speed/wps": 12970.669814355608, "speed/FLOPS": 203722193002556.2, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.459095478057861, "optim/lr": 0.0029838792975348245, "optim/total_tokens": 1984430080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303156, "loss/out": 5.742314338684082, "created_at": "2025-01-14T23:40:25.077593+00:00"} {"global_step": 7571, "acc_step": 0, "speed/wps": 12978.273067578419, "speed/FLOPS": 203841612542385.25, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.115447998046875, "optim/lr": 0.0029838667674652697, "optim/total_tokens": 1984692224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362577, "loss/out": 5.791349411010742, "created_at": "2025-01-14T23:40:30.132020+00:00"} {"global_step": 7572, "acc_step": 0, "speed/wps": 12966.12171080036, "speed/FLOPS": 203650758786471.8, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.399105072021484, "optim/lr": 0.002983854232554339, "optim/total_tokens": 1984954368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455003, "loss/out": 5.717801094055176, "created_at": "2025-01-14T23:40:35.191200+00:00"} {"global_step": 7573, "acc_step": 0, "speed/wps": 12972.97075517601, "speed/FLOPS": 203758332439964.62, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 117.51581573486328, "optim/lr": 0.002983841692802072, "optim/total_tokens": 1985216512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462672, "loss/out": 5.643948554992676, "created_at": "2025-01-14T23:40:40.246532+00:00"} {"global_step": 7574, "acc_step": 0, "speed/wps": 12974.251758015245, "speed/FLOPS": 203778452349837.2, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.475670099258423, "optim/lr": 0.002983829148208511, "optim/total_tokens": 1985478656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 5.754759788513184, "created_at": "2025-01-14T23:40:45.302038+00:00"} {"global_step": 7575, "acc_step": 0, "speed/wps": 12968.668632915942, "speed/FLOPS": 203690761698134.78, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.036428451538086, "optim/lr": 0.002983816598773696, "optim/total_tokens": 1985740800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433464, "loss/out": 5.824296951293945, "created_at": "2025-01-14T23:40:50.361314+00:00"} {"global_step": 7576, "acc_step": 0, "speed/wps": 12976.312530019622, "speed/FLOPS": 203810819605963.44, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.243100166320801, "optim/lr": 0.002983804044497668, "optim/total_tokens": 1986002944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450756, "loss/out": 5.984497547149658, "created_at": "2025-01-14T23:40:55.420205+00:00"} {"global_step": 7577, "acc_step": 0, "speed/wps": 12968.309737684329, "speed/FLOPS": 203685124755348.56, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.0429694652557373, "optim/lr": 0.0029837914853804697, "optim/total_tokens": 1986265088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460287, "loss/out": 5.667020797729492, "created_at": "2025-01-14T23:41:00.477728+00:00"} {"global_step": 7578, "acc_step": 0, "speed/wps": 12963.667584579489, "speed/FLOPS": 203612213361850.22, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 49.55451965332031, "optim/lr": 0.0029837789214221403, "optim/total_tokens": 1986527232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 410588, "loss/out": 5.6947832107543945, "created_at": "2025-01-14T23:41:05.544810+00:00"} {"global_step": 7579, "acc_step": 0, "speed/wps": 12978.50224575051, "speed/FLOPS": 203845212100501.72, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0006, "optim/grad_norm": 32.89274215698242, "optim/lr": 0.002983766352622721, "optim/total_tokens": 1986789376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287670, "loss/out": 5.978135108947754, "created_at": "2025-01-14T23:41:10.598933+00:00"} {"global_step": 7580, "acc_step": 0, "speed/wps": 12967.292209870802, "speed/FLOPS": 203669143082807.38, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.038095951080322, "optim/lr": 0.0029837537789822533, "optim/total_tokens": 1987051520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 343042, "loss/out": 6.275414943695068, "created_at": "2025-01-14T23:41:15.664529+00:00"} {"global_step": 7581, "acc_step": 0, "speed/wps": 12970.43469443878, "speed/FLOPS": 203718500121173.78, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 70.05718231201172, "optim/lr": 0.0029837412005007777, "optim/total_tokens": 1987313664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 396093, "loss/out": 5.835111618041992, "created_at": "2025-01-14T23:41:20.723020+00:00"} {"global_step": 7582, "acc_step": 0, "speed/wps": 12973.054863158317, "speed/FLOPS": 203759653471403.25, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.7876248359680176, "optim/lr": 0.002983728617178336, "optim/total_tokens": 1987575808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436381, "loss/out": 5.891547203063965, "created_at": "2025-01-14T23:41:25.783618+00:00"} {"global_step": 7583, "acc_step": 0, "speed/wps": 12972.646019197875, "speed/FLOPS": 203753232015194.03, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 9.554707527160645, "optim/lr": 0.0029837160290149684, "optim/total_tokens": 1987837952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 5.922370910644531, "created_at": "2025-01-14T23:41:30.840252+00:00"} {"global_step": 7584, "acc_step": 0, "speed/wps": 12970.269351752177, "speed/FLOPS": 203715903187076.03, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0009, "optim/grad_norm": 29.546939849853516, "optim/lr": 0.002983703436010717, "optim/total_tokens": 1988100096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440628, "loss/out": 5.841643333435059, "created_at": "2025-01-14T23:41:35.896684+00:00"} {"global_step": 7585, "acc_step": 0, "speed/wps": 12968.909530358816, "speed/FLOPS": 203694545323502.9, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.269259929656982, "optim/lr": 0.002983690838165621, "optim/total_tokens": 1988362240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467734, "loss/out": 5.82645320892334, "created_at": "2025-01-14T23:41:40.953943+00:00"} {"global_step": 7586, "acc_step": 0, "speed/wps": 12964.399661102592, "speed/FLOPS": 203623711629624.12, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0009, "optim/grad_norm": 3.3797359466552734, "optim/lr": 0.0029836782354797234, "optim/total_tokens": 1988624384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472260, "loss/out": 5.6992316246032715, "created_at": "2025-01-14T23:41:46.017150+00:00"} {"global_step": 7587, "acc_step": 0, "speed/wps": 12974.720974607306, "speed/FLOPS": 203785822041186.66, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.822031021118164, "optim/lr": 0.0029836656279530653, "optim/total_tokens": 1988886528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334076, "loss/out": 5.6616973876953125, "created_at": "2025-01-14T23:41:51.074683+00:00"} {"global_step": 7588, "acc_step": 0, "speed/wps": 12969.333104589312, "speed/FLOPS": 203701198138844.8, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 33.66073989868164, "optim/lr": 0.0029836530155856866, "optim/total_tokens": 1989148672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356674, "loss/out": 5.660910606384277, "created_at": "2025-01-14T23:41:56.132577+00:00"} {"global_step": 7589, "acc_step": 0, "speed/wps": 12966.356781890534, "speed/FLOPS": 203654450900964.53, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 35.5588493347168, "optim/lr": 0.0029836403983776287, "optim/total_tokens": 1989410816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443835, "loss/out": 5.74039363861084, "created_at": "2025-01-14T23:42:01.191784+00:00"} {"global_step": 7590, "acc_step": 0, "speed/wps": 12964.854060256699, "speed/FLOPS": 203630848592749.25, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 99.17788696289062, "optim/lr": 0.0029836277763289334, "optim/total_tokens": 1989672960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 5.873144626617432, "created_at": "2025-01-14T23:42:06.253208+00:00"} {"global_step": 7591, "acc_step": 0, "speed/wps": 12967.755065957837, "speed/FLOPS": 203676412873685.2, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.7084574699401855, "optim/lr": 0.0029836151494396414, "optim/total_tokens": 1989935104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444357, "loss/out": 5.957526206970215, "created_at": "2025-01-14T23:42:11.311855+00:00"} {"global_step": 7592, "acc_step": 0, "speed/wps": 12965.249487342362, "speed/FLOPS": 203637059318503.34, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.834610462188721, "optim/lr": 0.002983602517709795, "optim/total_tokens": 1990197248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458972, "loss/out": 5.806602478027344, "created_at": "2025-01-14T23:42:16.374555+00:00"} {"global_step": 7593, "acc_step": 0, "speed/wps": 12962.390776828044, "speed/FLOPS": 203592159341596.06, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.97756290435791, "optim/lr": 0.0029835898811394336, "optim/total_tokens": 1990459392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480663, "loss/out": 5.783184051513672, "created_at": "2025-01-14T23:42:21.438538+00:00"} {"global_step": 7594, "acc_step": 0, "speed/wps": 12964.25224941197, "speed/FLOPS": 203621396326458.0, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.361098051071167, "optim/lr": 0.0029835772397286, "optim/total_tokens": 1990721536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 5.726902961730957, "created_at": "2025-01-14T23:42:26.503184+00:00"} {"global_step": 7595, "acc_step": 0, "speed/wps": 12968.508660199177, "speed/FLOPS": 203688249106792.44, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.44869613647461, "optim/lr": 0.0029835645934773343, "optim/total_tokens": 1990983680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427066, "loss/out": 5.794812202453613, "created_at": "2025-01-14T23:42:31.561430+00:00"} {"global_step": 7596, "acc_step": 0, "speed/wps": 12968.787422973182, "speed/FLOPS": 203692627459218.38, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 40.3796501159668, "optim/lr": 0.0029835519423856783, "optim/total_tokens": 1991245824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304215, "loss/out": 5.831648349761963, "created_at": "2025-01-14T23:42:36.618190+00:00"} {"global_step": 7597, "acc_step": 0, "speed/wps": 12969.080493644597, "speed/FLOPS": 203697230536834.25, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7260754108428955, "optim/lr": 0.002983539286453673, "optim/total_tokens": 1991507968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360941, "loss/out": 5.808436393737793, "created_at": "2025-01-14T23:42:41.680391+00:00"} {"global_step": 7598, "acc_step": 0, "speed/wps": 12967.027262739635, "speed/FLOPS": 203664981724036.9, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.683379650115967, "optim/lr": 0.0029835266256813603, "optim/total_tokens": 1991770112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447269, "loss/out": 5.5648298263549805, "created_at": "2025-01-14T23:42:46.738710+00:00"} {"global_step": 7599, "acc_step": 0, "speed/wps": 12968.326935713432, "speed/FLOPS": 203685394874028.2, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.62128734588623, "optim/lr": 0.0029835139600687807, "optim/total_tokens": 1992032256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 5.818054676055908, "created_at": "2025-01-14T23:42:51.795936+00:00"} {"global_step": 7600, "acc_step": 0, "speed/wps": 12959.827318238751, "speed/FLOPS": 203551896701889.97, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.9959940910339355, "optim/lr": 0.002983501289615976, "optim/total_tokens": 1992294400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432352, "loss/out": 5.814422607421875, "created_at": "2025-01-14T23:42:56.861606+00:00"} {"global_step": 7601, "acc_step": 0, "speed/wps": 12964.230717621025, "speed/FLOPS": 203621058140093.44, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.168465614318848, "optim/lr": 0.002983488614322988, "optim/total_tokens": 1992556544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460840, "loss/out": 5.648078918457031, "created_at": "2025-01-14T23:43:01.925653+00:00"} {"global_step": 7602, "acc_step": 0, "speed/wps": 12967.113451389241, "speed/FLOPS": 203666335435214.6, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.9813814163208, "optim/lr": 0.0029834759341898576, "optim/total_tokens": 1992818688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317586, "loss/out": 5.615529537200928, "created_at": "2025-01-14T23:43:06.989123+00:00"} {"global_step": 7603, "acc_step": 0, "speed/wps": 12967.729455215287, "speed/FLOPS": 203676010621780.6, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.351592063903809, "optim/lr": 0.002983463249216625, "optim/total_tokens": 1993080832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443015, "loss/out": 5.707095146179199, "created_at": "2025-01-14T23:43:12.056668+00:00"} {"global_step": 7604, "acc_step": 0, "speed/wps": 12971.589712539047, "speed/FLOPS": 203736641267599.7, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.831485748291016, "optim/lr": 0.002983450559403334, "optim/total_tokens": 1993342976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 5.670022964477539, "created_at": "2025-01-14T23:43:17.116431+00:00"} {"global_step": 7605, "acc_step": 0, "speed/wps": 12968.72337296633, "speed/FLOPS": 203691621465846.8, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.001, "optim/grad_norm": 640.1736450195312, "optim/lr": 0.002983437864750024, "optim/total_tokens": 1993605120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438773, "loss/out": 5.594577312469482, "created_at": "2025-01-14T23:43:22.178309+00:00"} {"global_step": 7606, "acc_step": 0, "speed/wps": 12967.75516632672, "speed/FLOPS": 203676414450116.4, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.686203956604004, "optim/lr": 0.002983425165256737, "optim/total_tokens": 1993867264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434230, "loss/out": 5.713468551635742, "created_at": "2025-01-14T23:43:27.235788+00:00"} {"global_step": 7607, "acc_step": 0, "speed/wps": 12973.194627795046, "speed/FLOPS": 203761848667079.8, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.245280742645264, "optim/lr": 0.0029834124609235155, "optim/total_tokens": 1994129408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431320, "loss/out": 5.685324192047119, "created_at": "2025-01-14T23:43:32.295343+00:00"} {"global_step": 7608, "acc_step": 0, "speed/wps": 12965.145823289518, "speed/FLOPS": 203635431132104.1, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.315993309020996, "optim/lr": 0.002983399751750399, "optim/total_tokens": 1994391552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 5.708047866821289, "created_at": "2025-01-14T23:43:37.361757+00:00"} {"global_step": 7609, "acc_step": 0, "speed/wps": 12966.144768367147, "speed/FLOPS": 203651120937242.62, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.73021125793457, "optim/lr": 0.0029833870377374305, "optim/total_tokens": 1994653696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435329, "loss/out": 5.703808784484863, "created_at": "2025-01-14T23:43:42.420163+00:00"} {"global_step": 7610, "acc_step": 0, "speed/wps": 12977.218527189669, "speed/FLOPS": 203825049536489.56, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.648564338684082, "optim/lr": 0.0029833743188846505, "optim/total_tokens": 1994915840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316287, "loss/out": 5.661285400390625, "created_at": "2025-01-14T23:43:47.475379+00:00"} {"global_step": 7611, "acc_step": 0, "speed/wps": 12968.761915707542, "speed/FLOPS": 203692226832560.84, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.650069236755371, "optim/lr": 0.0029833615951921012, "optim/total_tokens": 1995177984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325381, "loss/out": 5.6560564041137695, "created_at": "2025-01-14T23:43:52.537539+00:00"} {"global_step": 7612, "acc_step": 0, "speed/wps": 12971.390566375785, "speed/FLOPS": 203733513403450.16, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 153.0658416748047, "optim/lr": 0.002983348866659824, "optim/total_tokens": 1995440128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452330, "loss/out": 5.6466755867004395, "created_at": "2025-01-14T23:43:57.593765+00:00"} {"global_step": 7613, "acc_step": 0, "speed/wps": 12970.757418995941, "speed/FLOPS": 203723568953814.03, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 70.3022232055664, "optim/lr": 0.00298333613328786, "optim/total_tokens": 1995702272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 5.610806941986084, "created_at": "2025-01-14T23:44:02.656540+00:00"} {"global_step": 7614, "acc_step": 0, "speed/wps": 12965.327327245817, "speed/FLOPS": 203638281901150.34, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 26.64715003967285, "optim/lr": 0.0029833233950762516, "optim/total_tokens": 1995964416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426517, "loss/out": 5.716266632080078, "created_at": "2025-01-14T23:44:07.715395+00:00"} {"global_step": 7615, "acc_step": 0, "speed/wps": 12970.645382717408, "speed/FLOPS": 203721809270105.7, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.353842735290527, "optim/lr": 0.0029833106520250393, "optim/total_tokens": 1996226560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 5.802857875823975, "created_at": "2025-01-14T23:44:12.775875+00:00"} {"global_step": 7616, "acc_step": 0, "speed/wps": 12965.791763159688, "speed/FLOPS": 203645576505379.5, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.018932342529297, "optim/lr": 0.002983297904134265, "optim/total_tokens": 1996488704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423342, "loss/out": 5.66533899307251, "created_at": "2025-01-14T23:44:17.834496+00:00"} {"global_step": 7617, "acc_step": 0, "speed/wps": 12965.16419537417, "speed/FLOPS": 203635719690939.4, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.567348003387451, "optim/lr": 0.002983285151403971, "optim/total_tokens": 1996750848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289838, "loss/out": 5.611350059509277, "created_at": "2025-01-14T23:44:22.893552+00:00"} {"global_step": 7618, "acc_step": 0, "speed/wps": 12973.443383148817, "speed/FLOPS": 203765755711736.47, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.57968521118164, "optim/lr": 0.002983272393834198, "optim/total_tokens": 1997012992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 392005, "loss/out": 5.62122917175293, "created_at": "2025-01-14T23:44:27.949049+00:00"} {"global_step": 7619, "acc_step": 0, "speed/wps": 12967.75601946229, "speed/FLOPS": 203676427849782.88, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 21.660266876220703, "optim/lr": 0.002983259631424988, "optim/total_tokens": 1997275136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446220, "loss/out": 5.6297993659973145, "created_at": "2025-01-14T23:44:33.007970+00:00"} {"global_step": 7620, "acc_step": 0, "speed/wps": 12976.21342204524, "speed/FLOPS": 203809262979037.12, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.518497467041016, "optim/lr": 0.0029832468641763825, "optim/total_tokens": 1997537280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438769, "loss/out": 5.790517330169678, "created_at": "2025-01-14T23:44:38.064161+00:00"} {"global_step": 7621, "acc_step": 0, "speed/wps": 12974.755899938327, "speed/FLOPS": 203786370591502.78, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0009, "optim/grad_norm": 7.289455890655518, "optim/lr": 0.0029832340920884237, "optim/total_tokens": 1997799424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439586, "loss/out": 5.717308521270752, "created_at": "2025-01-14T23:44:43.120019+00:00"} {"global_step": 7622, "acc_step": 0, "speed/wps": 12975.50845623307, "speed/FLOPS": 203798190522232.44, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.219750165939331, "optim/lr": 0.0029832213151611527, "optim/total_tokens": 1998061568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462963, "loss/out": 5.5600738525390625, "created_at": "2025-01-14T23:44:48.180587+00:00"} {"global_step": 7623, "acc_step": 0, "speed/wps": 12979.096467785566, "speed/FLOPS": 203854545174108.94, "speed/curr_iter_time": 1.2715, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.948034286499023, "optim/lr": 0.0029832085333946117, "optim/total_tokens": 1998323712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 5.588542938232422, "created_at": "2025-01-14T23:44:53.239314+00:00"} {"global_step": 7624, "acc_step": 0, "speed/wps": 12976.027030985366, "speed/FLOPS": 203806335451313.62, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 107.08036804199219, "optim/lr": 0.0029831957467888414, "optim/total_tokens": 1998585856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 5.670966148376465, "created_at": "2025-01-14T23:44:58.302339+00:00"} {"global_step": 7625, "acc_step": 0, "speed/wps": 12970.932390345215, "speed/FLOPS": 203726317119290.2, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9355355501174927, "optim/lr": 0.0029831829553438843, "optim/total_tokens": 1998848000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 5.8028669357299805, "created_at": "2025-01-14T23:45:03.362152+00:00"} {"global_step": 7626, "acc_step": 0, "speed/wps": 12972.915285689109, "speed/FLOPS": 203757461215449.78, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.795891761779785, "optim/lr": 0.002983170159059782, "optim/total_tokens": 1999110144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440921, "loss/out": 5.835803031921387, "created_at": "2025-01-14T23:45:08.421720+00:00"} {"global_step": 7627, "acc_step": 0, "speed/wps": 12973.559220619998, "speed/FLOPS": 203767575098398.72, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0012, "optim/grad_norm": 36.35470199584961, "optim/lr": 0.002983157357936577, "optim/total_tokens": 1999372288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330623, "loss/out": 5.898148536682129, "created_at": "2025-01-14T23:45:13.509033+00:00"} {"global_step": 7628, "acc_step": 0, "speed/wps": 12969.876006536371, "speed/FLOPS": 203709725159948.84, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 162.6997528076172, "optim/lr": 0.002983144551974309, "optim/total_tokens": 1999634432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 6.189108371734619, "created_at": "2025-01-14T23:45:18.566732+00:00"} {"global_step": 7629, "acc_step": 0, "speed/wps": 12981.693032319105, "speed/FLOPS": 203895327788161.75, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.8443827629089355, "optim/lr": 0.0029831317411730222, "optim/total_tokens": 1999896576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461913, "loss/out": 6.406589984893799, "created_at": "2025-01-14T23:45:23.623866+00:00"} {"global_step": 7630, "acc_step": 0, "speed/wps": 12971.357591138752, "speed/FLOPS": 203732995482039.8, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.713265895843506, "optim/lr": 0.0029831189255327567, "optim/total_tokens": 2000158720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450227, "loss/out": 6.2271728515625, "created_at": "2025-01-14T23:45:28.680606+00:00"} {"global_step": 7631, "acc_step": 0, "speed/wps": 12977.876355032897, "speed/FLOPS": 203835381626717.66, "speed/curr_iter_time": 1.2722, "speed/data_load_time": 0.0007, "optim/grad_norm": 552.7214965820312, "optim/lr": 0.0029831061050535554, "optim/total_tokens": 2000420864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 6.212049961090088, "created_at": "2025-01-14T23:45:33.734265+00:00"} {"global_step": 7632, "acc_step": 0, "speed/wps": 12971.687231494543, "speed/FLOPS": 203738172936800.84, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.963919639587402, "optim/lr": 0.002983093279735459, "optim/total_tokens": 2000683008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444641, "loss/out": 6.149289608001709, "created_at": "2025-01-14T23:45:38.790581+00:00"} {"global_step": 7633, "acc_step": 0, "speed/wps": 12973.597894961902, "speed/FLOPS": 203768182532082.44, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.297979354858398, "optim/lr": 0.0029830804495785106, "optim/total_tokens": 2000945152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436080, "loss/out": 6.049415588378906, "created_at": "2025-01-14T23:45:43.845840+00:00"} {"global_step": 7634, "acc_step": 0, "speed/wps": 12970.803457212614, "speed/FLOPS": 203724292047270.03, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 18.193178176879883, "optim/lr": 0.0029830676145827513, "optim/total_tokens": 2001207296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439054, "loss/out": 5.966663837432861, "created_at": "2025-01-14T23:45:48.901995+00:00"} {"global_step": 7635, "acc_step": 0, "speed/wps": 12974.930132278649, "speed/FLOPS": 203789107149827.88, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.374302864074707, "optim/lr": 0.0029830547747482228, "optim/total_tokens": 2001469440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328221, "loss/out": 6.078900337219238, "created_at": "2025-01-14T23:45:53.962429+00:00"} {"global_step": 7636, "acc_step": 0, "speed/wps": 12974.436555109056, "speed/FLOPS": 203781354842133.3, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.6941134929656982, "optim/lr": 0.0029830419300749676, "optim/total_tokens": 2001731584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353982, "loss/out": 6.139163494110107, "created_at": "2025-01-14T23:45:59.020388+00:00"} {"global_step": 7637, "acc_step": 0, "speed/wps": 12976.735300511984, "speed/FLOPS": 203817459797493.5, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.395966529846191, "optim/lr": 0.002983029080563027, "optim/total_tokens": 2001993728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 404667, "loss/out": 6.047791481018066, "created_at": "2025-01-14T23:46:04.079018+00:00"} {"global_step": 7638, "acc_step": 0, "speed/wps": 12973.69086357926, "speed/FLOPS": 203769642731970.34, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 217.42884826660156, "optim/lr": 0.002983016226212444, "optim/total_tokens": 2002255872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436096, "loss/out": 6.003645896911621, "created_at": "2025-01-14T23:46:09.135388+00:00"} {"global_step": 7639, "acc_step": 0, "speed/wps": 12977.621405531549, "speed/FLOPS": 203831377294461.66, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.223355531692505, "optim/lr": 0.0029830033670232596, "optim/total_tokens": 2002518016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454182, "loss/out": 6.18803071975708, "created_at": "2025-01-14T23:46:14.193439+00:00"} {"global_step": 7640, "acc_step": 0, "speed/wps": 12976.472207162731, "speed/FLOPS": 203813327554915.34, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.6736066341400146, "optim/lr": 0.0029829905029955155, "optim/total_tokens": 2002780160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481474, "loss/out": 5.889758586883545, "created_at": "2025-01-14T23:46:19.251760+00:00"} {"global_step": 7641, "acc_step": 0, "speed/wps": 12973.854724369916, "speed/FLOPS": 203772216390855.9, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.52930212020874, "optim/lr": 0.002982977634129254, "optim/total_tokens": 2003042304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418566, "loss/out": 6.0052313804626465, "created_at": "2025-01-14T23:46:24.314785+00:00"} {"global_step": 7642, "acc_step": 0, "speed/wps": 12977.26556618719, "speed/FLOPS": 203825788348584.0, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.169214248657227, "optim/lr": 0.0029829647604245175, "optim/total_tokens": 2003304448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480954, "loss/out": 5.993416786193848, "created_at": "2025-01-14T23:46:29.369643+00:00"} {"global_step": 7643, "acc_step": 0, "speed/wps": 12971.444841623419, "speed/FLOPS": 203734365870796.25, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.3545773029327393, "optim/lr": 0.0029829518818813476, "optim/total_tokens": 2003566592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426544, "loss/out": 5.914831161499023, "created_at": "2025-01-14T23:46:34.437077+00:00"} {"global_step": 7644, "acc_step": 0, "speed/wps": 12977.707450257312, "speed/FLOPS": 203832728745117.62, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.284114360809326, "optim/lr": 0.002982938998499787, "optim/total_tokens": 2003828736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305035, "loss/out": 5.834529876708984, "created_at": "2025-01-14T23:46:39.497066+00:00"} {"global_step": 7645, "acc_step": 0, "speed/wps": 12970.40853520032, "speed/FLOPS": 203718089254387.22, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.7844904661178589, "optim/lr": 0.0029829261102798763, "optim/total_tokens": 2004090880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448890, "loss/out": 5.939660549163818, "created_at": "2025-01-14T23:46:44.568172+00:00"} {"global_step": 7646, "acc_step": 0, "speed/wps": 12975.909039042985, "speed/FLOPS": 203804482225720.62, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.054738521575928, "optim/lr": 0.002982913217221659, "optim/total_tokens": 2004353024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 5.73002290725708, "created_at": "2025-01-14T23:46:49.627070+00:00"} {"global_step": 7647, "acc_step": 0, "speed/wps": 12975.415928113598, "speed/FLOPS": 203796737240969.2, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 24.351673126220703, "optim/lr": 0.002982900319325176, "optim/total_tokens": 2004615168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 5.848513126373291, "created_at": "2025-01-14T23:46:54.681749+00:00"} {"global_step": 7648, "acc_step": 0, "speed/wps": 12973.719913555518, "speed/FLOPS": 203770099001766.3, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.931868076324463, "optim/lr": 0.00298288741659047, "optim/total_tokens": 2004877312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457361, "loss/out": 5.908127784729004, "created_at": "2025-01-14T23:46:59.737142+00:00"} {"global_step": 7649, "acc_step": 0, "speed/wps": 12971.159224392413, "speed/FLOPS": 203729879859704.1, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.789771556854248, "optim/lr": 0.0029828745090175832, "optim/total_tokens": 2005139456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421476, "loss/out": 5.839133262634277, "created_at": "2025-01-14T23:47:04.798475+00:00"} {"global_step": 7650, "acc_step": 0, "speed/wps": 12973.649332744042, "speed/FLOPS": 203768990433133.88, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.313760757446289, "optim/lr": 0.0029828615966065583, "optim/total_tokens": 2005401600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 5.8891754150390625, "created_at": "2025-01-14T23:47:09.854225+00:00"} {"global_step": 7651, "acc_step": 0, "speed/wps": 12973.007907297833, "speed/FLOPS": 203758915965090.25, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0009, "optim/grad_norm": 6.391777992248535, "optim/lr": 0.0029828486793574355, "optim/total_tokens": 2005663744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443543, "loss/out": 5.734846115112305, "created_at": "2025-01-14T23:47:14.910029+00:00"} {"global_step": 7652, "acc_step": 0, "speed/wps": 12971.10960914635, "speed/FLOPS": 203729100584086.16, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.361722946166992, "optim/lr": 0.002982835757270259, "optim/total_tokens": 2005925888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 341862, "loss/out": 5.7345733642578125, "created_at": "2025-01-14T23:47:19.971534+00:00"} {"global_step": 7653, "acc_step": 0, "speed/wps": 12974.84777988712, "speed/FLOPS": 203787813692354.88, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.784390926361084, "optim/lr": 0.00298282283034507, "optim/total_tokens": 2006188032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332595, "loss/out": 5.773892402648926, "created_at": "2025-01-14T23:47:25.027334+00:00"} {"global_step": 7654, "acc_step": 0, "speed/wps": 12971.14119396103, "speed/FLOPS": 203729596667002.8, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.356750965118408, "optim/lr": 0.0029828098985819106, "optim/total_tokens": 2006450176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 389608, "loss/out": 5.784311771392822, "created_at": "2025-01-14T23:47:30.084748+00:00"} {"global_step": 7655, "acc_step": 0, "speed/wps": 12969.607898856511, "speed/FLOPS": 203705514160418.06, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 14.461593627929688, "optim/lr": 0.0029827969619808234, "optim/total_tokens": 2006712320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449657, "loss/out": 5.7564826011657715, "created_at": "2025-01-14T23:47:35.163743+00:00"} {"global_step": 7656, "acc_step": 0, "speed/wps": 12971.517328085865, "speed/FLOPS": 203735504370295.8, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.483456611633301, "optim/lr": 0.00298278402054185, "optim/total_tokens": 2006974464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 404430, "loss/out": 5.805909156799316, "created_at": "2025-01-14T23:47:40.219564+00:00"} {"global_step": 7657, "acc_step": 0, "speed/wps": 12979.128141780977, "speed/FLOPS": 203855042657731.5, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.9386348724365234, "optim/lr": 0.0029827710742650333, "optim/total_tokens": 2007236608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 5.720099449157715, "created_at": "2025-01-14T23:47:45.277509+00:00"} {"global_step": 7658, "acc_step": 0, "speed/wps": 12972.5612229797, "speed/FLOPS": 203751900174065.38, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.182152271270752, "optim/lr": 0.0029827581231504155, "optim/total_tokens": 2007498752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432668, "loss/out": 5.699197769165039, "created_at": "2025-01-14T23:47:50.333305+00:00"} {"global_step": 7659, "acc_step": 0, "speed/wps": 12975.306744572352, "speed/FLOPS": 203795022363420.0, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0008, "optim/grad_norm": 14.841435432434082, "optim/lr": 0.002982745167198038, "optim/total_tokens": 2007760896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435028, "loss/out": 5.760637283325195, "created_at": "2025-01-14T23:47:55.393006+00:00"} {"global_step": 7660, "acc_step": 0, "speed/wps": 12968.708334523704, "speed/FLOPS": 203691385266440.4, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.427140712738037, "optim/lr": 0.0029827322064079433, "optim/total_tokens": 2008023040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317507, "loss/out": 5.758297443389893, "created_at": "2025-01-14T23:48:00.455428+00:00"} {"global_step": 7661, "acc_step": 0, "speed/wps": 12977.020005076944, "speed/FLOPS": 203821931473911.25, "speed/curr_iter_time": 1.2721, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.083953857421875, "optim/lr": 0.0029827192407801753, "optim/total_tokens": 2008285184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299119, "loss/out": 5.7079548835754395, "created_at": "2025-01-14T23:48:05.509468+00:00"} {"global_step": 7662, "acc_step": 0, "speed/wps": 12960.901585890128, "speed/FLOPS": 203568769551553.28, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0009, "optim/grad_norm": 48.08882141113281, "optim/lr": 0.002982706270314774, "optim/total_tokens": 2008547328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354800, "loss/out": 5.693731307983398, "created_at": "2025-01-14T23:48:10.569606+00:00"} {"global_step": 7663, "acc_step": 0, "speed/wps": 12972.050661112, "speed/FLOPS": 203743881098347.62, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.796886920928955, "optim/lr": 0.0029826932950117835, "optim/total_tokens": 2008809472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450746, "loss/out": 5.651191711425781, "created_at": "2025-01-14T23:48:15.627388+00:00"} {"global_step": 7664, "acc_step": 0, "speed/wps": 12963.422774233613, "speed/FLOPS": 203608368278963.56, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 32.76729965209961, "optim/lr": 0.002982680314871245, "optim/total_tokens": 2009071616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474123, "loss/out": 5.516725540161133, "created_at": "2025-01-14T23:48:20.687316+00:00"} {"global_step": 7665, "acc_step": 0, "speed/wps": 12966.24584896763, "speed/FLOPS": 203652708546972.5, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.489850044250488, "optim/lr": 0.002982667329893201, "optim/total_tokens": 2009333760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 5.817427635192871, "created_at": "2025-01-14T23:48:25.776038+00:00"} {"global_step": 7666, "acc_step": 0, "speed/wps": 12964.552698801977, "speed/FLOPS": 203626115296988.8, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.025038719177246, "optim/lr": 0.0029826543400776947, "optim/total_tokens": 2009595904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425492, "loss/out": 5.730658531188965, "created_at": "2025-01-14T23:48:30.835399+00:00"} {"global_step": 7667, "acc_step": 0, "speed/wps": 12962.041760386654, "speed/FLOPS": 203586677558785.2, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.352428913116455, "optim/lr": 0.0029826413454247676, "optim/total_tokens": 2009858048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302050, "loss/out": 5.6544108390808105, "created_at": "2025-01-14T23:48:35.896784+00:00"} {"global_step": 7668, "acc_step": 0, "speed/wps": 12965.781177457306, "speed/FLOPS": 203645410242377.3, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.1834404468536377, "optim/lr": 0.0029826283459344623, "optim/total_tokens": 2010120192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352115, "loss/out": 5.747781753540039, "created_at": "2025-01-14T23:48:40.955979+00:00"} {"global_step": 7669, "acc_step": 0, "speed/wps": 12952.494974752544, "speed/FLOPS": 203436732171743.28, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.388080596923828, "optim/lr": 0.0029826153416068215, "optim/total_tokens": 2010382336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452093, "loss/out": 5.598464012145996, "created_at": "2025-01-14T23:48:46.044801+00:00"} {"global_step": 7670, "acc_step": 0, "speed/wps": 12962.784721215341, "speed/FLOPS": 203598346779534.88, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.024566650390625, "optim/lr": 0.002982602332441888, "optim/total_tokens": 2010644480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461913, "loss/out": 5.662773609161377, "created_at": "2025-01-14T23:48:51.109123+00:00"} {"global_step": 7671, "acc_step": 0, "speed/wps": 12967.273186574112, "speed/FLOPS": 203668844295792.12, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.355442523956299, "optim/lr": 0.002982589318439703, "optim/total_tokens": 2010906624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431605, "loss/out": 5.61058235168457, "created_at": "2025-01-14T23:48:56.167765+00:00"} {"global_step": 7672, "acc_step": 0, "speed/wps": 12964.005876495483, "speed/FLOPS": 203617526701252.72, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.243021011352539, "optim/lr": 0.0029825762996003096, "optim/total_tokens": 2011168768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293489, "loss/out": 5.560620307922363, "created_at": "2025-01-14T23:49:01.226926+00:00"} {"global_step": 7673, "acc_step": 0, "speed/wps": 12959.201998294888, "speed/FLOPS": 203542075192891.97, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.258397102355957, "optim/lr": 0.00298256327592375, "optim/total_tokens": 2011430912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 346546, "loss/out": 5.684808731079102, "created_at": "2025-01-14T23:49:06.291308+00:00"} {"global_step": 7674, "acc_step": 0, "speed/wps": 12972.92462743211, "speed/FLOPS": 203757607940360.53, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.104719638824463, "optim/lr": 0.0029825502474100677, "optim/total_tokens": 2011693056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353457, "loss/out": 5.805677890777588, "created_at": "2025-01-14T23:49:11.351713+00:00"} {"global_step": 7675, "acc_step": 0, "speed/wps": 12966.873060316912, "speed/FLOPS": 203662559763091.7, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 32.44154357910156, "optim/lr": 0.002982537214059305, "optim/total_tokens": 2011955200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430283, "loss/out": 5.681260108947754, "created_at": "2025-01-14T23:49:16.413899+00:00"} {"global_step": 7676, "acc_step": 0, "speed/wps": 12962.168826527068, "speed/FLOPS": 203588673307127.56, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 27.905412673950195, "optim/lr": 0.0029825241758715025, "optim/total_tokens": 2012217344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 5.658058166503906, "created_at": "2025-01-14T23:49:21.473978+00:00"} {"global_step": 7677, "acc_step": 0, "speed/wps": 12961.737676572558, "speed/FLOPS": 203581901504629.88, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.301274299621582, "optim/lr": 0.0029825111328467056, "optim/total_tokens": 2012479488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427332, "loss/out": 5.792576313018799, "created_at": "2025-01-14T23:49:26.537128+00:00"} {"global_step": 7678, "acc_step": 0, "speed/wps": 12962.899804870116, "speed/FLOPS": 203600154326474.62, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.359692573547363, "optim/lr": 0.002982498084984955, "optim/total_tokens": 2012741632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 5.7117018699646, "created_at": "2025-01-14T23:49:31.596376+00:00"} {"global_step": 7679, "acc_step": 0, "speed/wps": 12969.281344872887, "speed/FLOPS": 203700385181373.06, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.4490408897399902, "optim/lr": 0.002982485032286293, "optim/total_tokens": 2013003776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427047, "loss/out": 5.715627193450928, "created_at": "2025-01-14T23:49:36.654841+00:00"} {"global_step": 7680, "acc_step": 0, "speed/wps": 12958.283782220395, "speed/FLOPS": 203527653347680.94, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.202327251434326, "optim/lr": 0.0029824719747507636, "optim/total_tokens": 2013265920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320284, "loss/out": 5.654262542724609, "created_at": "2025-01-14T23:49:41.716213+00:00"} {"global_step": 7681, "acc_step": 0, "speed/wps": 12965.809781111278, "speed/FLOPS": 203645859502068.47, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.867847442626953, "optim/lr": 0.002982458912378409, "optim/total_tokens": 2013528064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425464, "loss/out": 5.678577423095703, "created_at": "2025-01-14T23:49:46.774474+00:00"} {"global_step": 7682, "acc_step": 0, "speed/wps": 12962.524994129126, "speed/FLOPS": 203594267408743.47, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.63449764251709, "optim/lr": 0.0029824458451692707, "optim/total_tokens": 2013790208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464007, "loss/out": 5.728673934936523, "created_at": "2025-01-14T23:49:51.836845+00:00"} {"global_step": 7683, "acc_step": 0, "speed/wps": 12961.054285576916, "speed/FLOPS": 203571167909965.66, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.173203945159912, "optim/lr": 0.0029824327731233923, "optim/total_tokens": 2014052352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460840, "loss/out": 5.824484348297119, "created_at": "2025-01-14T23:49:56.901871+00:00"} {"global_step": 7684, "acc_step": 0, "speed/wps": 12964.56739559851, "speed/FLOPS": 203626346130373.84, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.397674560546875, "optim/lr": 0.0029824196962408168, "optim/total_tokens": 2014314496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436381, "loss/out": 5.684820175170898, "created_at": "2025-01-14T23:50:01.963097+00:00"} {"global_step": 7685, "acc_step": 0, "speed/wps": 12963.081429822709, "speed/FLOPS": 203603006995933.94, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.860856533050537, "optim/lr": 0.0029824066145215863, "optim/total_tokens": 2014576640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 5.722829341888428, "created_at": "2025-01-14T23:50:07.030667+00:00"} {"global_step": 7686, "acc_step": 0, "speed/wps": 12957.766238113882, "speed/FLOPS": 203519524606308.03, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.6497702598571777, "optim/lr": 0.0029823935279657435, "optim/total_tokens": 2014838784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345608, "loss/out": 5.743481636047363, "created_at": "2025-01-14T23:50:12.097643+00:00"} {"global_step": 7687, "acc_step": 0, "speed/wps": 12947.506976226628, "speed/FLOPS": 203358388800665.53, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0012, "optim/grad_norm": 3.443183660507202, "optim/lr": 0.002982380436573331, "optim/total_tokens": 2015100928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 339699, "loss/out": 5.684666633605957, "created_at": "2025-01-14T23:50:17.167029+00:00"} {"global_step": 7688, "acc_step": 0, "speed/wps": 12966.220356476975, "speed/FLOPS": 203652308152376.44, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5646650791168213, "optim/lr": 0.0029823673403443916, "optim/total_tokens": 2015363072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444359, "loss/out": 5.637802600860596, "created_at": "2025-01-14T23:50:22.226350+00:00"} {"global_step": 7689, "acc_step": 0, "speed/wps": 12953.40834564481, "speed/FLOPS": 203451077916709.47, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.889641761779785, "optim/lr": 0.0029823542392789686, "optim/total_tokens": 2015625216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455808, "loss/out": 5.595094680786133, "created_at": "2025-01-14T23:50:27.291611+00:00"} {"global_step": 7690, "acc_step": 0, "speed/wps": 12965.976756534259, "speed/FLOPS": 203648482080534.66, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.836212158203125, "optim/lr": 0.002982341133377104, "optim/total_tokens": 2015887360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445976, "loss/out": 5.827824592590332, "created_at": "2025-01-14T23:50:32.350961+00:00"} {"global_step": 7691, "acc_step": 0, "speed/wps": 12961.694271640608, "speed/FLOPS": 203581219770528.97, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 11.004140853881836, "optim/lr": 0.0029823280226388403, "optim/total_tokens": 2016149504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443307, "loss/out": 5.765377998352051, "created_at": "2025-01-14T23:50:37.412175+00:00"} {"global_step": 7692, "acc_step": 0, "speed/wps": 12960.27652032322, "speed/FLOPS": 203558952037894.9, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.808080673217773, "optim/lr": 0.002982314907064221, "optim/total_tokens": 2016411648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450990, "loss/out": 5.796168327331543, "created_at": "2025-01-14T23:50:42.477962+00:00"} {"global_step": 7693, "acc_step": 0, "speed/wps": 12954.709062850368, "speed/FLOPS": 203471507467795.5, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 21.45697593688965, "optim/lr": 0.0029823017866532887, "optim/total_tokens": 2016673792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306907, "loss/out": 5.604928016662598, "created_at": "2025-01-14T23:50:47.541062+00:00"} {"global_step": 7694, "acc_step": 0, "speed/wps": 12957.198708730844, "speed/FLOPS": 203510610777480.3, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 216.0966796875, "optim/lr": 0.002982288661406086, "optim/total_tokens": 2016935936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372023, "loss/out": 5.683480739593506, "created_at": "2025-01-14T23:50:52.605139+00:00"} {"global_step": 7695, "acc_step": 0, "speed/wps": 12955.950115351716, "speed/FLOPS": 203490999902714.0, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.5590102672576904, "optim/lr": 0.002982275531322657, "optim/total_tokens": 2017198080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457665, "loss/out": 5.879489898681641, "created_at": "2025-01-14T23:50:57.667110+00:00"} {"global_step": 7696, "acc_step": 0, "speed/wps": 12950.259207001704, "speed/FLOPS": 203401616366949.5, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0008, "optim/grad_norm": 8.999627113342285, "optim/lr": 0.0029822623964030426, "optim/total_tokens": 2017460224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459237, "loss/out": 5.756260871887207, "created_at": "2025-01-14T23:51:02.734317+00:00"} {"global_step": 7697, "acc_step": 0, "speed/wps": 12957.121171580346, "speed/FLOPS": 203509392949984.53, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.344846725463867, "optim/lr": 0.002982249256647286, "optim/total_tokens": 2017722368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 5.8651933670043945, "created_at": "2025-01-14T23:51:07.797213+00:00"} {"global_step": 7698, "acc_step": 0, "speed/wps": 12951.458106340087, "speed/FLOPS": 203420446728519.16, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.95000696182251, "optim/lr": 0.0029822361120554314, "optim/total_tokens": 2017984512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438531, "loss/out": 5.753653049468994, "created_at": "2025-01-14T23:51:12.862214+00:00"} {"global_step": 7699, "acc_step": 0, "speed/wps": 12954.29637111964, "speed/FLOPS": 203465025577066.56, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 35.65391159057617, "optim/lr": 0.0029822229626275203, "optim/total_tokens": 2018246656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293833, "loss/out": 5.786909103393555, "created_at": "2025-01-14T23:51:17.925893+00:00"} {"global_step": 7700, "acc_step": 0, "speed/wps": 12956.64695758444, "speed/FLOPS": 203501944767540.53, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 26.703876495361328, "optim/lr": 0.002982209808363596, "optim/total_tokens": 2018508800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 347078, "loss/out": 5.668500900268555, "created_at": "2025-01-14T23:51:22.991401+00:00"} {"global_step": 7701, "acc_step": 0, "speed/wps": 12955.367021969882, "speed/FLOPS": 203481841619898.03, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.626669883728027, "optim/lr": 0.0029821966492637017, "optim/total_tokens": 2018770944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436151, "loss/out": 5.870176792144775, "created_at": "2025-01-14T23:51:28.061701+00:00"} {"global_step": 7702, "acc_step": 0, "speed/wps": 12954.480931676091, "speed/FLOPS": 203467924354218.38, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0009, "optim/grad_norm": 8.68172550201416, "optim/lr": 0.0029821834853278814, "optim/total_tokens": 2019033088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 5.62801456451416, "created_at": "2025-01-14T23:51:33.125834+00:00"} {"global_step": 7703, "acc_step": 0, "speed/wps": 12953.268444291181, "speed/FLOPS": 203448880573706.12, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 27.11077117919922, "optim/lr": 0.0029821703165561755, "optim/total_tokens": 2019295232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433179, "loss/out": 5.69605827331543, "created_at": "2025-01-14T23:51:38.189083+00:00"} {"global_step": 7704, "acc_step": 0, "speed/wps": 12946.149284263733, "speed/FLOPS": 203337064382724.0, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.935639381408691, "optim/lr": 0.0029821571429486287, "optim/total_tokens": 2019557376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 5.700748443603516, "created_at": "2025-01-14T23:51:43.260196+00:00"} {"global_step": 7705, "acc_step": 0, "speed/wps": 12958.7233221247, "speed/FLOPS": 203534556925869.2, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.081262111663818, "optim/lr": 0.002982143964505284, "optim/total_tokens": 2019819520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337010, "loss/out": 5.752096176147461, "created_at": "2025-01-14T23:51:48.322268+00:00"} {"global_step": 7706, "acc_step": 0, "speed/wps": 12953.102387335912, "speed/FLOPS": 203446272421037.44, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5708413124084473, "optim/lr": 0.0029821307812261835, "optim/total_tokens": 2020081664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356961, "loss/out": 5.5906853675842285, "created_at": "2025-01-14T23:51:53.387722+00:00"} {"global_step": 7707, "acc_step": 0, "speed/wps": 12951.388916488275, "speed/FLOPS": 203419360006819.62, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.518770456314087, "optim/lr": 0.0029821175931113706, "optim/total_tokens": 2020343808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461385, "loss/out": 5.645001411437988, "created_at": "2025-01-14T23:51:58.452866+00:00"} {"global_step": 7708, "acc_step": 0, "speed/wps": 12944.625396768128, "speed/FLOPS": 203313129635564.66, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.165246963500977, "optim/lr": 0.0029821044001608885, "optim/total_tokens": 2020605952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435053, "loss/out": 5.642017364501953, "created_at": "2025-01-14T23:52:03.524073+00:00"} {"global_step": 7709, "acc_step": 0, "speed/wps": 12949.447873472654, "speed/FLOPS": 203388873258986.7, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.6018927097320557, "optim/lr": 0.00298209120237478, "optim/total_tokens": 2020868096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461077, "loss/out": 5.607356548309326, "created_at": "2025-01-14T23:52:08.593795+00:00"} {"global_step": 7710, "acc_step": 0, "speed/wps": 12951.485054492809, "speed/FLOPS": 203420869986288.28, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.677412748336792, "optim/lr": 0.0029820779997530894, "optim/total_tokens": 2021130240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301284, "loss/out": 5.677499771118164, "created_at": "2025-01-14T23:52:13.657984+00:00"} {"global_step": 7711, "acc_step": 0, "speed/wps": 12950.418760784682, "speed/FLOPS": 203404122378361.03, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.931748867034912, "optim/lr": 0.0029820647922958574, "optim/total_tokens": 2021392384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440390, "loss/out": 5.705657482147217, "created_at": "2025-01-14T23:52:18.722038+00:00"} {"global_step": 7712, "acc_step": 0, "speed/wps": 12950.475270259642, "speed/FLOPS": 203405009937317.44, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.8293380737304688, "optim/lr": 0.002982051580003129, "optim/total_tokens": 2021654528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444881, "loss/out": 5.589876174926758, "created_at": "2025-01-14T23:52:23.790979+00:00"} {"global_step": 7713, "acc_step": 0, "speed/wps": 12948.13222108221, "speed/FLOPS": 203368209130298.94, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 8.653247833251953, "optim/lr": 0.0029820383628749473, "optim/total_tokens": 2021916672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476715, "loss/out": 5.673130512237549, "created_at": "2025-01-14T23:52:28.856425+00:00"} {"global_step": 7714, "acc_step": 0, "speed/wps": 12947.89793896987, "speed/FLOPS": 203364529407787.28, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.0624125003814697, "optim/lr": 0.0029820251409113543, "optim/total_tokens": 2022178816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 5.645798206329346, "created_at": "2025-01-14T23:52:33.932194+00:00"} {"global_step": 7715, "acc_step": 0, "speed/wps": 12952.941587201967, "speed/FLOPS": 203443746833971.3, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.3920180797576904, "optim/lr": 0.0029820119141123935, "optim/total_tokens": 2022440960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 5.623324871063232, "created_at": "2025-01-14T23:52:39.002218+00:00"} {"global_step": 7716, "acc_step": 0, "speed/wps": 12945.127984264964, "speed/FLOPS": 203321023462830.06, "speed/curr_iter_time": 1.2778, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.795531272888184, "optim/lr": 0.0029819986824781083, "optim/total_tokens": 2022703104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302050, "loss/out": 5.560579776763916, "created_at": "2025-01-14T23:52:44.071482+00:00"} {"global_step": 7717, "acc_step": 0, "speed/wps": 12954.303475375567, "speed/FLOPS": 203465137159168.16, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.876334190368652, "optim/lr": 0.0029819854460085425, "optim/total_tokens": 2022965248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451002, "loss/out": 5.631642818450928, "created_at": "2025-01-14T23:52:49.138397+00:00"} {"global_step": 7718, "acc_step": 0, "speed/wps": 12954.748776816501, "speed/FLOPS": 203472131230206.3, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.05259370803833, "optim/lr": 0.0029819722047037373, "optim/total_tokens": 2023227392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 5.722491264343262, "created_at": "2025-01-14T23:52:54.204499+00:00"} {"global_step": 7719, "acc_step": 0, "speed/wps": 12951.11426204639, "speed/FLOPS": 203415046181395.06, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.698647499084473, "optim/lr": 0.0029819589585637378, "optim/total_tokens": 2023489536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429217, "loss/out": 5.728550434112549, "created_at": "2025-01-14T23:52:59.273441+00:00"} {"global_step": 7720, "acc_step": 0, "speed/wps": 12955.061047449462, "speed/FLOPS": 203477035869601.78, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.616145133972168, "optim/lr": 0.0029819457075885873, "optim/total_tokens": 2023751680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442256, "loss/out": 5.583522796630859, "created_at": "2025-01-14T23:53:04.339482+00:00"} {"global_step": 7721, "acc_step": 0, "speed/wps": 12956.184591148622, "speed/FLOPS": 203494682667307.94, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.615926742553711, "optim/lr": 0.0029819324517783275, "optim/total_tokens": 2024013824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293250, "loss/out": 5.757449626922607, "created_at": "2025-01-14T23:53:09.401791+00:00"} {"global_step": 7722, "acc_step": 0, "speed/wps": 12949.478073208948, "speed/FLOPS": 203389347587343.94, "speed/curr_iter_time": 1.2774, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.278181076049805, "optim/lr": 0.0029819191911330026, "optim/total_tokens": 2024275968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 5.666079521179199, "created_at": "2025-01-14T23:53:14.476153+00:00"} {"global_step": 7723, "acc_step": 0, "speed/wps": 12960.15067281181, "speed/FLOPS": 203556975429795.6, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.577259540557861, "optim/lr": 0.0029819059256526565, "optim/total_tokens": 2024538112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460271, "loss/out": 5.676357746124268, "created_at": "2025-01-14T23:53:19.540940+00:00"} {"global_step": 7724, "acc_step": 0, "speed/wps": 12953.57808020937, "speed/FLOPS": 203453743831285.62, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4097399711608887, "optim/lr": 0.0029818926553373307, "optim/total_tokens": 2024800256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460287, "loss/out": 5.649642467498779, "created_at": "2025-01-14T23:53:24.605098+00:00"} {"global_step": 7725, "acc_step": 0, "speed/wps": 12951.994072646503, "speed/FLOPS": 203428864815856.28, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.661720275878906, "optim/lr": 0.0029818793801870704, "optim/total_tokens": 2025062400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 5.72598934173584, "created_at": "2025-01-14T23:53:29.674520+00:00"} {"global_step": 7726, "acc_step": 0, "speed/wps": 12958.364719410165, "speed/FLOPS": 203528924577465.62, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 26.06694984436035, "optim/lr": 0.0029818661002019176, "optim/total_tokens": 2025324544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338354, "loss/out": 5.841053485870361, "created_at": "2025-01-14T23:53:34.735825+00:00"} {"global_step": 7727, "acc_step": 0, "speed/wps": 12948.409646310738, "speed/FLOPS": 203372566474735.16, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.06611442565918, "optim/lr": 0.002981852815381916, "optim/total_tokens": 2025586688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322598, "loss/out": 5.897171497344971, "created_at": "2025-01-14T23:53:39.801147+00:00"} {"global_step": 7728, "acc_step": 0, "speed/wps": 12952.550810949087, "speed/FLOPS": 203437609155936.97, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.001, "optim/grad_norm": 1.955257534980774, "optim/lr": 0.0029818395257271094, "optim/total_tokens": 2025848832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358828, "loss/out": 5.879412651062012, "created_at": "2025-01-14T23:53:44.865350+00:00"} {"global_step": 7729, "acc_step": 0, "speed/wps": 12948.641792983972, "speed/FLOPS": 203376212657241.25, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0009, "optim/grad_norm": 7.320980548858643, "optim/lr": 0.0029818262312375403, "optim/total_tokens": 2026110976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432941, "loss/out": 5.641676902770996, "created_at": "2025-01-14T23:53:49.934483+00:00"} {"global_step": 7730, "acc_step": 0, "speed/wps": 12950.613876671663, "speed/FLOPS": 203407186941483.4, "speed/curr_iter_time": 1.2782, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.1378366947174072, "optim/lr": 0.0029818129319132526, "optim/total_tokens": 2026373120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456556, "loss/out": 5.7587785720825195, "created_at": "2025-01-14T23:53:54.998798+00:00"} {"global_step": 7731, "acc_step": 0, "speed/wps": 12950.722797581137, "speed/FLOPS": 203408897694039.72, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.4981210231781006, "optim/lr": 0.00298179962775429, "optim/total_tokens": 2026635264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456851, "loss/out": 5.5844221115112305, "created_at": "2025-01-14T23:54:00.063029+00:00"} {"global_step": 7732, "acc_step": 0, "speed/wps": 12948.699466397187, "speed/FLOPS": 203377118497447.38, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.1112589836120605, "optim/lr": 0.0029817863187606955, "optim/total_tokens": 2026897408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 5.626729965209961, "created_at": "2025-01-14T23:54:05.131924+00:00"} {"global_step": 7733, "acc_step": 0, "speed/wps": 12950.823768965001, "speed/FLOPS": 203410483588372.53, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.497909545898438, "optim/lr": 0.002981773004932513, "optim/total_tokens": 2027159552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351583, "loss/out": 5.754510879516602, "created_at": "2025-01-14T23:54:10.195914+00:00"} {"global_step": 7734, "acc_step": 0, "speed/wps": 12953.390517132904, "speed/FLOPS": 203450797895431.03, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.863629698753357, "optim/lr": 0.0029817596862697845, "optim/total_tokens": 2027421696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422813, "loss/out": 5.61968469619751, "created_at": "2025-01-14T23:54:15.262074+00:00"} {"global_step": 7735, "acc_step": 0, "speed/wps": 12950.36299540163, "speed/FLOPS": 203403246506390.56, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7037094831466675, "optim/lr": 0.002981746362772555, "optim/total_tokens": 2027683840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 5.654080867767334, "created_at": "2025-01-14T23:54:20.330116+00:00"} {"global_step": 7736, "acc_step": 0, "speed/wps": 12960.193554212137, "speed/FLOPS": 203557648941112.97, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.198664665222168, "optim/lr": 0.0029817330344408676, "optim/total_tokens": 2027945984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433993, "loss/out": 5.618656158447266, "created_at": "2025-01-14T23:54:25.391697+00:00"} {"global_step": 7737, "acc_step": 0, "speed/wps": 12956.008166374575, "speed/FLOPS": 203491911673798.47, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.938563346862793, "optim/lr": 0.002981719701274765, "optim/total_tokens": 2028208128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 5.5746893882751465, "created_at": "2025-01-14T23:54:30.453876+00:00"} {"global_step": 7738, "acc_step": 0, "speed/wps": 12954.601381798046, "speed/FLOPS": 203469816188899.5, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.0699310302734375, "optim/lr": 0.002981706363274292, "optim/total_tokens": 2028470272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425730, "loss/out": 5.681365966796875, "created_at": "2025-01-14T23:54:35.521202+00:00"} {"global_step": 7739, "acc_step": 0, "speed/wps": 12949.906406090402, "speed/FLOPS": 203396075143838.25, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0009, "optim/grad_norm": 30.66353416442871, "optim/lr": 0.0029816930204394913, "optim/total_tokens": 2028732416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 341986, "loss/out": 5.57847785949707, "created_at": "2025-01-14T23:54:40.587755+00:00"} {"global_step": 7740, "acc_step": 0, "speed/wps": 12947.48039958373, "speed/FLOPS": 203357971377968.7, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8560633659362793, "optim/lr": 0.0029816796727704068, "optim/total_tokens": 2028994560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421165, "loss/out": 5.669548511505127, "created_at": "2025-01-14T23:54:45.653425+00:00"} {"global_step": 7741, "acc_step": 0, "speed/wps": 12951.89331186699, "speed/FLOPS": 203427282229354.22, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.984642028808594, "optim/lr": 0.002981666320267082, "optim/total_tokens": 2029256704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425201, "loss/out": 5.624251365661621, "created_at": "2025-01-14T23:54:50.716835+00:00"} {"global_step": 7742, "acc_step": 0, "speed/wps": 12956.095315447463, "speed/FLOPS": 203493280469743.56, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.0172579288482666, "optim/lr": 0.0029816529629295594, "optim/total_tokens": 2029518848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 5.633302688598633, "created_at": "2025-01-14T23:54:55.779044+00:00"} {"global_step": 7743, "acc_step": 0, "speed/wps": 12952.26327636937, "speed/FLOPS": 203433093030248.6, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 21.034698486328125, "optim/lr": 0.002981639600757884, "optim/total_tokens": 2029780992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 5.542794227600098, "created_at": "2025-01-14T23:55:00.847433+00:00"} {"global_step": 7744, "acc_step": 0, "speed/wps": 12952.580665153986, "speed/FLOPS": 203438078057248.88, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.507978916168213, "optim/lr": 0.0029816262337520986, "optim/total_tokens": 2030043136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457095, "loss/out": 5.785587310791016, "created_at": "2025-01-14T23:55:05.910803+00:00"} {"global_step": 7745, "acc_step": 0, "speed/wps": 12956.586932699565, "speed/FLOPS": 203501001994240.72, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6273326873779297, "optim/lr": 0.0029816128619122475, "optim/total_tokens": 2030305280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307899, "loss/out": 5.741794586181641, "created_at": "2025-01-14T23:55:10.978872+00:00"} {"global_step": 7746, "acc_step": 0, "speed/wps": 12952.77682133612, "speed/FLOPS": 203441158959636.12, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.654495120048523, "optim/lr": 0.0029815994852383735, "optim/total_tokens": 2030567424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438233, "loss/out": 5.635212421417236, "created_at": "2025-01-14T23:55:16.047730+00:00"} {"global_step": 7747, "acc_step": 0, "speed/wps": 12954.003348218428, "speed/FLOPS": 203460423249747.34, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6530356407165527, "optim/lr": 0.002981586103730521, "optim/total_tokens": 2030829568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 5.6483917236328125, "created_at": "2025-01-14T23:55:21.111524+00:00"} {"global_step": 7748, "acc_step": 0, "speed/wps": 12959.541856764028, "speed/FLOPS": 203547413137164.28, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5319018363952637, "optim/lr": 0.002981572717388733, "optim/total_tokens": 2031091712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 5.676423072814941, "created_at": "2025-01-14T23:55:26.172406+00:00"} {"global_step": 7749, "acc_step": 0, "speed/wps": 12956.070984006794, "speed/FLOPS": 203492898311032.94, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.851426601409912, "optim/lr": 0.0029815593262130543, "optim/total_tokens": 2031353856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462150, "loss/out": 5.447206497192383, "created_at": "2025-01-14T23:55:31.241230+00:00"} {"global_step": 7750, "acc_step": 0, "speed/wps": 12954.544120848164, "speed/FLOPS": 203468916826996.8, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0013, "optim/grad_norm": 17.01348304748535, "optim/lr": 0.002981545930203527, "optim/total_tokens": 2031616000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431605, "loss/out": 5.501028060913086, "created_at": "2025-01-14T23:55:36.306372+00:00"} {"global_step": 7751, "acc_step": 0, "speed/wps": 12961.743299179894, "speed/FLOPS": 203581989815404.34, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9354076385498047, "optim/lr": 0.0029815325293601957, "optim/total_tokens": 2031878144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311717, "loss/out": 5.6465229988098145, "created_at": "2025-01-14T23:55:41.371147+00:00"} {"global_step": 7752, "acc_step": 0, "speed/wps": 12954.132269507629, "speed/FLOPS": 203462448135750.66, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.2217934131622314, "optim/lr": 0.0029815191236831043, "optim/total_tokens": 2032140288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431605, "loss/out": 5.51319694519043, "created_at": "2025-01-14T23:55:46.438963+00:00"} {"global_step": 7753, "acc_step": 0, "speed/wps": 12954.046558575126, "speed/FLOPS": 203461101927776.53, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.34332799911499, "optim/lr": 0.002981505713172296, "optim/total_tokens": 2032402432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 5.544767379760742, "created_at": "2025-01-14T23:55:51.506772+00:00"} {"global_step": 7754, "acc_step": 0, "speed/wps": 12946.597755397195, "speed/FLOPS": 203344108238138.94, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.875698089599609, "optim/lr": 0.0029814922978278146, "optim/total_tokens": 2032664576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460556, "loss/out": 5.613320350646973, "created_at": "2025-01-14T23:55:56.573114+00:00"} {"global_step": 7755, "acc_step": 0, "speed/wps": 12954.48574191106, "speed/FLOPS": 203467999905568.53, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.032600402832031, "optim/lr": 0.0029814788776497047, "optim/total_tokens": 2032926720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 5.510165691375732, "created_at": "2025-01-14T23:56:01.636010+00:00"} {"global_step": 7756, "acc_step": 0, "speed/wps": 12951.728043458028, "speed/FLOPS": 203424686461888.88, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4618135690689087, "optim/lr": 0.002981465452638009, "optim/total_tokens": 2033188864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433179, "loss/out": 5.678133010864258, "created_at": "2025-01-14T23:56:06.703383+00:00"} {"global_step": 7757, "acc_step": 0, "speed/wps": 12952.649665085153, "speed/FLOPS": 203439161795983.72, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.790198802947998, "optim/lr": 0.002981452022792772, "optim/total_tokens": 2033451008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351058, "loss/out": 5.464278221130371, "created_at": "2025-01-14T23:56:11.767231+00:00"} {"global_step": 7758, "acc_step": 0, "speed/wps": 12952.019499692307, "speed/FLOPS": 203429264182551.06, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.8661105632781982, "optim/lr": 0.002981438588114037, "optim/total_tokens": 2033713152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441442, "loss/out": 5.619478702545166, "created_at": "2025-01-14T23:56:16.830898+00:00"} {"global_step": 7759, "acc_step": 0, "speed/wps": 12949.482982198566, "speed/FLOPS": 203389424689771.56, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.5583415031433105, "optim/lr": 0.002981425148601848, "optim/total_tokens": 2033975296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 5.542941093444824, "created_at": "2025-01-14T23:56:21.897112+00:00"} {"global_step": 7760, "acc_step": 0, "speed/wps": 12947.661271088265, "speed/FLOPS": 203360812213490.9, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.8193917274475098, "optim/lr": 0.0029814117042562495, "optim/total_tokens": 2034237440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453131, "loss/out": 5.606337070465088, "created_at": "2025-01-14T23:56:26.963440+00:00"} {"global_step": 7761, "acc_step": 0, "speed/wps": 12956.2482191748, "speed/FLOPS": 203495682032893.12, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.509958267211914, "optim/lr": 0.0029813982550772845, "optim/total_tokens": 2034499584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457095, "loss/out": 5.528340816497803, "created_at": "2025-01-14T23:56:32.030239+00:00"} {"global_step": 7762, "acc_step": 0, "speed/wps": 12957.035817131802, "speed/FLOPS": 203508052341080.84, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.074505090713501, "optim/lr": 0.0029813848010649974, "optim/total_tokens": 2034761728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304572, "loss/out": 5.50240421295166, "created_at": "2025-01-14T23:56:37.097616+00:00"} {"global_step": 7763, "acc_step": 0, "speed/wps": 12955.752031125161, "speed/FLOPS": 203487888717739.72, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0009, "optim/grad_norm": 11.218536376953125, "optim/lr": 0.0029813713422194316, "optim/total_tokens": 2035023872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432365, "loss/out": 5.655523300170898, "created_at": "2025-01-14T23:56:42.161713+00:00"} {"global_step": 7764, "acc_step": 0, "speed/wps": 12949.237472938146, "speed/FLOPS": 203385568629471.6, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.004136800765991, "optim/lr": 0.0029813578785406318, "optim/total_tokens": 2035286016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479095, "loss/out": 5.599340915679932, "created_at": "2025-01-14T23:56:47.226306+00:00"} {"global_step": 7765, "acc_step": 0, "speed/wps": 12945.711220898871, "speed/FLOPS": 203330183995616.7, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.4080581665039062, "optim/lr": 0.002981344410028641, "optim/total_tokens": 2035548160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423342, "loss/out": 5.481196880340576, "created_at": "2025-01-14T23:56:52.292314+00:00"} {"global_step": 7766, "acc_step": 0, "speed/wps": 12951.090375974749, "speed/FLOPS": 203414671017817.03, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.821562647819519, "optim/lr": 0.0029813309366835032, "optim/total_tokens": 2035810304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 5.661867141723633, "created_at": "2025-01-14T23:56:57.356843+00:00"} {"global_step": 7767, "acc_step": 0, "speed/wps": 12950.027400325896, "speed/FLOPS": 203397975524570.2, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.9088163375854492, "optim/lr": 0.0029813174585052635, "optim/total_tokens": 2036072448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 5.58527135848999, "created_at": "2025-01-14T23:57:02.428575+00:00"} {"global_step": 7768, "acc_step": 0, "speed/wps": 12954.228448244381, "speed/FLOPS": 203463958754974.78, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.4040262699127197, "optim/lr": 0.002981303975493965, "optim/total_tokens": 2036334592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315105, "loss/out": 5.523728847503662, "created_at": "2025-01-14T23:57:07.495084+00:00"} {"global_step": 7769, "acc_step": 0, "speed/wps": 12957.999732136037, "speed/FLOPS": 203523191950780.38, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.5991536378860474, "optim/lr": 0.0029812904876496517, "optim/total_tokens": 2036596736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390956, "loss/out": 5.53753662109375, "created_at": "2025-01-14T23:57:12.561822+00:00"} {"global_step": 7770, "acc_step": 0, "speed/wps": 12953.3002355611, "speed/FLOPS": 203449379899285.97, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.5477705001831055, "optim/lr": 0.002981276994972368, "optim/total_tokens": 2036858880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470403, "loss/out": 5.489312171936035, "created_at": "2025-01-14T23:57:17.629518+00:00"} {"global_step": 7771, "acc_step": 0, "speed/wps": 12949.999404866148, "speed/FLOPS": 203397535817405.06, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.566165447235107, "optim/lr": 0.0029812634974621575, "optim/total_tokens": 2037121024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448083, "loss/out": 5.464144706726074, "created_at": "2025-01-14T23:57:22.694160+00:00"} {"global_step": 7772, "acc_step": 0, "speed/wps": 12954.550111641036, "speed/FLOPS": 203469010920631.1, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.311394691467285, "optim/lr": 0.002981249995119064, "optim/total_tokens": 2037383168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 5.574936866760254, "created_at": "2025-01-14T23:57:27.758746+00:00"} {"global_step": 7773, "acc_step": 0, "speed/wps": 12953.868940097529, "speed/FLOPS": 203458312185512.5, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8011184930801392, "optim/lr": 0.002981236487943132, "optim/total_tokens": 2037645312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427589, "loss/out": 5.493658065795898, "created_at": "2025-01-14T23:57:32.843846+00:00"} {"global_step": 7774, "acc_step": 0, "speed/wps": 12956.911752666952, "speed/FLOPS": 203506103738177.94, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.835229873657227, "optim/lr": 0.002981222975934406, "optim/total_tokens": 2037907456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443306, "loss/out": 5.512973785400391, "created_at": "2025-01-14T23:57:37.906167+00:00"} {"global_step": 7775, "acc_step": 0, "speed/wps": 12951.627928058479, "speed/FLOPS": 203423114011965.97, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6502360105514526, "optim/lr": 0.002981209459092929, "optim/total_tokens": 2038169600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 5.648523807525635, "created_at": "2025-01-14T23:57:42.974061+00:00"} {"global_step": 7776, "acc_step": 0, "speed/wps": 12942.055501202509, "speed/FLOPS": 203272765894300.03, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.8091859817504883, "optim/lr": 0.002981195937418746, "optim/total_tokens": 2038431744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 5.502276420593262, "created_at": "2025-01-14T23:57:48.041575+00:00"} {"global_step": 7777, "acc_step": 0, "speed/wps": 12945.834763727164, "speed/FLOPS": 203332124405500.66, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3544294834136963, "optim/lr": 0.0029811824109119006, "optim/total_tokens": 2038693888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438240, "loss/out": 5.466499328613281, "created_at": "2025-01-14T23:57:53.108877+00:00"} {"global_step": 7778, "acc_step": 0, "speed/wps": 12955.612825728993, "speed/FLOPS": 203485702305704.62, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.266432046890259, "optim/lr": 0.002981168879572438, "optim/total_tokens": 2038956032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349477, "loss/out": 5.448200225830078, "created_at": "2025-01-14T23:57:58.176055+00:00"} {"global_step": 7779, "acc_step": 0, "speed/wps": 12951.847114573811, "speed/FLOPS": 203426556637383.12, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.094688892364502, "optim/lr": 0.0029811553434004002, "optim/total_tokens": 2039218176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440921, "loss/out": 5.642457485198975, "created_at": "2025-01-14T23:58:03.240196+00:00"} {"global_step": 7780, "acc_step": 0, "speed/wps": 12952.92785377139, "speed/FLOPS": 203443531131572.38, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5481224060058594, "optim/lr": 0.0029811418023958335, "optim/total_tokens": 2039480320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 5.497286796569824, "created_at": "2025-01-14T23:58:08.310260+00:00"} {"global_step": 7781, "acc_step": 0, "speed/wps": 12955.289878822094, "speed/FLOPS": 203480629980756.22, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.100506544113159, "optim/lr": 0.002981128256558781, "optim/total_tokens": 2039742464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 124038, "loss/out": 5.518775939941406, "created_at": "2025-01-14T23:58:13.435522+00:00"} {"global_step": 7782, "acc_step": 0, "speed/wps": 12954.848282215346, "speed/FLOPS": 203473694099229.9, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.1362335681915283, "optim/lr": 0.002981114705889287, "optim/total_tokens": 2040004608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456054, "loss/out": 5.584764003753662, "created_at": "2025-01-14T23:58:18.501401+00:00"} {"global_step": 7783, "acc_step": 0, "speed/wps": 12955.498331771236, "speed/FLOPS": 203483904020766.88, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.611169815063477, "optim/lr": 0.002981101150387396, "optim/total_tokens": 2040266752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 5.528019905090332, "created_at": "2025-01-14T23:58:23.566284+00:00"} {"global_step": 7784, "acc_step": 0, "speed/wps": 12954.083247065091, "speed/FLOPS": 203461678170926.72, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.660855293273926, "optim/lr": 0.002981087590053152, "optim/total_tokens": 2040528896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423026, "loss/out": 5.535460948944092, "created_at": "2025-01-14T23:58:28.633535+00:00"} {"global_step": 7785, "acc_step": 0, "speed/wps": 12947.901464955745, "speed/FLOPS": 203364584788240.62, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7867661714553833, "optim/lr": 0.002981074024886599, "optim/total_tokens": 2040791040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451559, "loss/out": 5.548720836639404, "created_at": "2025-01-14T23:58:33.699107+00:00"} {"global_step": 7786, "acc_step": 0, "speed/wps": 12961.68681070111, "speed/FLOPS": 203581102586221.75, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.035210371017456, "optim/lr": 0.002981060454887782, "optim/total_tokens": 2041053184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306907, "loss/out": 5.442469596862793, "created_at": "2025-01-14T23:58:38.759475+00:00"} {"global_step": 7787, "acc_step": 0, "speed/wps": 12956.335392946932, "speed/FLOPS": 203497051216773.78, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.920427322387695, "optim/lr": 0.0029810468800567456, "optim/total_tokens": 2041315328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 382419, "loss/out": 5.548189640045166, "created_at": "2025-01-14T23:58:43.822836+00:00"} {"global_step": 7788, "acc_step": 0, "speed/wps": 12951.202201019549, "speed/FLOPS": 203416427383809.34, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.804797410964966, "optim/lr": 0.002981033300393532, "optim/total_tokens": 2041577472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460578, "loss/out": 5.570889472961426, "created_at": "2025-01-14T23:58:48.890016+00:00"} {"global_step": 7789, "acc_step": 0, "speed/wps": 12954.328808943774, "speed/FLOPS": 203465535057668.8, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.46071457862854, "optim/lr": 0.002981019715898187, "optim/total_tokens": 2041839616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427589, "loss/out": 5.455506324768066, "created_at": "2025-01-14T23:58:53.953264+00:00"} {"global_step": 7790, "acc_step": 0, "speed/wps": 12948.399739423645, "speed/FLOPS": 203372410873459.88, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.6019036769866943, "optim/lr": 0.0029810061265707556, "optim/total_tokens": 2042101760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 5.57954216003418, "created_at": "2025-01-14T23:58:59.018071+00:00"} {"global_step": 7791, "acc_step": 0, "speed/wps": 12955.365240246894, "speed/FLOPS": 203481813635490.16, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3028461933135986, "optim/lr": 0.0029809925324112804, "optim/total_tokens": 2042363904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326440, "loss/out": 5.572592735290527, "created_at": "2025-01-14T23:59:04.086913+00:00"} {"global_step": 7792, "acc_step": 0, "speed/wps": 12950.642959249926, "speed/FLOPS": 203407643723338.7, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1666059494018555, "optim/lr": 0.0029809789334198076, "optim/total_tokens": 2042626048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 400071, "loss/out": 5.540570259094238, "created_at": "2025-01-14T23:59:09.155150+00:00"} {"global_step": 7793, "acc_step": 0, "speed/wps": 12956.890042464876, "speed/FLOPS": 203505762749621.4, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.648221492767334, "optim/lr": 0.0029809653295963796, "optim/total_tokens": 2042888192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440919, "loss/out": 5.537668704986572, "created_at": "2025-01-14T23:59:14.219112+00:00"} {"global_step": 7794, "acc_step": 0, "speed/wps": 12951.815595045186, "speed/FLOPS": 203426061579873.7, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.249185085296631, "optim/lr": 0.0029809517209410425, "optim/total_tokens": 2043150336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 5.627391815185547, "created_at": "2025-01-14T23:59:19.286390+00:00"} {"global_step": 7795, "acc_step": 0, "speed/wps": 12950.468482423175, "speed/FLOPS": 203404903325018.28, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4422365427017212, "optim/lr": 0.0029809381074538395, "optim/total_tokens": 2043412480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430268, "loss/out": 5.638438701629639, "created_at": "2025-01-14T23:59:24.354553+00:00"} {"global_step": 7796, "acc_step": 0, "speed/wps": 12960.271605533775, "speed/FLOPS": 203558874844373.06, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.3065967559814453, "optim/lr": 0.0029809244891348154, "optim/total_tokens": 2043674624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 5.415465354919434, "created_at": "2025-01-14T23:59:29.415213+00:00"} {"global_step": 7797, "acc_step": 0, "speed/wps": 12960.199829504598, "speed/FLOPS": 203557747503204.22, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.7402663230895996, "optim/lr": 0.0029809108659840156, "optim/total_tokens": 2043936768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424678, "loss/out": 5.560371398925781, "created_at": "2025-01-14T23:59:34.480001+00:00"} {"global_step": 7798, "acc_step": 0, "speed/wps": 12954.356463020638, "speed/FLOPS": 203465969402946.66, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.0044007301330566, "optim/lr": 0.0029808972380014833, "optim/total_tokens": 2044198912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360416, "loss/out": 5.607478618621826, "created_at": "2025-01-14T23:59:39.544663+00:00"} {"global_step": 7799, "acc_step": 0, "speed/wps": 12951.837490874508, "speed/FLOPS": 203426405483961.22, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.4048333168029785, "optim/lr": 0.0029808836051872633, "optim/total_tokens": 2044461056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450756, "loss/out": 5.391594886779785, "created_at": "2025-01-14T23:59:44.612891+00:00"} {"global_step": 7800, "acc_step": 0, "speed/wps": 12948.964060636505, "speed/FLOPS": 203381274313566.44, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.907949686050415, "optim/lr": 0.0029808699675414, "optim/total_tokens": 2044723200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461100, "loss/out": 5.3940229415893555, "created_at": "2025-01-14T23:59:49.680007+00:00"} {"global_step": 7801, "acc_step": 0, "speed/wps": 12952.085895123888, "speed/FLOPS": 203430307014040.94, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0009, "optim/grad_norm": 9.852300643920898, "optim/lr": 0.002980856325063938, "optim/total_tokens": 2044985344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 501369, "loss/out": 5.641318321228027, "created_at": "2025-01-14T23:59:54.748474+00:00"} {"global_step": 7802, "acc_step": 0, "speed/wps": 12953.428675501744, "speed/FLOPS": 203451397225048.75, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.824118137359619, "optim/lr": 0.0029808426777549225, "optim/total_tokens": 2045247488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325819, "loss/out": 5.525539398193359, "created_at": "2025-01-14T23:59:59.815998+00:00"} {"global_step": 7803, "acc_step": 0, "speed/wps": 12956.708333312059, "speed/FLOPS": 203502908757681.84, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.886094093322754, "optim/lr": 0.002980829025614397, "optim/total_tokens": 2045509632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 408354, "loss/out": 5.4583353996276855, "created_at": "2025-01-15T00:00:04.886155+00:00"} {"global_step": 7804, "acc_step": 0, "speed/wps": 12958.099662112432, "speed/FLOPS": 203524761488377.84, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.4167802333831787, "optim/lr": 0.002980815368642406, "optim/total_tokens": 2045771776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 5.571462631225586, "created_at": "2025-01-15T00:00:09.949269+00:00"} {"global_step": 7805, "acc_step": 0, "speed/wps": 12955.426589520914, "speed/FLOPS": 203482777210141.38, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.021286964416504, "optim/lr": 0.002980801706838995, "optim/total_tokens": 2046033920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423364, "loss/out": 5.568745136260986, "created_at": "2025-01-15T00:00:15.016447+00:00"} {"global_step": 7806, "acc_step": 0, "speed/wps": 12950.937038598462, "speed/FLOPS": 203412262643614.8, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0006, "optim/grad_norm": 17.182626724243164, "optim/lr": 0.002980788040204208, "optim/total_tokens": 2046296064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 5.568238735198975, "created_at": "2025-01-15T00:00:20.084283+00:00"} {"global_step": 7807, "acc_step": 0, "speed/wps": 12954.107385033189, "speed/FLOPS": 203462057290884.72, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.908668041229248, "optim/lr": 0.0029807743687380894, "optim/total_tokens": 2046558208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 5.5259857177734375, "created_at": "2025-01-15T00:00:25.148200+00:00"} {"global_step": 7808, "acc_step": 0, "speed/wps": 12957.754775514984, "speed/FLOPS": 203519344570441.06, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.905213475227356, "optim/lr": 0.002980760692440684, "optim/total_tokens": 2046820352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436096, "loss/out": 5.480989456176758, "created_at": "2025-01-15T00:00:30.210026+00:00"} {"global_step": 7809, "acc_step": 0, "speed/wps": 12951.0977603631, "speed/FLOPS": 203414786999783.3, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.547037124633789, "optim/lr": 0.0029807470113120365, "optim/total_tokens": 2047082496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321540, "loss/out": 5.4889068603515625, "created_at": "2025-01-15T00:00:35.273962+00:00"} {"global_step": 7810, "acc_step": 0, "speed/wps": 12946.429508983812, "speed/FLOPS": 203341465697021.97, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.001, "optim/grad_norm": 6.287168025970459, "optim/lr": 0.002980733325352192, "optim/total_tokens": 2047344640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423893, "loss/out": 5.3998003005981445, "created_at": "2025-01-15T00:00:40.344437+00:00"} {"global_step": 7811, "acc_step": 0, "speed/wps": 12962.746202054497, "speed/FLOPS": 203597741783182.97, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 36.0522346496582, "optim/lr": 0.0029807196345611946, "optim/total_tokens": 2047606784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 5.460845470428467, "created_at": "2025-01-15T00:00:45.404187+00:00"} {"global_step": 7812, "acc_step": 0, "speed/wps": 12955.142321229874, "speed/FLOPS": 203478312386005.03, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3452322483062744, "optim/lr": 0.0029807059389390885, "optim/total_tokens": 2047868928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441679, "loss/out": 5.518855094909668, "created_at": "2025-01-15T00:00:50.469853+00:00"} {"global_step": 7813, "acc_step": 0, "speed/wps": 12956.022963075737, "speed/FLOPS": 203492144076323.03, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.555515766143799, "optim/lr": 0.002980692238485919, "optim/total_tokens": 2048131072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439871, "loss/out": 5.537403583526611, "created_at": "2025-01-15T00:00:55.537216+00:00"} {"global_step": 7814, "acc_step": 0, "speed/wps": 12956.184705651312, "speed/FLOPS": 203494684465730.03, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8025622367858887, "optim/lr": 0.002980678533201731, "optim/total_tokens": 2048393216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292724, "loss/out": 5.51930046081543, "created_at": "2025-01-15T00:01:00.599024+00:00"} {"global_step": 7815, "acc_step": 0, "speed/wps": 12961.72381943288, "speed/FLOPS": 203581683859209.5, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.010763645172119, "optim/lr": 0.0029806648230865686, "optim/total_tokens": 2048655360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 395808, "loss/out": 5.426800727844238, "created_at": "2025-01-15T00:01:05.665855+00:00"} {"global_step": 7816, "acc_step": 0, "speed/wps": 12958.25655759953, "speed/FLOPS": 203527225747599.72, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.923098564147949, "optim/lr": 0.0029806511081404774, "optim/total_tokens": 2048917504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469875, "loss/out": 5.422054767608643, "created_at": "2025-01-15T00:01:10.727952+00:00"} {"global_step": 7817, "acc_step": 0, "speed/wps": 12955.317768362265, "speed/FLOPS": 203481068024308.88, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8752208948135376, "optim/lr": 0.0029806373883635015, "optim/total_tokens": 2049179648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440637, "loss/out": 5.423490047454834, "created_at": "2025-01-15T00:01:15.798056+00:00"} {"global_step": 7818, "acc_step": 0, "speed/wps": 12963.282923944626, "speed/FLOPS": 203606171738000.6, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.6746163368225098, "optim/lr": 0.0029806236637556854, "optim/total_tokens": 2049441792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479622, "loss/out": 5.474058628082275, "created_at": "2025-01-15T00:01:20.861248+00:00"} {"global_step": 7819, "acc_step": 0, "speed/wps": 12952.024985138776, "speed/FLOPS": 203429350339025.66, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.8757855892181396, "optim/lr": 0.0029806099343170746, "optim/total_tokens": 2049703936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438769, "loss/out": 5.51234245300293, "created_at": "2025-01-15T00:01:25.929376+00:00"} {"global_step": 7820, "acc_step": 0, "speed/wps": 12948.88588096185, "speed/FLOPS": 203380046394345.88, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.001, "optim/grad_norm": 2.5243990421295166, "optim/lr": 0.0029805962000477126, "optim/total_tokens": 2049966080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307440, "loss/out": 5.563546180725098, "created_at": "2025-01-15T00:01:30.995275+00:00"} {"global_step": 7821, "acc_step": 0, "speed/wps": 12957.403105010133, "speed/FLOPS": 203513821101916.47, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0009, "optim/grad_norm": 3.443857192993164, "optim/lr": 0.0029805824609476462, "optim/total_tokens": 2050228224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355086, "loss/out": 5.617371559143066, "created_at": "2025-01-15T00:01:36.056918+00:00"} {"global_step": 7822, "acc_step": 0, "speed/wps": 12951.536815283991, "speed/FLOPS": 203421682960640.6, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.5219173431396484, "optim/lr": 0.002980568717016919, "optim/total_tokens": 2050490368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430268, "loss/out": 5.439074993133545, "created_at": "2025-01-15T00:01:41.122346+00:00"} {"global_step": 7823, "acc_step": 0, "speed/wps": 12953.433184513377, "speed/FLOPS": 203451468045272.47, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.805031776428223, "optim/lr": 0.0029805549682555763, "optim/total_tokens": 2050752512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 5.550383567810059, "created_at": "2025-01-15T00:01:46.210180+00:00"} {"global_step": 7824, "acc_step": 0, "speed/wps": 12958.105334059941, "speed/FLOPS": 203524850574107.53, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6693793535232544, "optim/lr": 0.002980541214663662, "optim/total_tokens": 2051014656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440105, "loss/out": 5.430097579956055, "created_at": "2025-01-15T00:01:51.273233+00:00"} {"global_step": 7825, "acc_step": 0, "speed/wps": 12962.061728167333, "speed/FLOPS": 203586991180219.34, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.5417168140411377, "optim/lr": 0.0029805274562412216, "optim/total_tokens": 2051276800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317586, "loss/out": 5.408883094787598, "created_at": "2025-01-15T00:01:56.353317+00:00"} {"global_step": 7826, "acc_step": 0, "speed/wps": 12954.54854716593, "speed/FLOPS": 203468986348399.72, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0008, "optim/grad_norm": 9.929473876953125, "optim/lr": 0.0029805136929883006, "optim/total_tokens": 2051538944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 5.461992263793945, "created_at": "2025-01-15T00:02:01.421975+00:00"} {"global_step": 7827, "acc_step": 0, "speed/wps": 12951.91248077711, "speed/FLOPS": 203427583303428.5, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.687953233718872, "optim/lr": 0.002980499924904943, "optim/total_tokens": 2051801088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437195, "loss/out": 5.501343727111816, "created_at": "2025-01-15T00:02:06.487853+00:00"} {"global_step": 7828, "acc_step": 0, "speed/wps": 12950.85898525952, "speed/FLOPS": 203411036708670.2, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.832938194274902, "optim/lr": 0.0029804861519911944, "optim/total_tokens": 2052063232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 5.485033988952637, "created_at": "2025-01-15T00:02:11.552235+00:00"} {"global_step": 7829, "acc_step": 0, "speed/wps": 12956.085215503055, "speed/FLOPS": 203493121836237.84, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5388767719268799, "optim/lr": 0.002980472374247099, "optim/total_tokens": 2052325376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326583, "loss/out": 5.548494338989258, "created_at": "2025-01-15T00:02:16.615904+00:00"} {"global_step": 7830, "acc_step": 0, "speed/wps": 12957.97596144138, "speed/FLOPS": 203522818599356.44, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.2820091247558594, "optim/lr": 0.0029804585916727027, "optim/total_tokens": 2052587520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356143, "loss/out": 5.550788402557373, "created_at": "2025-01-15T00:02:21.677695+00:00"} {"global_step": 7831, "acc_step": 0, "speed/wps": 12957.386828274535, "speed/FLOPS": 203513565453417.3, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.700331449508667, "optim/lr": 0.00298044480426805, "optim/total_tokens": 2052849664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433179, "loss/out": 5.470028877258301, "created_at": "2025-01-15T00:02:26.744690+00:00"} {"global_step": 7832, "acc_step": 0, "speed/wps": 12956.979021355026, "speed/FLOPS": 203507160285361.0, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.497720241546631, "optim/lr": 0.002980431012033185, "optim/total_tokens": 2053111808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 5.552618026733398, "created_at": "2025-01-15T00:02:31.806225+00:00"} {"global_step": 7833, "acc_step": 0, "speed/wps": 12955.769021735021, "speed/FLOPS": 203488155578615.1, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.634089469909668, "optim/lr": 0.0029804172149681544, "optim/total_tokens": 2053373952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 5.4772772789001465, "created_at": "2025-01-15T00:02:36.868681+00:00"} {"global_step": 7834, "acc_step": 0, "speed/wps": 12952.783270649054, "speed/FLOPS": 203441260254958.06, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.973125457763672, "optim/lr": 0.002980403413073002, "optim/total_tokens": 2053636096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428948, "loss/out": 5.529358863830566, "created_at": "2025-01-15T00:02:41.934461+00:00"} {"global_step": 7835, "acc_step": 0, "speed/wps": 12950.104619042253, "speed/FLOPS": 203399188350620.22, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.5324339866638184, "optim/lr": 0.0029803896063477727, "optim/total_tokens": 2053898240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 5.466058731079102, "created_at": "2025-01-15T00:02:47.001253+00:00"} {"global_step": 7836, "acc_step": 0, "speed/wps": 12946.881546401333, "speed/FLOPS": 203348565565828.62, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.3894968032836914, "optim/lr": 0.0029803757947925126, "optim/total_tokens": 2054160384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329804, "loss/out": 5.532038688659668, "created_at": "2025-01-15T00:02:52.067103+00:00"} {"global_step": 7837, "acc_step": 0, "speed/wps": 12952.480674851033, "speed/FLOPS": 203436507572139.94, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 10.612208366394043, "optim/lr": 0.002980361978407266, "optim/total_tokens": 2054422528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361049, "loss/out": 5.51870584487915, "created_at": "2025-01-15T00:02:57.136590+00:00"} {"global_step": 7838, "acc_step": 0, "speed/wps": 12960.582263837247, "speed/FLOPS": 203563754159918.3, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.68458366394043, "optim/lr": 0.0029803481571920778, "optim/total_tokens": 2054684672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452847, "loss/out": 5.379298210144043, "created_at": "2025-01-15T00:03:02.198473+00:00"} {"global_step": 7839, "acc_step": 0, "speed/wps": 12957.143333478032, "speed/FLOPS": 203509741033041.34, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0009, "optim/grad_norm": 3.835563898086548, "optim/lr": 0.0029803343311469934, "optim/total_tokens": 2054946816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 5.371730804443359, "created_at": "2025-01-15T00:03:07.261065+00:00"} {"global_step": 7840, "acc_step": 0, "speed/wps": 12958.518781579854, "speed/FLOPS": 203531344335544.22, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.233603477478027, "optim/lr": 0.002980320500272059, "optim/total_tokens": 2055208960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450462, "loss/out": 5.432668209075928, "created_at": "2025-01-15T00:03:12.323135+00:00"} {"global_step": 7841, "acc_step": 0, "speed/wps": 12957.033865559004, "speed/FLOPS": 203508021688948.3, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 8.647305488586426, "optim/lr": 0.002980306664567318, "optim/total_tokens": 2055471104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459743, "loss/out": 5.431711196899414, "created_at": "2025-01-15T00:03:17.386059+00:00"} {"global_step": 7842, "acc_step": 0, "speed/wps": 12952.912770874073, "speed/FLOPS": 203443294233943.94, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0012, "optim/grad_norm": 5.050390243530273, "optim/lr": 0.0029802928240328153, "optim/total_tokens": 2055733248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319090, "loss/out": 5.500395774841309, "created_at": "2025-01-15T00:03:22.456449+00:00"} {"global_step": 7843, "acc_step": 0, "speed/wps": 12963.353186629945, "speed/FLOPS": 203607275310024.6, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.85148024559021, "optim/lr": 0.0029802789786685983, "optim/total_tokens": 2055995392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 409163, "loss/out": 5.488891124725342, "created_at": "2025-01-15T00:03:27.515487+00:00"} {"global_step": 7844, "acc_step": 0, "speed/wps": 12963.020862930438, "speed/FLOPS": 203602055709663.2, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.143100738525391, "optim/lr": 0.0029802651284747106, "optim/total_tokens": 2056257536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431605, "loss/out": 5.431291103363037, "created_at": "2025-01-15T00:03:32.575733+00:00"} {"global_step": 7845, "acc_step": 0, "speed/wps": 12964.317986901859, "speed/FLOPS": 203622428824069.47, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.898960828781128, "optim/lr": 0.002980251273451197, "optim/total_tokens": 2056519680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480663, "loss/out": 5.426406383514404, "created_at": "2025-01-15T00:03:37.641198+00:00"} {"global_step": 7846, "acc_step": 0, "speed/wps": 12955.717799582542, "speed/FLOPS": 203487351064323.8, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.194217681884766, "optim/lr": 0.002980237413598104, "optim/total_tokens": 2056781824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446742, "loss/out": 5.4501752853393555, "created_at": "2025-01-15T00:03:42.711610+00:00"} {"global_step": 7847, "acc_step": 0, "speed/wps": 12961.331044875838, "speed/FLOPS": 203575514795062.38, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.896444320678711, "optim/lr": 0.0029802235489154757, "optim/total_tokens": 2057043968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440628, "loss/out": 5.48492431640625, "created_at": "2025-01-15T00:03:47.774391+00:00"} {"global_step": 7848, "acc_step": 0, "speed/wps": 12963.36585319466, "speed/FLOPS": 203607474255829.12, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.453674793243408, "optim/lr": 0.002980209679403358, "optim/total_tokens": 2057306112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 5.346668243408203, "created_at": "2025-01-15T00:03:52.833594+00:00"} {"global_step": 7849, "acc_step": 0, "speed/wps": 12963.121873338083, "speed/FLOPS": 203603642216918.1, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.581659317016602, "optim/lr": 0.002980195805061796, "optim/total_tokens": 2057568256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444878, "loss/out": 5.456668853759766, "created_at": "2025-01-15T00:03:57.892786+00:00"} {"global_step": 7850, "acc_step": 0, "speed/wps": 12958.791855342326, "speed/FLOPS": 203535633334226.5, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.146742582321167, "optim/lr": 0.002980181925890835, "optim/total_tokens": 2057830400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310132, "loss/out": 5.582774639129639, "created_at": "2025-01-15T00:04:02.957840+00:00"} {"global_step": 7851, "acc_step": 0, "speed/wps": 12958.667423152798, "speed/FLOPS": 203533678955702.22, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.473205327987671, "optim/lr": 0.00298016804189052, "optim/total_tokens": 2058092544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 410549, "loss/out": 5.626551628112793, "created_at": "2025-01-15T00:04:08.019843+00:00"} {"global_step": 7852, "acc_step": 0, "speed/wps": 12954.492023601568, "speed/FLOPS": 203468098568149.9, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.322014808654785, "optim/lr": 0.002980154153060897, "optim/total_tokens": 2058354688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442778, "loss/out": 5.580338478088379, "created_at": "2025-01-15T00:04:13.082549+00:00"} {"global_step": 7853, "acc_step": 0, "speed/wps": 12954.373550132148, "speed/FLOPS": 203466237779513.06, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.9885241985321045, "optim/lr": 0.0029801402594020103, "optim/total_tokens": 2058616832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 5.508182525634766, "created_at": "2025-01-15T00:04:18.151625+00:00"} {"global_step": 7854, "acc_step": 0, "speed/wps": 12960.49151676894, "speed/FLOPS": 203562328852511.03, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.63757610321045, "optim/lr": 0.0029801263609139056, "optim/total_tokens": 2058878976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463750, "loss/out": 5.404128074645996, "created_at": "2025-01-15T00:04:23.217207+00:00"} {"global_step": 7855, "acc_step": 0, "speed/wps": 12963.666901546023, "speed/FLOPS": 203612202633871.06, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.096504211425781, "optim/lr": 0.002980112457596629, "optim/total_tokens": 2059141120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430567, "loss/out": 5.458535194396973, "created_at": "2025-01-15T00:04:28.306487+00:00"} {"global_step": 7856, "acc_step": 0, "speed/wps": 12954.825477070408, "speed/FLOPS": 203473335913091.12, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.105942726135254, "optim/lr": 0.0029800985494502245, "optim/total_tokens": 2059403264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456330, "loss/out": 5.553091049194336, "created_at": "2025-01-15T00:04:33.369330+00:00"} {"global_step": 7857, "acc_step": 0, "speed/wps": 12951.897888942722, "speed/FLOPS": 203427354118618.16, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.796144962310791, "optim/lr": 0.002980084636474739, "optim/total_tokens": 2059665408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446505, "loss/out": 5.508346080780029, "created_at": "2025-01-15T00:04:38.438069+00:00"} {"global_step": 7858, "acc_step": 0, "speed/wps": 12953.84646509225, "speed/FLOPS": 203457959184670.72, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0009, "optim/grad_norm": 2.188187599182129, "optim/lr": 0.0029800707186702162, "optim/total_tokens": 2059927552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 5.459999084472656, "created_at": "2025-01-15T00:04:43.505858+00:00"} {"global_step": 7859, "acc_step": 0, "speed/wps": 12963.256155929374, "speed/FLOPS": 203605751309537.94, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.316823959350586, "optim/lr": 0.0029800567960367036, "optim/total_tokens": 2060189696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443022, "loss/out": 5.355289459228516, "created_at": "2025-01-15T00:04:48.566226+00:00"} {"global_step": 7860, "acc_step": 0, "speed/wps": 12962.746791863243, "speed/FLOPS": 203597751046939.78, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0597777366638184, "optim/lr": 0.0029800428685742447, "optim/total_tokens": 2060451840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444641, "loss/out": 5.574413299560547, "created_at": "2025-01-15T00:04:53.626806+00:00"} {"global_step": 7861, "acc_step": 0, "speed/wps": 12957.712239524348, "speed/FLOPS": 203518676484257.56, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0011, "optim/grad_norm": 2.9172849655151367, "optim/lr": 0.002980028936282886, "optim/total_tokens": 2060713984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305274, "loss/out": 5.486327648162842, "created_at": "2025-01-15T00:04:58.688054+00:00"} {"global_step": 7862, "acc_step": 0, "speed/wps": 12956.114725800517, "speed/FLOPS": 203493585336010.88, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.8999767303466797, "optim/lr": 0.0029800149991626723, "optim/total_tokens": 2060976128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374658, "loss/out": 5.451009273529053, "created_at": "2025-01-15T00:05:03.752299+00:00"} {"global_step": 7863, "acc_step": 0, "speed/wps": 12959.070792396702, "speed/FLOPS": 203540014423965.16, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.179091691970825, "optim/lr": 0.00298000105721365, "optim/total_tokens": 2061238272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440628, "loss/out": 5.415529251098633, "created_at": "2025-01-15T00:05:08.816933+00:00"} {"global_step": 7864, "acc_step": 0, "speed/wps": 12960.951727042984, "speed/FLOPS": 203569557087260.06, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.217724561691284, "optim/lr": 0.002979987110435863, "optim/total_tokens": 2061500416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 5.461676597595215, "created_at": "2025-01-15T00:05:13.880248+00:00"} {"global_step": 7865, "acc_step": 0, "speed/wps": 12966.931625026069, "speed/FLOPS": 203663479602325.34, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 17.407634735107422, "optim/lr": 0.002979973158829359, "optim/total_tokens": 2061762560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437195, "loss/out": 5.484323024749756, "created_at": "2025-01-15T00:05:18.938709+00:00"} {"global_step": 7866, "acc_step": 0, "speed/wps": 12954.301228913133, "speed/FLOPS": 203465101875388.66, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.2108213901519775, "optim/lr": 0.002979959202394182, "optim/total_tokens": 2062024704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 5.484277248382568, "created_at": "2025-01-15T00:05:24.007988+00:00"} {"global_step": 7867, "acc_step": 0, "speed/wps": 12961.9783547938, "speed/FLOPS": 203585681686821.0, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 19.85356903076172, "optim/lr": 0.0029799452411303773, "optim/total_tokens": 2062286848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433993, "loss/out": 5.386289596557617, "created_at": "2025-01-15T00:05:29.072805+00:00"} {"global_step": 7868, "acc_step": 0, "speed/wps": 12958.874061317969, "speed/FLOPS": 203536924492036.8, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.180646896362305, "optim/lr": 0.0029799312750379914, "optim/total_tokens": 2062548992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354562, "loss/out": 5.5996928215026855, "created_at": "2025-01-15T00:05:34.139219+00:00"} {"global_step": 7869, "acc_step": 0, "speed/wps": 12964.480125348395, "speed/FLOPS": 203624975431173.9, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.313648223876953, "optim/lr": 0.0029799173041170695, "optim/total_tokens": 2062811136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457902, "loss/out": 5.637226104736328, "created_at": "2025-01-15T00:05:39.198299+00:00"} {"global_step": 7870, "acc_step": 0, "speed/wps": 12965.014112422645, "speed/FLOPS": 203633362431950.12, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.319911956787109, "optim/lr": 0.002979903328367657, "optim/total_tokens": 2063073280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456045, "loss/out": 5.477910995483398, "created_at": "2025-01-15T00:05:44.261841+00:00"} {"global_step": 7871, "acc_step": 0, "speed/wps": 12962.97883681755, "speed/FLOPS": 203601395631810.38, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.493762969970703, "optim/lr": 0.0029798893477898003, "optim/total_tokens": 2063335424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443307, "loss/out": 5.416553497314453, "created_at": "2025-01-15T00:05:49.324172+00:00"} {"global_step": 7872, "acc_step": 0, "speed/wps": 12964.007247410633, "speed/FLOPS": 203617548233358.97, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.9565000534057617, "optim/lr": 0.0029798753623835436, "optim/total_tokens": 2063597568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 5.636117935180664, "created_at": "2025-01-15T00:05:54.387739+00:00"} {"global_step": 7873, "acc_step": 0, "speed/wps": 12969.93712039304, "speed/FLOPS": 203710685037047.06, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.042043685913086, "optim/lr": 0.002979861372148934, "optim/total_tokens": 2063859712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430283, "loss/out": 5.404865741729736, "created_at": "2025-01-15T00:05:59.444418+00:00"} {"global_step": 7874, "acc_step": 0, "speed/wps": 12963.406649540326, "speed/FLOPS": 203608115018497.84, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.689115524291992, "optim/lr": 0.0029798473770860164, "optim/total_tokens": 2064121856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330623, "loss/out": 5.424746036529541, "created_at": "2025-01-15T00:06:04.509277+00:00"} {"global_step": 7875, "acc_step": 0, "speed/wps": 12960.679206770772, "speed/FLOPS": 203565276781902.88, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 31.90140151977539, "optim/lr": 0.0029798333771948364, "optim/total_tokens": 2064384000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355857, "loss/out": 5.549195289611816, "created_at": "2025-01-15T00:06:09.574750+00:00"} {"global_step": 7876, "acc_step": 0, "speed/wps": 12964.206967051834, "speed/FLOPS": 203620685104767.6, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 72.92033386230469, "optim/lr": 0.0029798193724754393, "optim/total_tokens": 2064646144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439583, "loss/out": 5.466673374176025, "created_at": "2025-01-15T00:06:14.635549+00:00"} {"global_step": 7877, "acc_step": 0, "speed/wps": 12959.978928069168, "speed/FLOPS": 203554277942610.4, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0009, "optim/grad_norm": 7.406352996826172, "optim/lr": 0.002979805362927871, "optim/total_tokens": 2064908288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 5.668733596801758, "created_at": "2025-01-15T00:06:19.696448+00:00"} {"global_step": 7878, "acc_step": 0, "speed/wps": 12964.136613905417, "speed/FLOPS": 203619580111927.75, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.3505728244781494, "optim/lr": 0.0029797913485521787, "optim/total_tokens": 2065170432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420954, "loss/out": 5.505112171173096, "created_at": "2025-01-15T00:06:24.762474+00:00"} {"global_step": 7879, "acc_step": 0, "speed/wps": 12962.460454386528, "speed/FLOPS": 203593253723399.06, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.117116928100586, "optim/lr": 0.0029797773293484065, "optim/total_tokens": 2065432576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 5.4662065505981445, "created_at": "2025-01-15T00:06:29.823133+00:00"} {"global_step": 7880, "acc_step": 0, "speed/wps": 12959.096777048833, "speed/FLOPS": 203540422548633.53, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.6393985748291016, "optim/lr": 0.0029797633053166004, "optim/total_tokens": 2065694720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327688, "loss/out": 5.496016979217529, "created_at": "2025-01-15T00:06:34.884180+00:00"} {"global_step": 7881, "acc_step": 0, "speed/wps": 12965.63405685686, "speed/FLOPS": 203643099511182.4, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 19.08400535583496, "optim/lr": 0.0029797492764568065, "optim/total_tokens": 2065956864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363627, "loss/out": 5.591976165771484, "created_at": "2025-01-15T00:06:39.943025+00:00"} {"global_step": 7882, "acc_step": 0, "speed/wps": 12959.547157647203, "speed/FLOPS": 203547496394818.72, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.9275777339935303, "optim/lr": 0.0029797352427690703, "optim/total_tokens": 2066219008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433464, "loss/out": 5.534255504608154, "created_at": "2025-01-15T00:06:45.007693+00:00"} {"global_step": 7883, "acc_step": 0, "speed/wps": 12958.98042208189, "speed/FLOPS": 203538595033989.2, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0009, "optim/grad_norm": 11.229572296142578, "optim/lr": 0.0029797212042534374, "optim/total_tokens": 2066481152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453375, "loss/out": 5.448864459991455, "created_at": "2025-01-15T00:06:50.068502+00:00"} {"global_step": 7884, "acc_step": 0, "speed/wps": 12952.764697598874, "speed/FLOPS": 203440968539682.94, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.0009, "optim/grad_norm": 12.271888732910156, "optim/lr": 0.0029797071609099545, "optim/total_tokens": 2066743296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 5.550799369812012, "created_at": "2025-01-15T00:06:55.136145+00:00"} {"global_step": 7885, "acc_step": 0, "speed/wps": 12958.938446027963, "speed/FLOPS": 203537935742381.16, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 54.468605041503906, "optim/lr": 0.0029796931127386663, "optim/total_tokens": 2067005440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435338, "loss/out": 5.48426628112793, "created_at": "2025-01-15T00:07:00.201515+00:00"} {"global_step": 7886, "acc_step": 0, "speed/wps": 12948.492397286644, "speed/FLOPS": 203373866192523.5, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.182957649230957, "optim/lr": 0.0029796790597396198, "optim/total_tokens": 2067267584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304859, "loss/out": 5.614692211151123, "created_at": "2025-01-15T00:07:05.266853+00:00"} {"global_step": 7887, "acc_step": 0, "speed/wps": 12958.593949097827, "speed/FLOPS": 203532524944704.97, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.468642234802246, "optim/lr": 0.0029796650019128595, "optim/total_tokens": 2067529728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298696, "loss/out": 5.599315643310547, "created_at": "2025-01-15T00:07:10.334135+00:00"} {"global_step": 7888, "acc_step": 0, "speed/wps": 12958.635175984538, "speed/FLOPS": 203533172469612.44, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 15.567874908447266, "optim/lr": 0.002979650939258432, "optim/total_tokens": 2067791872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 5.624274253845215, "created_at": "2025-01-15T00:07:15.400481+00:00"} {"global_step": 7889, "acc_step": 0, "speed/wps": 12955.13990752033, "speed/FLOPS": 203478274475379.88, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 27.295150756835938, "optim/lr": 0.002979636871776383, "optim/total_tokens": 2068054016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 5.599452972412109, "created_at": "2025-01-15T00:07:20.463051+00:00"} {"global_step": 7890, "acc_step": 0, "speed/wps": 12956.077207598948, "speed/FLOPS": 203492996061099.84, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.001, "optim/grad_norm": 75.1855239868164, "optim/lr": 0.002979622799466759, "optim/total_tokens": 2068316160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429217, "loss/out": 5.55417537689209, "created_at": "2025-01-15T00:07:25.531105+00:00"} {"global_step": 7891, "acc_step": 0, "speed/wps": 12955.292011140518, "speed/FLOPS": 203480663471747.2, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 21.514467239379883, "optim/lr": 0.0029796087223296058, "optim/total_tokens": 2068578304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456567, "loss/out": 5.646538734436035, "created_at": "2025-01-15T00:07:30.597502+00:00"} {"global_step": 7892, "acc_step": 0, "speed/wps": 12956.931697488144, "speed/FLOPS": 203506416999001.8, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.209616184234619, "optim/lr": 0.002979594640364968, "optim/total_tokens": 2068840448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298347, "loss/out": 5.714663505554199, "created_at": "2025-01-15T00:07:35.664528+00:00"} {"global_step": 7893, "acc_step": 0, "speed/wps": 12962.12206404487, "speed/FLOPS": 203587938838086.25, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.211735725402832, "optim/lr": 0.002979580553572893, "optim/total_tokens": 2069102592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440637, "loss/out": 5.624720573425293, "created_at": "2025-01-15T00:07:40.728456+00:00"} {"global_step": 7894, "acc_step": 0, "speed/wps": 12954.455316102247, "speed/FLOPS": 203467522026431.7, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.957240104675293, "optim/lr": 0.002979566461953427, "optim/total_tokens": 2069364736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 5.482112884521484, "created_at": "2025-01-15T00:07:45.791390+00:00"} {"global_step": 7895, "acc_step": 0, "speed/wps": 12956.881286832539, "speed/FLOPS": 203505625230384.03, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 42.63806915283203, "optim/lr": 0.002979552365506614, "optim/total_tokens": 2069626880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429454, "loss/out": 5.455487251281738, "created_at": "2025-01-15T00:07:50.857445+00:00"} {"global_step": 7896, "acc_step": 0, "speed/wps": 12956.914758700666, "speed/FLOPS": 203506150952068.16, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.998974084854126, "optim/lr": 0.002979538264232502, "optim/total_tokens": 2069889024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 5.684647560119629, "created_at": "2025-01-15T00:07:55.923377+00:00"} {"global_step": 7897, "acc_step": 0, "speed/wps": 12964.363136282378, "speed/FLOPS": 203623137957132.28, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.444282531738281, "optim/lr": 0.0029795241581311357, "optim/total_tokens": 2070151168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323942, "loss/out": 5.594164848327637, "created_at": "2025-01-15T00:08:00.986301+00:00"} {"global_step": 7898, "acc_step": 0, "speed/wps": 12955.79642434556, "speed/FLOPS": 203488585974268.78, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 20.32212257385254, "optim/lr": 0.0029795100472025625, "optim/total_tokens": 2070413312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 389370, "loss/out": 5.5872297286987305, "created_at": "2025-01-15T00:08:06.050837+00:00"} {"global_step": 7899, "acc_step": 0, "speed/wps": 12963.22717667447, "speed/FLOPS": 203605296150518.03, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.446282386779785, "optim/lr": 0.0029794959314468268, "optim/total_tokens": 2070675456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 5.547861099243164, "created_at": "2025-01-15T00:08:11.114715+00:00"} {"global_step": 7900, "acc_step": 0, "speed/wps": 12957.622067106417, "speed/FLOPS": 203517260202524.06, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 21.273548126220703, "optim/lr": 0.002979481810863976, "optim/total_tokens": 2070937600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428403, "loss/out": 5.856986045837402, "created_at": "2025-01-15T00:08:16.178500+00:00"} {"global_step": 7901, "acc_step": 0, "speed/wps": 12956.814763499118, "speed/FLOPS": 203504580390020.03, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.391249656677246, "optim/lr": 0.002979467685454056, "optim/total_tokens": 2071199744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 5.669747352600098, "created_at": "2025-01-15T00:08:21.243463+00:00"} {"global_step": 7902, "acc_step": 0, "speed/wps": 12962.58264801106, "speed/FLOPS": 203595172942183.97, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 11.239799499511719, "optim/lr": 0.0029794535552171114, "optim/total_tokens": 2071461888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428623, "loss/out": 5.537352561950684, "created_at": "2025-01-15T00:08:26.309795+00:00"} {"global_step": 7903, "acc_step": 0, "speed/wps": 12964.731630326725, "speed/FLOPS": 203628925662462.66, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 14.573296546936035, "optim/lr": 0.002979439420153191, "optim/total_tokens": 2071724032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 5.501667499542236, "created_at": "2025-01-15T00:08:31.369787+00:00"} {"global_step": 7904, "acc_step": 0, "speed/wps": 12961.872095230108, "speed/FLOPS": 203584012734357.28, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.769988536834717, "optim/lr": 0.0029794252802623376, "optim/total_tokens": 2071986176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419902, "loss/out": 5.724720001220703, "created_at": "2025-01-15T00:08:36.430503+00:00"} {"global_step": 7905, "acc_step": 0, "speed/wps": 12957.999653393379, "speed/FLOPS": 203523190714018.72, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.78680682182312, "optim/lr": 0.0029794111355446002, "optim/total_tokens": 2072248320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 344264, "loss/out": 5.561676979064941, "created_at": "2025-01-15T00:08:41.491725+00:00"} {"global_step": 7906, "acc_step": 0, "speed/wps": 12958.596185117543, "speed/FLOPS": 203532560064467.22, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.316187381744385, "optim/lr": 0.0029793969860000234, "optim/total_tokens": 2072510464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444359, "loss/out": 5.569752216339111, "created_at": "2025-01-15T00:08:46.557164+00:00"} {"global_step": 7907, "acc_step": 0, "speed/wps": 12961.157767233351, "speed/FLOPS": 203572793231578.7, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.4212310314178467, "optim/lr": 0.002979382831628654, "optim/total_tokens": 2072772608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 5.625213623046875, "created_at": "2025-01-15T00:08:51.623247+00:00"} {"global_step": 7908, "acc_step": 0, "speed/wps": 12959.364555340904, "speed/FLOPS": 203544628374676.44, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.2501378059387207, "optim/lr": 0.002979368672430538, "optim/total_tokens": 2073034752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450462, "loss/out": 5.601135730743408, "created_at": "2025-01-15T00:08:56.690091+00:00"} {"global_step": 7909, "acc_step": 0, "speed/wps": 12961.224337557693, "speed/FLOPS": 203573838809999.7, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.915597677230835, "optim/lr": 0.0029793545084057216, "optim/total_tokens": 2073296896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440399, "loss/out": 5.526126861572266, "created_at": "2025-01-15T00:09:01.754258+00:00"} {"global_step": 7910, "acc_step": 0, "speed/wps": 12954.61440330919, "speed/FLOPS": 203470020709625.44, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.410319805145264, "optim/lr": 0.002979340339554251, "optim/total_tokens": 2073559040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317346, "loss/out": 5.588251113891602, "created_at": "2025-01-15T00:09:06.820038+00:00"} {"global_step": 7911, "acc_step": 0, "speed/wps": 12962.79346811266, "speed/FLOPS": 203598484161576.78, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 21.85784149169922, "optim/lr": 0.0029793261658761726, "optim/total_tokens": 2073821184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441971, "loss/out": 5.555832862854004, "created_at": "2025-01-15T00:09:11.879428+00:00"} {"global_step": 7912, "acc_step": 0, "speed/wps": 12956.915498280603, "speed/FLOPS": 203506162568187.34, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.8026347160339355, "optim/lr": 0.002979311987371532, "optim/total_tokens": 2074083328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470403, "loss/out": 5.54099178314209, "created_at": "2025-01-15T00:09:16.940871+00:00"} {"global_step": 7913, "acc_step": 0, "speed/wps": 12968.616429173155, "speed/FLOPS": 203689941766618.84, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.0157206058502197, "optim/lr": 0.002979297804040376, "optim/total_tokens": 2074345472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469353, "loss/out": 5.47137975692749, "created_at": "2025-01-15T00:09:22.002835+00:00"} {"global_step": 7914, "acc_step": 0, "speed/wps": 12965.167678217047, "speed/FLOPS": 203635774393772.75, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.5271236896514893, "optim/lr": 0.0029792836158827516, "optim/total_tokens": 2074607616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 5.535153388977051, "created_at": "2025-01-15T00:09:27.067611+00:00"} {"global_step": 7915, "acc_step": 0, "speed/wps": 12961.088710047707, "speed/FLOPS": 203571708593583.62, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.314786911010742, "optim/lr": 0.0029792694228987036, "optim/total_tokens": 2074869760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451808, "loss/out": 5.539820194244385, "created_at": "2025-01-15T00:09:32.130051+00:00"} {"global_step": 7916, "acc_step": 0, "speed/wps": 12958.597879433957, "speed/FLOPS": 203532586676034.72, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.228243350982666, "optim/lr": 0.0029792552250882793, "optim/total_tokens": 2075131904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452847, "loss/out": 5.518457412719727, "created_at": "2025-01-15T00:09:37.196296+00:00"} {"global_step": 7917, "acc_step": 0, "speed/wps": 12963.10155841605, "speed/FLOPS": 203603323143151.97, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.42167854309082, "optim/lr": 0.0029792410224515242, "optim/total_tokens": 2075394048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308138, "loss/out": 5.582736015319824, "created_at": "2025-01-15T00:09:42.259269+00:00"} {"global_step": 7918, "acc_step": 0, "speed/wps": 12957.110275560484, "speed/FLOPS": 203509221813019.9, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.245019435882568, "optim/lr": 0.0029792268149884853, "optim/total_tokens": 2075656192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453660, "loss/out": 5.485653877258301, "created_at": "2025-01-15T00:09:47.327005+00:00"} {"global_step": 7919, "acc_step": 0, "speed/wps": 12961.962687337778, "speed/FLOPS": 203585435607894.84, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.4712677001953125, "optim/lr": 0.0029792126026992093, "optim/total_tokens": 2075918336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 5.538083076477051, "created_at": "2025-01-15T00:09:52.391782+00:00"} {"global_step": 7920, "acc_step": 0, "speed/wps": 12971.303665713585, "speed/FLOPS": 203732148509136.28, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0007, "optim/grad_norm": 20.746753692626953, "optim/lr": 0.0029791983855837418, "optim/total_tokens": 2076180480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430484, "loss/out": 5.48353385925293, "created_at": "2025-01-15T00:09:57.451819+00:00"} {"global_step": 7921, "acc_step": 0, "speed/wps": 12959.778393504148, "speed/FLOPS": 203551128271703.4, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0009, "optim/grad_norm": 5.307722568511963, "optim/lr": 0.002979184163642129, "optim/total_tokens": 2076442624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 5.504619598388672, "created_at": "2025-01-15T00:10:02.517519+00:00"} {"global_step": 7922, "acc_step": 0, "speed/wps": 12963.470389120394, "speed/FLOPS": 203609116136190.16, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.2404041290283203, "optim/lr": 0.0029791699368744186, "optim/total_tokens": 2076704768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333837, "loss/out": 5.499680519104004, "created_at": "2025-01-15T00:10:07.578482+00:00"} {"global_step": 7923, "acc_step": 0, "speed/wps": 12962.453183587479, "speed/FLOPS": 203593139525508.88, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.381978988647461, "optim/lr": 0.0029791557052806557, "optim/total_tokens": 2076966912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 386163, "loss/out": 5.461499214172363, "created_at": "2025-01-15T00:10:12.643577+00:00"} {"global_step": 7924, "acc_step": 0, "speed/wps": 12966.016451218544, "speed/FLOPS": 203649105540097.56, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0009, "optim/grad_norm": 3.997575283050537, "optim/lr": 0.002979141468860888, "optim/total_tokens": 2077229056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456851, "loss/out": 5.44301700592041, "created_at": "2025-01-15T00:10:17.706442+00:00"} {"global_step": 7925, "acc_step": 0, "speed/wps": 12964.088002945022, "speed/FLOPS": 203618816609996.28, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.116880416870117, "optim/lr": 0.00297912722761516, "optim/total_tokens": 2077491200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 5.507406234741211, "created_at": "2025-01-15T00:10:22.765342+00:00"} {"global_step": 7926, "acc_step": 0, "speed/wps": 12961.148189375286, "speed/FLOPS": 203572642798156.4, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.081113576889038, "optim/lr": 0.0029791129815435204, "optim/total_tokens": 2077753344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437201, "loss/out": 5.507692337036133, "created_at": "2025-01-15T00:10:27.829624+00:00"} {"global_step": 7927, "acc_step": 0, "speed/wps": 12964.675380932782, "speed/FLOPS": 203628042188436.0, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.461568832397461, "optim/lr": 0.002979098730646014, "optim/total_tokens": 2078015488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 414842, "loss/out": 5.468455791473389, "created_at": "2025-01-15T00:10:32.890584+00:00"} {"global_step": 7928, "acc_step": 0, "speed/wps": 12961.19290128704, "speed/FLOPS": 203573345060171.12, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 15.998598098754883, "optim/lr": 0.002979084474922688, "optim/total_tokens": 2078277632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 5.523426055908203, "created_at": "2025-01-15T00:10:37.954623+00:00"} {"global_step": 7929, "acc_step": 0, "speed/wps": 12970.153733193518, "speed/FLOPS": 203714087238735.75, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.485944747924805, "optim/lr": 0.002979070214373589, "optim/total_tokens": 2078539776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427351, "loss/out": 5.581789016723633, "created_at": "2025-01-15T00:10:43.016904+00:00"} {"global_step": 7930, "acc_step": 0, "speed/wps": 12965.953021068768, "speed/FLOPS": 203648109282433.2, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.1497514247894287, "optim/lr": 0.002979055948998763, "optim/total_tokens": 2078801920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 340518, "loss/out": 5.498628616333008, "created_at": "2025-01-15T00:10:48.075002+00:00"} {"global_step": 7931, "acc_step": 0, "speed/wps": 12968.627741302793, "speed/FLOPS": 203690119439159.3, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.4456539154052734, "optim/lr": 0.002979041678798257, "optim/total_tokens": 2079064064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435044, "loss/out": 5.530586242675781, "created_at": "2025-01-15T00:10:53.141050+00:00"} {"global_step": 7932, "acc_step": 0, "speed/wps": 12971.30325445414, "speed/FLOPS": 203732142049741.56, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.5253329277038574, "optim/lr": 0.002979027403772118, "optim/total_tokens": 2079326208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 5.392449378967285, "created_at": "2025-01-15T00:10:58.202919+00:00"} {"global_step": 7933, "acc_step": 0, "speed/wps": 12966.30623575009, "speed/FLOPS": 203653657004371.2, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.891656875610352, "optim/lr": 0.0029790131239203917, "optim/total_tokens": 2079588352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445933, "loss/out": 5.5033183097839355, "created_at": "2025-01-15T00:11:03.265610+00:00"} {"global_step": 7934, "acc_step": 0, "speed/wps": 12970.998559019843, "speed/FLOPS": 203727356389250.53, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0012, "optim/grad_norm": 7.4281158447265625, "optim/lr": 0.002978998839243125, "optim/total_tokens": 2079850496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 5.432865142822266, "created_at": "2025-01-15T00:11:08.322705+00:00"} {"global_step": 7935, "acc_step": 0, "speed/wps": 12973.574313368705, "speed/FLOPS": 203767812150756.78, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.909205436706543, "optim/lr": 0.002978984549740365, "optim/total_tokens": 2080112640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411402, "loss/out": 5.583549499511719, "created_at": "2025-01-15T00:11:13.378574+00:00"} {"global_step": 7936, "acc_step": 0, "speed/wps": 12965.027881049191, "speed/FLOPS": 203633578687150.12, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.263024806976318, "optim/lr": 0.0029789702554121568, "optim/total_tokens": 2080374784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303511, "loss/out": 5.557428359985352, "created_at": "2025-01-15T00:11:18.441819+00:00"} {"global_step": 7937, "acc_step": 0, "speed/wps": 12968.755039375084, "speed/FLOPS": 203692118830310.53, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.432344436645508, "optim/lr": 0.0029789559562585486, "optim/total_tokens": 2080636928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448890, "loss/out": 5.542585372924805, "created_at": "2025-01-15T00:11:23.508524+00:00"} {"global_step": 7938, "acc_step": 0, "speed/wps": 12969.118438558833, "speed/FLOPS": 203697826513854.1, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.359936237335205, "optim/lr": 0.002978941652279587, "optim/total_tokens": 2080899072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 5.471996784210205, "created_at": "2025-01-15T00:11:28.566713+00:00"} {"global_step": 7939, "acc_step": 0, "speed/wps": 12970.877696199002, "speed/FLOPS": 203725458072566.3, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.1754329204559326, "optim/lr": 0.0029789273434753178, "optim/total_tokens": 2081161216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 5.639978885650635, "created_at": "2025-01-15T00:11:33.634949+00:00"} {"global_step": 7940, "acc_step": 0, "speed/wps": 12970.471559492431, "speed/FLOPS": 203719079137499.28, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.576259136199951, "optim/lr": 0.002978913029845788, "optim/total_tokens": 2081423360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 5.490815162658691, "created_at": "2025-01-15T00:11:38.695215+00:00"} {"global_step": 7941, "acc_step": 0, "speed/wps": 12970.665326807373, "speed/FLOPS": 203722122519444.62, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.001, "optim/grad_norm": 2.578044891357422, "optim/lr": 0.0029788987113910444, "optim/total_tokens": 2081685504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441442, "loss/out": 5.4693756103515625, "created_at": "2025-01-15T00:11:43.753503+00:00"} {"global_step": 7942, "acc_step": 0, "speed/wps": 12964.328677696445, "speed/FLOPS": 203622596737689.16, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0009, "optim/grad_norm": 7.271699905395508, "optim/lr": 0.0029788843881111337, "optim/total_tokens": 2081947648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452615, "loss/out": 5.522348403930664, "created_at": "2025-01-15T00:11:48.812650+00:00"} {"global_step": 7943, "acc_step": 0, "speed/wps": 12965.618638568498, "speed/FLOPS": 203642857345776.75, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.001, "optim/grad_norm": 22.17057991027832, "optim/lr": 0.0029788700600061027, "optim/total_tokens": 2082209792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 5.590721130371094, "created_at": "2025-01-15T00:11:53.871015+00:00"} {"global_step": 7944, "acc_step": 0, "speed/wps": 12969.987343433779, "speed/FLOPS": 203711473858915.56, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.214590072631836, "optim/lr": 0.0029788557270759987, "optim/total_tokens": 2082471936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427880, "loss/out": 5.543402671813965, "created_at": "2025-01-15T00:11:58.929524+00:00"} {"global_step": 7945, "acc_step": 0, "speed/wps": 12964.670577459237, "speed/FLOPS": 203627966743283.3, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.838422775268555, "optim/lr": 0.0029788413893208675, "optim/total_tokens": 2082734080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308260, "loss/out": 5.630586624145508, "created_at": "2025-01-15T00:12:03.992037+00:00"} {"global_step": 7946, "acc_step": 0, "speed/wps": 12968.678649715488, "speed/FLOPS": 203690919025736.2, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.090062141418457, "optim/lr": 0.0029788270467407557, "optim/total_tokens": 2082996224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452093, "loss/out": 5.6489410400390625, "created_at": "2025-01-15T00:12:09.050272+00:00"} {"global_step": 7947, "acc_step": 0, "speed/wps": 12966.089624819117, "speed/FLOPS": 203650254832045.62, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0496954917907715, "optim/lr": 0.0029788126993357113, "optim/total_tokens": 2083258368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 5.509020805358887, "created_at": "2025-01-15T00:12:14.112703+00:00"} {"global_step": 7948, "acc_step": 0, "speed/wps": 12965.014986696118, "speed/FLOPS": 203633376163616.4, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.4638724327087402, "optim/lr": 0.00297879834710578, "optim/total_tokens": 2083520512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 5.465572357177734, "created_at": "2025-01-15T00:12:19.172988+00:00"} {"global_step": 7949, "acc_step": 0, "speed/wps": 12957.622532376838, "speed/FLOPS": 203517267510235.38, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.9883899688720703, "optim/lr": 0.0029787839900510095, "optim/total_tokens": 2083782656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 5.497800350189209, "created_at": "2025-01-15T00:12:24.235545+00:00"} {"global_step": 7950, "acc_step": 0, "speed/wps": 12966.61798623531, "speed/FLOPS": 203658553474131.6, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.523205757141113, "optim/lr": 0.002978769628171446, "optim/total_tokens": 2084044800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443022, "loss/out": 5.377857208251953, "created_at": "2025-01-15T00:12:29.299194+00:00"} {"global_step": 7951, "acc_step": 0, "speed/wps": 12965.911002221263, "speed/FLOPS": 203647449318693.16, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.9387996196746826, "optim/lr": 0.002978755261467136, "optim/total_tokens": 2084306944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308786, "loss/out": 5.514141082763672, "created_at": "2025-01-15T00:12:34.357573+00:00"} {"global_step": 7952, "acc_step": 0, "speed/wps": 12955.497818948954, "speed/FLOPS": 203483895966188.25, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.717944622039795, "optim/lr": 0.002978740889938128, "optim/total_tokens": 2084569088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 343732, "loss/out": 5.561824321746826, "created_at": "2025-01-15T00:12:39.428386+00:00"} {"global_step": 7953, "acc_step": 0, "speed/wps": 12966.074511254077, "speed/FLOPS": 203650017452738.47, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5305848121643066, "optim/lr": 0.002978726513584467, "optim/total_tokens": 2084831232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 403274, "loss/out": 5.504547595977783, "created_at": "2025-01-15T00:12:44.486238+00:00"} {"global_step": 7954, "acc_step": 0, "speed/wps": 12961.33243909902, "speed/FLOPS": 203575536693253.3, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.590911388397217, "optim/lr": 0.0029787121324062016, "optim/total_tokens": 2085093376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450994, "loss/out": 5.539625644683838, "created_at": "2025-01-15T00:12:49.546269+00:00"} {"global_step": 7955, "acc_step": 0, "speed/wps": 12960.024202380022, "speed/FLOPS": 203554989037875.8, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1120879650115967, "optim/lr": 0.0029786977464033772, "optim/total_tokens": 2085355520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 5.565701484680176, "created_at": "2025-01-15T00:12:54.611164+00:00"} {"global_step": 7956, "acc_step": 0, "speed/wps": 12955.22620296547, "speed/FLOPS": 203479629863928.16, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.889670729637146, "optim/lr": 0.0029786833555760415, "optim/total_tokens": 2085617664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455517, "loss/out": 5.414675712585449, "created_at": "2025-01-15T00:12:59.677987+00:00"} {"global_step": 7957, "acc_step": 0, "speed/wps": 12960.886582292635, "speed/FLOPS": 203568533899437.56, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0977613925933838, "optim/lr": 0.002978668959924241, "optim/total_tokens": 2085879808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447839, "loss/out": 5.470545768737793, "created_at": "2025-01-15T00:13:04.745445+00:00"} {"global_step": 7958, "acc_step": 0, "speed/wps": 12971.831971523916, "speed/FLOPS": 203740446277855.78, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.765844702720642, "optim/lr": 0.0029786545594480238, "optim/total_tokens": 2086141952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303928, "loss/out": 5.375428199768066, "created_at": "2025-01-15T00:13:09.801477+00:00"} {"global_step": 7959, "acc_step": 0, "speed/wps": 12968.56935486463, "speed/FLOPS": 203689202399916.6, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.5631086826324463, "optim/lr": 0.0029786401541474355, "optim/total_tokens": 2086404096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352687, "loss/out": 5.5375847816467285, "created_at": "2025-01-15T00:13:14.861411+00:00"} {"global_step": 7960, "acc_step": 0, "speed/wps": 12962.779114419774, "speed/FLOPS": 203598258717105.97, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.3378868103027344, "optim/lr": 0.002978625744022524, "optim/total_tokens": 2086666240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440390, "loss/out": 5.364381313323975, "created_at": "2025-01-15T00:13:19.921106+00:00"} {"global_step": 7961, "acc_step": 0, "speed/wps": 12967.783310300776, "speed/FLOPS": 203676856489902.84, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.07139778137207, "optim/lr": 0.002978611329073336, "optim/total_tokens": 2086928384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 5.549978733062744, "created_at": "2025-01-15T00:13:24.978670+00:00"} {"global_step": 7962, "acc_step": 0, "speed/wps": 12960.152692157777, "speed/FLOPS": 203557007146398.9, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.6265437602996826, "optim/lr": 0.0029785969092999184, "optim/total_tokens": 2087190528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 412410, "loss/out": 5.427236557006836, "created_at": "2025-01-15T00:13:30.039289+00:00"} {"global_step": 7963, "acc_step": 0, "speed/wps": 12969.88168401688, "speed/FLOPS": 203709814332581.9, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5717960596084595, "optim/lr": 0.0029785824847023187, "optim/total_tokens": 2087452672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459237, "loss/out": 5.514998435974121, "created_at": "2025-01-15T00:13:35.096048+00:00"} {"global_step": 7964, "acc_step": 0, "speed/wps": 12964.447111192729, "speed/FLOPS": 203624456898492.97, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.768789768218994, "optim/lr": 0.0029785680552805835, "optim/total_tokens": 2087714816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458410, "loss/out": 5.54774284362793, "created_at": "2025-01-15T00:13:40.155742+00:00"} {"global_step": 7965, "acc_step": 0, "speed/wps": 12971.637552429944, "speed/FLOPS": 203737392658825.12, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.707768440246582, "optim/lr": 0.00297855362103476, "optim/total_tokens": 2087976960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442784, "loss/out": 5.499068260192871, "created_at": "2025-01-15T00:13:45.212609+00:00"} {"global_step": 7966, "acc_step": 0, "speed/wps": 12966.725253085151, "speed/FLOPS": 203660238247408.47, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.8023059368133545, "optim/lr": 0.0029785391819648955, "optim/total_tokens": 2088239104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441687, "loss/out": 5.441317558288574, "created_at": "2025-01-15T00:13:50.271812+00:00"} {"global_step": 7967, "acc_step": 0, "speed/wps": 12968.356960675408, "speed/FLOPS": 203685866457314.66, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.033673048019409, "optim/lr": 0.002978524738071037, "optim/total_tokens": 2088501248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422788, "loss/out": 5.531471252441406, "created_at": "2025-01-15T00:13:55.330310+00:00"} {"global_step": 7968, "acc_step": 0, "speed/wps": 12966.95016704343, "speed/FLOPS": 203663770830187.4, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.4148943424224854, "optim/lr": 0.0029785102893532312, "optim/total_tokens": 2088763392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331967, "loss/out": 5.486783027648926, "created_at": "2025-01-15T00:14:00.389100+00:00"} {"global_step": 7969, "acc_step": 0, "speed/wps": 12961.542881906169, "speed/FLOPS": 203578841986718.3, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3519541025161743, "optim/lr": 0.0029784958358115257, "optim/total_tokens": 2089025536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337583, "loss/out": 5.416872024536133, "created_at": "2025-01-15T00:14:05.449890+00:00"} {"global_step": 7970, "acc_step": 0, "speed/wps": 12970.31352655621, "speed/FLOPS": 203716597013075.9, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.538052558898926, "optim/lr": 0.002978481377445968, "optim/total_tokens": 2089287680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450227, "loss/out": 5.355312347412109, "created_at": "2025-01-15T00:14:10.511868+00:00"} {"global_step": 7971, "acc_step": 0, "speed/wps": 12963.140299388024, "speed/FLOPS": 203603931623352.47, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.14226496219635, "optim/lr": 0.0029784669142566044, "optim/total_tokens": 2089549824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 5.431256294250488, "created_at": "2025-01-15T00:14:15.576139+00:00"} {"global_step": 7972, "acc_step": 0, "speed/wps": 12976.233443083622, "speed/FLOPS": 203809577436956.7, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.88100814819336, "optim/lr": 0.002978452446243483, "optim/total_tokens": 2089811968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 5.350551605224609, "created_at": "2025-01-15T00:14:20.630379+00:00"} {"global_step": 7973, "acc_step": 0, "speed/wps": 12961.64099475217, "speed/FLOPS": 203580382983785.9, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2588884830474854, "optim/lr": 0.00297843797340665, "optim/total_tokens": 2090074112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444878, "loss/out": 5.453448295593262, "created_at": "2025-01-15T00:14:25.690439+00:00"} {"global_step": 7974, "acc_step": 0, "speed/wps": 12969.744460107155, "speed/FLOPS": 203707659042514.25, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 52.31321716308594, "optim/lr": 0.0029784234957461533, "optim/total_tokens": 2090336256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462938, "loss/out": 5.2799224853515625, "created_at": "2025-01-15T00:14:30.746981+00:00"} {"global_step": 7975, "acc_step": 0, "speed/wps": 12970.635217040872, "speed/FLOPS": 203721649604186.56, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.78459358215332, "optim/lr": 0.00297840901326204, "optim/total_tokens": 2090598400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415331, "loss/out": 5.481320858001709, "created_at": "2025-01-15T00:14:35.803156+00:00"} {"global_step": 7976, "acc_step": 0, "speed/wps": 12963.396406846481, "speed/FLOPS": 203607954142916.06, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.7430760860443115, "optim/lr": 0.0029783945259543574, "optim/total_tokens": 2090860544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335714, "loss/out": 5.48068904876709, "created_at": "2025-01-15T00:14:40.867094+00:00"} {"global_step": 7977, "acc_step": 0, "speed/wps": 12974.461584730196, "speed/FLOPS": 203781747966727.62, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.277202844619751, "optim/lr": 0.0029783800338231527, "optim/total_tokens": 2091122688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448605, "loss/out": 5.443142890930176, "created_at": "2025-01-15T00:14:45.922704+00:00"} {"global_step": 7978, "acc_step": 0, "speed/wps": 12971.090412062002, "speed/FLOPS": 203728799067496.94, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6025397777557373, "optim/lr": 0.002978365536868473, "optim/total_tokens": 2091384832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447269, "loss/out": 5.420834541320801, "created_at": "2025-01-15T00:14:50.979095+00:00"} {"global_step": 7979, "acc_step": 0, "speed/wps": 12967.699270658293, "speed/FLOPS": 203675536531835.06, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.706499695777893, "optim/lr": 0.002978351035090366, "optim/total_tokens": 2091646976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 5.42111873626709, "created_at": "2025-01-15T00:14:56.043078+00:00"} {"global_step": 7980, "acc_step": 0, "speed/wps": 12968.338500672771, "speed/FLOPS": 203685576517606.53, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5774579048156738, "optim/lr": 0.0029783365284888792, "optim/total_tokens": 2091909120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462963, "loss/out": 5.362729072570801, "created_at": "2025-01-15T00:15:01.102638+00:00"} {"global_step": 7981, "acc_step": 0, "speed/wps": 12933.638486275584, "speed/FLOPS": 203140565108682.12, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1809436082839966, "optim/lr": 0.002978322017064059, "optim/total_tokens": 2092171264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418037, "loss/out": 5.284091949462891, "created_at": "2025-01-15T00:15:06.174773+00:00"} {"global_step": 7982, "acc_step": 0, "speed/wps": 12971.529618465283, "speed/FLOPS": 203735697407593.3, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7730904817581177, "optim/lr": 0.0029783075008159533, "optim/total_tokens": 2092433408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 5.389413356781006, "created_at": "2025-01-15T00:15:11.236372+00:00"} {"global_step": 7983, "acc_step": 0, "speed/wps": 12965.037345205355, "speed/FLOPS": 203633727334727.8, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.518006682395935, "optim/lr": 0.0029782929797446097, "optim/total_tokens": 2092695552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450462, "loss/out": 5.40042781829834, "created_at": "2025-01-15T00:15:16.296740+00:00"} {"global_step": 7984, "acc_step": 0, "speed/wps": 12972.33006403164, "speed/FLOPS": 203748269505140.4, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0009, "optim/grad_norm": 2.8696541786193848, "optim/lr": 0.002978278453850075, "optim/total_tokens": 2092957696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308424, "loss/out": 5.469796657562256, "created_at": "2025-01-15T00:15:21.352220+00:00"} {"global_step": 7985, "acc_step": 0, "speed/wps": 12967.538370772352, "speed/FLOPS": 203673009378026.75, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 13.164052963256836, "optim/lr": 0.0029782639231323974, "optim/total_tokens": 2093219840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349715, "loss/out": 5.340353488922119, "created_at": "2025-01-15T00:15:26.409971+00:00"} {"global_step": 7986, "acc_step": 0, "speed/wps": 12973.615688250395, "speed/FLOPS": 203768462000128.75, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.8573129177093506, "optim/lr": 0.002978249387591624, "optim/total_tokens": 2093481984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448077, "loss/out": 5.5188727378845215, "created_at": "2025-01-15T00:15:31.465005+00:00"} {"global_step": 7987, "acc_step": 0, "speed/wps": 12970.586251313687, "speed/FLOPS": 203720880530155.2, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6265324354171753, "optim/lr": 0.0029782348472278015, "optim/total_tokens": 2093744128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 5.386513710021973, "created_at": "2025-01-15T00:15:36.526892+00:00"} {"global_step": 7988, "acc_step": 0, "speed/wps": 12969.8634491624, "speed/FLOPS": 203709527929134.97, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.660226821899414, "optim/lr": 0.002978220302040978, "optim/total_tokens": 2094006272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450990, "loss/out": 5.376425743103027, "created_at": "2025-01-15T00:15:41.583443+00:00"} {"global_step": 7989, "acc_step": 0, "speed/wps": 12959.759607057953, "speed/FLOPS": 203550833204750.62, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0962499380111694, "optim/lr": 0.0029782057520312006, "optim/total_tokens": 2094268416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457380, "loss/out": 5.311435699462891, "created_at": "2025-01-15T00:15:46.644165+00:00"} {"global_step": 7990, "acc_step": 0, "speed/wps": 12967.877658147536, "speed/FLOPS": 203678338352483.72, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.921830654144287, "optim/lr": 0.0029781911971985173, "optim/total_tokens": 2094530560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 5.342567443847656, "created_at": "2025-01-15T00:15:51.704021+00:00"} {"global_step": 7991, "acc_step": 0, "speed/wps": 12961.323555706107, "speed/FLOPS": 203575397167360.53, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.904386281967163, "optim/lr": 0.0029781766375429754, "optim/total_tokens": 2094792704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334608, "loss/out": 5.330047130584717, "created_at": "2025-01-15T00:15:56.771986+00:00"} {"global_step": 7992, "acc_step": 0, "speed/wps": 12965.90129310624, "speed/FLOPS": 203647296823700.03, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5866456031799316, "optim/lr": 0.002978162073064622, "optim/total_tokens": 2095054848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345442, "loss/out": 5.383134841918945, "created_at": "2025-01-15T00:16:01.836413+00:00"} {"global_step": 7993, "acc_step": 0, "speed/wps": 12965.926684001726, "speed/FLOPS": 203647695622604.38, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.241228699684143, "optim/lr": 0.0029781475037635053, "optim/total_tokens": 2095316992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 406962, "loss/out": 5.327695846557617, "created_at": "2025-01-15T00:16:06.896594+00:00"} {"global_step": 7994, "acc_step": 0, "speed/wps": 12959.011367931804, "speed/FLOPS": 203539081081086.16, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8089678287506104, "optim/lr": 0.002978132929639672, "optim/total_tokens": 2095579136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 5.290835380554199, "created_at": "2025-01-15T00:16:11.961333+00:00"} {"global_step": 7995, "acc_step": 0, "speed/wps": 12960.490134661957, "speed/FLOPS": 203562307144621.66, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.40023672580719, "optim/lr": 0.0029781183506931703, "optim/total_tokens": 2095841280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461622, "loss/out": 5.3429975509643555, "created_at": "2025-01-15T00:16:17.024344+00:00"} {"global_step": 7996, "acc_step": 0, "speed/wps": 12964.163087005481, "speed/FLOPS": 203619995908342.12, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3478895425796509, "optim/lr": 0.002978103766924048, "optim/total_tokens": 2096103424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436666, "loss/out": 5.2779951095581055, "created_at": "2025-01-15T00:16:22.083530+00:00"} {"global_step": 7997, "acc_step": 0, "speed/wps": 12967.506234765811, "speed/FLOPS": 203672504637884.53, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.6419312953948975, "optim/lr": 0.002978089178332352, "optim/total_tokens": 2096365568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 5.474644660949707, "created_at": "2025-01-15T00:16:27.141988+00:00"} {"global_step": 7998, "acc_step": 0, "speed/wps": 12964.361932485896, "speed/FLOPS": 203623119049854.34, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.976798415184021, "optim/lr": 0.0029780745849181302, "optim/total_tokens": 2096627712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424462, "loss/out": 5.287177562713623, "created_at": "2025-01-15T00:16:32.225652+00:00"} {"global_step": 7999, "acc_step": 0, "speed/wps": 12970.713274639767, "speed/FLOPS": 203722875606039.62, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.476611137390137, "optim/lr": 0.00297805998668143, "optim/total_tokens": 2096889856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 412087, "loss/out": 5.300370216369629, "created_at": "2025-01-15T00:16:37.284248+00:00"} {"global_step": 8000, "acc_step": 0, "speed/wps": 12961.377894758829, "speed/FLOPS": 203576250636853.22, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8420099020004272, "optim/lr": 0.0029780453836222996, "optim/total_tokens": 2097152000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448890, "loss/out": 5.425071716308594, "created_at": "2025-01-15T00:16:42.345530+00:00"} {"global_step": 8001, "acc_step": 0, "speed/wps": 12310.566278914946, "speed/FLOPS": 193354359901148.0, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.268, "optim/grad_norm": 1.6699917316436768, "optim/lr": 0.0029780307757407854, "optim/total_tokens": 2097414144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357486, "loss/out": 5.431225776672363, "created_at": "2025-01-15T00:16:47.674227+00:00"} {"global_step": 8002, "acc_step": 0, "speed/wps": 12986.60584567923, "speed/FLOPS": 203972490272895.53, "speed/curr_iter_time": 1.2712, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7835835218429565, "optim/lr": 0.002978016163036937, "optim/total_tokens": 2097676288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416629, "loss/out": 5.286376953125, "created_at": "2025-01-15T00:16:52.724325+00:00"} {"global_step": 8003, "acc_step": 0, "speed/wps": 12981.559366135998, "speed/FLOPS": 203893228377079.47, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6057538986206055, "optim/lr": 0.0029780015455108007, "optim/total_tokens": 2097938432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455808, "loss/out": 5.411839008331299, "created_at": "2025-01-15T00:16:57.776515+00:00"} {"global_step": 8004, "acc_step": 0, "speed/wps": 12988.723751231288, "speed/FLOPS": 204005754889897.78, "speed/curr_iter_time": 1.2717, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.2319395542144775, "optim/lr": 0.002977986923162424, "optim/total_tokens": 2098200576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444644, "loss/out": 5.428064346313477, "created_at": "2025-01-15T00:17:02.830230+00:00"} {"global_step": 8005, "acc_step": 0, "speed/wps": 12976.255246871091, "speed/FLOPS": 203809919895400.44, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2223155498504639, "optim/lr": 0.0029779722959918553, "optim/total_tokens": 2098462720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 5.34510612487793, "created_at": "2025-01-15T00:17:07.884540+00:00"} {"global_step": 8006, "acc_step": 0, "speed/wps": 12972.466768706738, "speed/FLOPS": 203750416639914.72, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0748218297958374, "optim/lr": 0.002977957663999142, "optim/total_tokens": 2098724864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469069, "loss/out": 5.410606384277344, "created_at": "2025-01-15T00:17:12.945988+00:00"} {"global_step": 8007, "acc_step": 0, "speed/wps": 12979.773359226716, "speed/FLOPS": 203865176684338.5, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2307953834533691, "optim/lr": 0.0029779430271843323, "optim/total_tokens": 2098987008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 5.314661502838135, "created_at": "2025-01-15T00:17:17.998773+00:00"} {"global_step": 8008, "acc_step": 0, "speed/wps": 12981.88225894954, "speed/FLOPS": 203898299852417.34, "speed/curr_iter_time": 1.2719, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4386770725250244, "optim/lr": 0.0029779283855474736, "optim/total_tokens": 2099249152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 5.36094856262207, "created_at": "2025-01-15T00:17:23.051923+00:00"} {"global_step": 8009, "acc_step": 0, "speed/wps": 12977.400335820284, "speed/FLOPS": 203827905090864.62, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3546968698501587, "optim/lr": 0.0029779137390886137, "optim/total_tokens": 2099511296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454995, "loss/out": 5.265800476074219, "created_at": "2025-01-15T00:17:28.105991+00:00"} {"global_step": 8010, "acc_step": 0, "speed/wps": 12975.71675161357, "speed/FLOPS": 203801462087411.06, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.4959311485290527, "optim/lr": 0.0029778990878078, "optim/total_tokens": 2099773440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434515, "loss/out": 5.357555866241455, "created_at": "2025-01-15T00:17:33.165236+00:00"} {"global_step": 8011, "acc_step": 0, "speed/wps": 12982.766978366657, "speed/FLOPS": 203912195586595.06, "speed/curr_iter_time": 1.2721, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.792769432067871, "optim/lr": 0.002977884431705081, "optim/total_tokens": 2100035584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 5.304041385650635, "created_at": "2025-01-15T00:17:38.221692+00:00"} {"global_step": 8012, "acc_step": 0, "speed/wps": 12977.718412060594, "speed/FLOPS": 203832900915301.25, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.365074634552002, "optim/lr": 0.002977869770780504, "optim/total_tokens": 2100297728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433481, "loss/out": 5.360072135925293, "created_at": "2025-01-15T00:17:43.276488+00:00"} {"global_step": 8013, "acc_step": 0, "speed/wps": 12977.393124836908, "speed/FLOPS": 203827791832461.78, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0008, "optim/grad_norm": 13.279437065124512, "optim/lr": 0.002977855105034117, "optim/total_tokens": 2100559872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316000, "loss/out": 5.222574234008789, "created_at": "2025-01-15T00:17:48.333057+00:00"} {"global_step": 8014, "acc_step": 0, "speed/wps": 12976.625781281822, "speed/FLOPS": 203815739647488.62, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.090664863586426, "optim/lr": 0.002977840434465968, "optim/total_tokens": 2100822016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353744, "loss/out": 5.3214111328125, "created_at": "2025-01-15T00:17:53.388563+00:00"} {"global_step": 8015, "acc_step": 0, "speed/wps": 12976.328043146123, "speed/FLOPS": 203811063260932.38, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.6949985027313232, "optim/lr": 0.002977825759076105, "optim/total_tokens": 2101084160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455517, "loss/out": 5.480955123901367, "created_at": "2025-01-15T00:17:58.444675+00:00"} {"global_step": 8016, "acc_step": 0, "speed/wps": 12976.679073741023, "speed/FLOPS": 203816576678791.88, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0009, "optim/grad_norm": 2.0103342533111572, "optim/lr": 0.002977811078864575, "optim/total_tokens": 2101346304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455532, "loss/out": 5.344581604003906, "created_at": "2025-01-15T00:18:03.498719+00:00"} {"global_step": 8017, "acc_step": 0, "speed/wps": 12980.199113522242, "speed/FLOPS": 203871863740597.47, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.602428913116455, "optim/lr": 0.0029777963938314262, "optim/total_tokens": 2101608448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 5.269313812255859, "created_at": "2025-01-15T00:18:08.551696+00:00"} {"global_step": 8018, "acc_step": 0, "speed/wps": 12976.35955048482, "speed/FLOPS": 203811558126982.25, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.342781901359558, "optim/lr": 0.0029777817039767077, "optim/total_tokens": 2101870592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463770, "loss/out": 5.347649097442627, "created_at": "2025-01-15T00:18:13.609512+00:00"} {"global_step": 8019, "acc_step": 0, "speed/wps": 12970.518685863593, "speed/FLOPS": 203719819321916.8, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0009, "optim/grad_norm": 7.447512149810791, "optim/lr": 0.0029777670093004667, "optim/total_tokens": 2102132736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431842, "loss/out": 5.304441452026367, "created_at": "2025-01-15T00:18:18.670767+00:00"} {"global_step": 8020, "acc_step": 0, "speed/wps": 12973.395876750204, "speed/FLOPS": 203765009558465.62, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.9064271450042725, "optim/lr": 0.00297775230980275, "optim/total_tokens": 2102394880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432637, "loss/out": 5.358870983123779, "created_at": "2025-01-15T00:18:23.729122+00:00"} {"global_step": 8021, "acc_step": 0, "speed/wps": 12980.171756034528, "speed/FLOPS": 203871434053659.78, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.564682960510254, "optim/lr": 0.0029777376054836075, "optim/total_tokens": 2102657024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 5.5489888191223145, "created_at": "2025-01-15T00:18:28.781842+00:00"} {"global_step": 8022, "acc_step": 0, "speed/wps": 12977.426288725546, "speed/FLOPS": 203828312716904.8, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.852797746658325, "optim/lr": 0.0029777228963430856, "optim/total_tokens": 2102919168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460556, "loss/out": 5.2295637130737305, "created_at": "2025-01-15T00:18:33.838015+00:00"} {"global_step": 8023, "acc_step": 0, "speed/wps": 12978.301745524412, "speed/FLOPS": 203842062968934.7, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4231575727462769, "optim/lr": 0.0029777081823812333, "optim/total_tokens": 2103181312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432656, "loss/out": 5.330033779144287, "created_at": "2025-01-15T00:18:38.895302+00:00"} {"global_step": 8024, "acc_step": 0, "speed/wps": 12975.21910000144, "speed/FLOPS": 203793645784997.6, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.490643858909607, "optim/lr": 0.0029776934635980983, "optim/total_tokens": 2103443456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309312, "loss/out": 5.347035884857178, "created_at": "2025-01-15T00:18:43.950339+00:00"} {"global_step": 8025, "acc_step": 0, "speed/wps": 12977.877068286134, "speed/FLOPS": 203835392829339.84, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.001, "optim/grad_norm": 1.2936300039291382, "optim/lr": 0.0029776787399937283, "optim/total_tokens": 2103705600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366319, "loss/out": 5.201455593109131, "created_at": "2025-01-15T00:18:49.003929+00:00"} {"global_step": 8026, "acc_step": 0, "speed/wps": 12981.494188884457, "speed/FLOPS": 203892204678781.1, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 18.532304763793945, "optim/lr": 0.0029776640115681714, "optim/total_tokens": 2103967744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438014, "loss/out": 5.365231513977051, "created_at": "2025-01-15T00:18:54.062033+00:00"} {"global_step": 8027, "acc_step": 0, "speed/wps": 12978.852970280774, "speed/FLOPS": 203850720711195.53, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.764453411102295, "optim/lr": 0.0029776492783214763, "optim/total_tokens": 2104229888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448605, "loss/out": 5.2746381759643555, "created_at": "2025-01-15T00:18:59.115775+00:00"} {"global_step": 8028, "acc_step": 0, "speed/wps": 12975.161472421292, "speed/FLOPS": 203792740664662.75, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.189140319824219, "optim/lr": 0.0029776345402536906, "optim/total_tokens": 2104492032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 5.374068260192871, "created_at": "2025-01-15T00:19:04.172283+00:00"} {"global_step": 8029, "acc_step": 0, "speed/wps": 12979.30287460665, "speed/FLOPS": 203857787076866.75, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.225241780281067, "optim/lr": 0.002977619797364862, "optim/total_tokens": 2104754176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459222, "loss/out": 5.384634017944336, "created_at": "2025-01-15T00:19:09.230951+00:00"} {"global_step": 8030, "acc_step": 0, "speed/wps": 12984.28679306853, "speed/FLOPS": 203936066364932.22, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.4360923767089844, "optim/lr": 0.0029776050496550396, "optim/total_tokens": 2105016320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424656, "loss/out": 5.333893775939941, "created_at": "2025-01-15T00:19:14.282345+00:00"} {"global_step": 8031, "acc_step": 0, "speed/wps": 12972.034345164595, "speed/FLOPS": 203743624833973.16, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.252691626548767, "optim/lr": 0.0029775902971242705, "optim/total_tokens": 2105278464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471217, "loss/out": 5.237018585205078, "created_at": "2025-01-15T00:19:19.339837+00:00"} {"global_step": 8032, "acc_step": 0, "speed/wps": 12978.074046460139, "speed/FLOPS": 203838486642238.97, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.306898832321167, "optim/lr": 0.0029775755397726035, "optim/total_tokens": 2105540608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494240, "loss/out": 5.293257713317871, "created_at": "2025-01-15T00:19:24.403611+00:00"} {"global_step": 8033, "acc_step": 0, "speed/wps": 12985.77631144974, "speed/FLOPS": 203959461297921.88, "speed/curr_iter_time": 1.2713, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8359999656677246, "optim/lr": 0.002977560777600086, "optim/total_tokens": 2105802752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 5.338357448577881, "created_at": "2025-01-15T00:19:29.455292+00:00"} {"global_step": 8034, "acc_step": 0, "speed/wps": 12983.385890901249, "speed/FLOPS": 203921916458425.78, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.227885127067566, "optim/lr": 0.002977546010606767, "optim/total_tokens": 2106064896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458430, "loss/out": 5.29813289642334, "created_at": "2025-01-15T00:19:34.508350+00:00"} {"global_step": 8035, "acc_step": 0, "speed/wps": 12976.5959214235, "speed/FLOPS": 203815270657381.94, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.700319766998291, "optim/lr": 0.0029775312387926947, "optim/total_tokens": 2106327040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424894, "loss/out": 5.260922431945801, "created_at": "2025-01-15T00:19:39.565376+00:00"} {"global_step": 8036, "acc_step": 0, "speed/wps": 12986.849058951804, "speed/FLOPS": 203976310271551.34, "speed/curr_iter_time": 1.2723, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.990857481956482, "optim/lr": 0.0029775164621579165, "optim/total_tokens": 2106589184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324848, "loss/out": 5.375302791595459, "created_at": "2025-01-15T00:19:44.617113+00:00"} {"global_step": 8037, "acc_step": 0, "speed/wps": 12981.629934851255, "speed/FLOPS": 203894336755724.16, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.5715235471725464, "optim/lr": 0.002977501680702481, "optim/total_tokens": 2106851328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300816, "loss/out": 5.215602397918701, "created_at": "2025-01-15T00:19:49.674596+00:00"} {"global_step": 8038, "acc_step": 0, "speed/wps": 12983.539456229142, "speed/FLOPS": 203924328412921.22, "speed/curr_iter_time": 1.2721, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.288569450378418, "optim/lr": 0.0029774868944264367, "optim/total_tokens": 2107113472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441687, "loss/out": 5.34776496887207, "created_at": "2025-01-15T00:19:54.734186+00:00"} {"global_step": 8039, "acc_step": 0, "speed/wps": 12981.294819911496, "speed/FLOPS": 203889073315099.06, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.5162346363067627, "optim/lr": 0.0029774721033298318, "optim/total_tokens": 2107375616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447032, "loss/out": 5.417269706726074, "created_at": "2025-01-15T00:19:59.794154+00:00"} {"global_step": 8040, "acc_step": 0, "speed/wps": 12981.388181456065, "speed/FLOPS": 203890539686448.72, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.9370899200439453, "optim/lr": 0.002977457307412714, "optim/total_tokens": 2107637760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 5.309506416320801, "created_at": "2025-01-15T00:20:04.851021+00:00"} {"global_step": 8041, "acc_step": 0, "speed/wps": 12983.82986884288, "speed/FLOPS": 203928889742088.72, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.154821753501892, "optim/lr": 0.002977442506675132, "optim/total_tokens": 2107899904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430791, "loss/out": 5.266953468322754, "created_at": "2025-01-15T00:20:09.902428+00:00"} {"global_step": 8042, "acc_step": 0, "speed/wps": 12982.058308948752, "speed/FLOPS": 203901064959573.03, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.377139925956726, "optim/lr": 0.002977427701117134, "optim/total_tokens": 2108162048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445933, "loss/out": 5.301033973693848, "created_at": "2025-01-15T00:20:14.959700+00:00"} {"global_step": 8043, "acc_step": 0, "speed/wps": 12981.630444954715, "speed/FLOPS": 203894344767599.94, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0475387573242188, "optim/lr": 0.002977412890738769, "optim/total_tokens": 2108424192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 5.247448921203613, "created_at": "2025-01-15T00:20:20.015930+00:00"} {"global_step": 8044, "acc_step": 0, "speed/wps": 12981.685658502827, "speed/FLOPS": 203895211972244.4, "speed/curr_iter_time": 1.2719, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.455298662185669, "optim/lr": 0.0029773980755400838, "optim/total_tokens": 2108686336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439820, "loss/out": 5.333737373352051, "created_at": "2025-01-15T00:20:25.073123+00:00"} {"global_step": 8045, "acc_step": 0, "speed/wps": 12984.529325296959, "speed/FLOPS": 203939875666853.9, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1762747764587402, "optim/lr": 0.002977383255521128, "optim/total_tokens": 2108948480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 339460, "loss/out": 5.361575126647949, "created_at": "2025-01-15T00:20:30.125353+00:00"} {"global_step": 8046, "acc_step": 0, "speed/wps": 12982.325930475181, "speed/FLOPS": 203905268323396.12, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.079620599746704, "optim/lr": 0.0029773684306819495, "optim/total_tokens": 2109210624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434807, "loss/out": 5.247392177581787, "created_at": "2025-01-15T00:20:35.177036+00:00"} {"global_step": 8047, "acc_step": 0, "speed/wps": 12979.732148563113, "speed/FLOPS": 203864529414237.25, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.2537988424301147, "optim/lr": 0.0029773536010225966, "optim/total_tokens": 2109472768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424156, "loss/out": 5.306221008300781, "created_at": "2025-01-15T00:20:40.232198+00:00"} {"global_step": 8048, "acc_step": 0, "speed/wps": 12979.424921533444, "speed/FLOPS": 203859703991562.12, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7897093296051025, "optim/lr": 0.002977338766543118, "optim/total_tokens": 2109734912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324713, "loss/out": 5.202512741088867, "created_at": "2025-01-15T00:20:45.288955+00:00"} {"global_step": 8049, "acc_step": 0, "speed/wps": 12981.812437100629, "speed/FLOPS": 203897203204334.94, "speed/curr_iter_time": 1.2722, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1053000688552856, "optim/lr": 0.002977323927243562, "optim/total_tokens": 2109997056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292773, "loss/out": 5.3534040451049805, "created_at": "2025-01-15T00:20:50.341935+00:00"} {"global_step": 8050, "acc_step": 0, "speed/wps": 12981.920956585045, "speed/FLOPS": 203898907651959.1, "speed/curr_iter_time": 1.2723, "speed/data_load_time": 0.0007, "optim/grad_norm": 63.210323333740234, "optim/lr": 0.0029773090831239765, "optim/total_tokens": 2110259200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358828, "loss/out": 5.285573959350586, "created_at": "2025-01-15T00:20:55.399118+00:00"} {"global_step": 8051, "acc_step": 0, "speed/wps": 12985.168694110926, "speed/FLOPS": 203949917832662.1, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.6271109580993652, "optim/lr": 0.002977294234184411, "optim/total_tokens": 2110521344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439054, "loss/out": 5.354217529296875, "created_at": "2025-01-15T00:21:00.451991+00:00"} {"global_step": 8052, "acc_step": 0, "speed/wps": 12974.997372873666, "speed/FLOPS": 203790163255770.88, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.633264183998108, "optim/lr": 0.002977279380424913, "optim/total_tokens": 2110783488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446503, "loss/out": 5.384512424468994, "created_at": "2025-01-15T00:21:05.506610+00:00"} {"global_step": 8053, "acc_step": 0, "speed/wps": 12983.842551498297, "speed/FLOPS": 203929088940619.84, "speed/curr_iter_time": 1.2719, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4997669458389282, "optim/lr": 0.0029772645218455315, "optim/total_tokens": 2111045632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465057, "loss/out": 5.314019203186035, "created_at": "2025-01-15T00:21:10.562281+00:00"} {"global_step": 8054, "acc_step": 0, "speed/wps": 12980.429566504547, "speed/FLOPS": 203875483321360.12, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4099080562591553, "optim/lr": 0.0029772496584463144, "optim/total_tokens": 2111307776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 5.379423141479492, "created_at": "2025-01-15T00:21:15.621079+00:00"} {"global_step": 8055, "acc_step": 0, "speed/wps": 12975.772094320544, "speed/FLOPS": 203802331320672.94, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 12.026701927185059, "optim/lr": 0.0029772347902273107, "optim/total_tokens": 2111569920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447554, "loss/out": 5.272459983825684, "created_at": "2025-01-15T00:21:20.675694+00:00"} {"global_step": 8056, "acc_step": 0, "speed/wps": 12970.309694299953, "speed/FLOPS": 203716536822225.3, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.455267310142517, "optim/lr": 0.0029772199171885687, "optim/total_tokens": 2111832064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 5.302606105804443, "created_at": "2025-01-15T00:21:25.736467+00:00"} {"global_step": 8057, "acc_step": 0, "speed/wps": 12979.059709291481, "speed/FLOPS": 203853967831447.88, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4561707973480225, "optim/lr": 0.0029772050393301374, "optim/total_tokens": 2112094208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432431, "loss/out": 5.274743556976318, "created_at": "2025-01-15T00:21:30.791594+00:00"} {"global_step": 8058, "acc_step": 0, "speed/wps": 12979.854208351566, "speed/FLOPS": 203866446530942.62, "speed/curr_iter_time": 1.2722, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.9155540466308594, "optim/lr": 0.0029771901566520644, "optim/total_tokens": 2112356352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345083, "loss/out": 5.272361755371094, "created_at": "2025-01-15T00:21:35.850482+00:00"} {"global_step": 8059, "acc_step": 0, "speed/wps": 12979.805321433225, "speed/FLOPS": 203865678694712.4, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.974045991897583, "optim/lr": 0.0029771752691543987, "optim/total_tokens": 2112618496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310132, "loss/out": 5.411452770233154, "created_at": "2025-01-15T00:21:40.903348+00:00"} {"global_step": 8060, "acc_step": 0, "speed/wps": 12978.77388588849, "speed/FLOPS": 203849478582141.38, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9386340379714966, "optim/lr": 0.0029771603768371895, "optim/total_tokens": 2112880640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321097, "loss/out": 5.370563507080078, "created_at": "2025-01-15T00:21:45.956833+00:00"} {"global_step": 8061, "acc_step": 0, "speed/wps": 12972.240044366346, "speed/FLOPS": 203746855622597.0, "speed/curr_iter_time": 1.272, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3744711875915527, "optim/lr": 0.0029771454797004847, "optim/total_tokens": 2113142784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359885, "loss/out": 5.213066101074219, "created_at": "2025-01-15T00:21:51.013493+00:00"} {"global_step": 8062, "acc_step": 0, "speed/wps": 12977.095712858469, "speed/FLOPS": 203823120568653.5, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3263332843780518, "optim/lr": 0.0029771305777443326, "optim/total_tokens": 2113404928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437710, "loss/out": 5.224798679351807, "created_at": "2025-01-15T00:21:56.077156+00:00"} {"global_step": 8063, "acc_step": 0, "speed/wps": 12974.892883021077, "speed/FLOPS": 203788522099051.88, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2459290027618408, "optim/lr": 0.002977115670968783, "optim/total_tokens": 2113667072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 5.299929618835449, "created_at": "2025-01-15T00:22:01.136043+00:00"} {"global_step": 8064, "acc_step": 0, "speed/wps": 12980.863246767089, "speed/FLOPS": 203882294865822.5, "speed/curr_iter_time": 1.2722, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.615811586380005, "optim/lr": 0.002977100759373883, "optim/total_tokens": 2113929216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441971, "loss/out": 5.314250946044922, "created_at": "2025-01-15T00:22:06.188473+00:00"} {"global_step": 8065, "acc_step": 0, "speed/wps": 12978.039952965533, "speed/FLOPS": 203837951157056.84, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.9937644004821777, "optim/lr": 0.0029770858429596825, "optim/total_tokens": 2114191360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424963, "loss/out": 5.353209495544434, "created_at": "2025-01-15T00:22:11.247421+00:00"} {"global_step": 8066, "acc_step": 0, "speed/wps": 12978.656973838635, "speed/FLOPS": 203847642317744.44, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.001, "optim/grad_norm": 3.490935802459717, "optim/lr": 0.0029770709217262297, "optim/total_tokens": 2114453504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 5.300739288330078, "created_at": "2025-01-15T00:22:16.303838+00:00"} {"global_step": 8067, "acc_step": 0, "speed/wps": 12981.324881004046, "speed/FLOPS": 203889545465865.8, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7400857210159302, "optim/lr": 0.0029770559956735733, "optim/total_tokens": 2114715648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457902, "loss/out": 5.375187873840332, "created_at": "2025-01-15T00:22:21.361644+00:00"} {"global_step": 8068, "acc_step": 0, "speed/wps": 12974.781732220697, "speed/FLOPS": 203786776322994.7, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.3168792724609375, "optim/lr": 0.0029770410648017618, "optim/total_tokens": 2114977792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313596, "loss/out": 5.368476867675781, "created_at": "2025-01-15T00:22:26.420756+00:00"} {"global_step": 8069, "acc_step": 0, "speed/wps": 12979.511610240614, "speed/FLOPS": 203861065556830.97, "speed/curr_iter_time": 1.2714, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5852000713348389, "optim/lr": 0.002977026129110844, "optim/total_tokens": 2115239936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354978, "loss/out": 5.33010721206665, "created_at": "2025-01-15T00:22:31.477911+00:00"} {"global_step": 8070, "acc_step": 0, "speed/wps": 12973.778138280015, "speed/FLOPS": 203771013501073.7, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.789086103439331, "optim/lr": 0.002977011188600869, "optim/total_tokens": 2115502080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440919, "loss/out": 5.258904933929443, "created_at": "2025-01-15T00:22:36.534017+00:00"} {"global_step": 8071, "acc_step": 0, "speed/wps": 12980.258715701835, "speed/FLOPS": 203872799874729.94, "speed/curr_iter_time": 1.2723, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7501418590545654, "optim/lr": 0.002976996243271886, "optim/total_tokens": 2115764224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440628, "loss/out": 5.34798526763916, "created_at": "2025-01-15T00:22:41.587636+00:00"} {"global_step": 8072, "acc_step": 0, "speed/wps": 12977.849004941882, "speed/FLOPS": 203834952055955.28, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.8742828369140625, "optim/lr": 0.002976981293123942, "optim/total_tokens": 2116026368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447554, "loss/out": 5.175487041473389, "created_at": "2025-01-15T00:22:46.644781+00:00"} {"global_step": 8073, "acc_step": 0, "speed/wps": 12980.555166441975, "speed/FLOPS": 203877456040969.84, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.638310432434082, "optim/lr": 0.0029769663381570876, "optim/total_tokens": 2116288512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440628, "loss/out": 5.382535934448242, "created_at": "2025-01-15T00:22:51.698933+00:00"} {"global_step": 8074, "acc_step": 0, "speed/wps": 12973.982785278497, "speed/FLOPS": 203774227763399.28, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.874506950378418, "optim/lr": 0.0029769513783713705, "optim/total_tokens": 2116550656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 5.487038612365723, "created_at": "2025-01-15T00:22:56.753993+00:00"} {"global_step": 8075, "acc_step": 0, "speed/wps": 12977.227968467083, "speed/FLOPS": 203825197824725.03, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.033053398132324, "optim/lr": 0.0029769364137668397, "optim/total_tokens": 2116812800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 5.385959625244141, "created_at": "2025-01-15T00:23:01.811459+00:00"} {"global_step": 8076, "acc_step": 0, "speed/wps": 12976.438373342036, "speed/FLOPS": 203812796148268.78, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.4258790016174316, "optim/lr": 0.0029769214443435446, "optim/total_tokens": 2117074944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422813, "loss/out": 5.376742362976074, "created_at": "2025-01-15T00:23:06.868295+00:00"} {"global_step": 8077, "acc_step": 0, "speed/wps": 12977.402705179333, "speed/FLOPS": 203827942304904.3, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.7231526374816895, "optim/lr": 0.0029769064701015333, "optim/total_tokens": 2117337088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 5.264759063720703, "created_at": "2025-01-15T00:23:11.925628+00:00"} {"global_step": 8078, "acc_step": 0, "speed/wps": 12971.727711987905, "speed/FLOPS": 203738808738575.12, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.261842727661133, "optim/lr": 0.002976891491040855, "optim/total_tokens": 2117599232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444119, "loss/out": 5.284278869628906, "created_at": "2025-01-15T00:23:16.985221+00:00"} {"global_step": 8079, "acc_step": 0, "speed/wps": 12976.806888885347, "speed/FLOPS": 203818584191269.9, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.827620506286621, "optim/lr": 0.0029768765071615582, "optim/total_tokens": 2117861376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328985, "loss/out": 5.313246726989746, "created_at": "2025-01-15T00:23:22.040159+00:00"} {"global_step": 8080, "acc_step": 0, "speed/wps": 12985.461992830411, "speed/FLOPS": 203954524492086.28, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.0173985958099365, "optim/lr": 0.002976861518463693, "optim/total_tokens": 2118123520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292773, "loss/out": 5.3508524894714355, "created_at": "2025-01-15T00:23:27.091648+00:00"} {"global_step": 8081, "acc_step": 0, "speed/wps": 12978.3747730105, "speed/FLOPS": 203843209965954.0, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7322454452514648, "optim/lr": 0.0029768465249473065, "optim/total_tokens": 2118385664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348659, "loss/out": 5.281074047088623, "created_at": "2025-01-15T00:23:32.148987+00:00"} {"global_step": 8082, "acc_step": 0, "speed/wps": 12975.267356645149, "speed/FLOPS": 203794403721898.6, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3055131435394287, "optim/lr": 0.0029768315266124486, "optim/total_tokens": 2118647808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431076, "loss/out": 5.190755844116211, "created_at": "2025-01-15T00:23:37.203497+00:00"} {"global_step": 8083, "acc_step": 0, "speed/wps": 12973.644771376688, "speed/FLOPS": 203768918790591.62, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9232361316680908, "optim/lr": 0.002976816523459169, "optim/total_tokens": 2118909952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457902, "loss/out": 5.332365036010742, "created_at": "2025-01-15T00:23:42.259852+00:00"} {"global_step": 8084, "acc_step": 0, "speed/wps": 12970.986826732615, "speed/FLOPS": 203727172117557.75, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3071542978286743, "optim/lr": 0.002976801515487515, "optim/total_tokens": 2119172096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 5.269426345825195, "created_at": "2025-01-15T00:23:47.317230+00:00"} {"global_step": 8085, "acc_step": 0, "speed/wps": 12976.39705942086, "speed/FLOPS": 203812147256365.44, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.9109129905700684, "optim/lr": 0.0029767865026975367, "optim/total_tokens": 2119434240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 5.221053123474121, "created_at": "2025-01-15T00:23:52.371121+00:00"} {"global_step": 8086, "acc_step": 0, "speed/wps": 12978.665883401054, "speed/FLOPS": 203847782254665.25, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.305440664291382, "optim/lr": 0.0029767714850892826, "optim/total_tokens": 2119696384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 5.2703375816345215, "created_at": "2025-01-15T00:23:57.431136+00:00"} {"global_step": 8087, "acc_step": 0, "speed/wps": 12970.927551152286, "speed/FLOPS": 203726241113115.5, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.3282827138900757, "optim/lr": 0.0029767564626628023, "optim/total_tokens": 2119958528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 5.2655744552612305, "created_at": "2025-01-15T00:24:02.490020+00:00"} {"global_step": 8088, "acc_step": 0, "speed/wps": 12968.786490828088, "speed/FLOPS": 203692612818598.78, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.7561143636703491, "optim/lr": 0.002976741435418144, "optim/total_tokens": 2120220672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455769, "loss/out": 5.181332111358643, "created_at": "2025-01-15T00:24:07.552396+00:00"} {"global_step": 8089, "acc_step": 0, "speed/wps": 12971.321211213522, "speed/FLOPS": 203732424085322.84, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1474052667617798, "optim/lr": 0.002976726403355357, "optim/total_tokens": 2120482816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 5.216414928436279, "created_at": "2025-01-15T00:24:12.613277+00:00"} {"global_step": 8090, "acc_step": 0, "speed/wps": 12971.451833190169, "speed/FLOPS": 203734475682959.47, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.421427607536316, "optim/lr": 0.0029767113664744905, "optim/total_tokens": 2120744960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325907, "loss/out": 5.299371242523193, "created_at": "2025-01-15T00:24:17.673160+00:00"} {"global_step": 8091, "acc_step": 0, "speed/wps": 12976.95120952262, "speed/FLOPS": 203820850945195.6, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2367159128189087, "optim/lr": 0.002976696324775594, "optim/total_tokens": 2121007104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303689, "loss/out": 5.219836235046387, "created_at": "2025-01-15T00:24:22.727435+00:00"} {"global_step": 8092, "acc_step": 0, "speed/wps": 12958.474610719053, "speed/FLOPS": 203530650571478.2, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.522721529006958, "optim/lr": 0.002976681278258716, "optim/total_tokens": 2121269248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444357, "loss/out": 5.3295793533325195, "created_at": "2025-01-15T00:24:27.793122+00:00"} {"global_step": 8093, "acc_step": 0, "speed/wps": 12976.83556753135, "speed/FLOPS": 203819034628813.97, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.933701515197754, "optim/lr": 0.002976666226923906, "optim/total_tokens": 2121531392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 5.320193290710449, "created_at": "2025-01-15T00:24:32.846901+00:00"} {"global_step": 8094, "acc_step": 0, "speed/wps": 12977.957611356533, "speed/FLOPS": 203836657868938.25, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5912940502166748, "optim/lr": 0.002976651170771212, "optim/total_tokens": 2121793536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497894, "loss/out": 5.248652935028076, "created_at": "2025-01-15T00:24:37.900941+00:00"} {"global_step": 8095, "acc_step": 0, "speed/wps": 12976.20803334079, "speed/FLOPS": 203809178342028.88, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5693575143814087, "optim/lr": 0.002976636109800684, "optim/total_tokens": 2122055680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450990, "loss/out": 5.21548318862915, "created_at": "2025-01-15T00:24:42.959121+00:00"} {"global_step": 8096, "acc_step": 0, "speed/wps": 12972.545649504775, "speed/FLOPS": 203751655571241.47, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.8719773292541504, "optim/lr": 0.0029766210440123715, "optim/total_tokens": 2122317824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453660, "loss/out": 5.288405418395996, "created_at": "2025-01-15T00:24:48.018708+00:00"} {"global_step": 8097, "acc_step": 0, "speed/wps": 12980.881483829919, "speed/FLOPS": 203882581303954.6, "speed/curr_iter_time": 1.2714, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5743504762649536, "optim/lr": 0.002976605973406323, "optim/total_tokens": 2122579968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 5.266900062561035, "created_at": "2025-01-15T00:24:53.071022+00:00"} {"global_step": 8098, "acc_step": 0, "speed/wps": 12977.914971346136, "speed/FLOPS": 203835988148981.44, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.671956181526184, "optim/lr": 0.002976590897982588, "optim/total_tokens": 2122842112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 5.216079235076904, "created_at": "2025-01-15T00:24:58.128901+00:00"} {"global_step": 8099, "acc_step": 0, "speed/wps": 12976.425066234506, "speed/FLOPS": 203812587141859.12, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2153644561767578, "optim/lr": 0.0029765758177412154, "optim/total_tokens": 2123104256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 5.315208911895752, "created_at": "2025-01-15T00:25:03.183213+00:00"} {"global_step": 8100, "acc_step": 0, "speed/wps": 12972.474823190041, "speed/FLOPS": 203750543146643.03, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.6863408088684082, "optim/lr": 0.0029765607326822548, "optim/total_tokens": 2123366400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463485, "loss/out": 5.223021984100342, "created_at": "2025-01-15T00:25:08.243562+00:00"} {"global_step": 8101, "acc_step": 0, "speed/wps": 12951.544977288642, "speed/FLOPS": 203421811156139.4, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.498647928237915, "optim/lr": 0.0029765456428057554, "optim/total_tokens": 2123628544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308260, "loss/out": 5.252359390258789, "created_at": "2025-01-15T00:25:13.307711+00:00"} {"global_step": 8102, "acc_step": 0, "speed/wps": 12976.162631529927, "speed/FLOPS": 203808465244200.66, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6142425537109375, "optim/lr": 0.0029765305481117654, "optim/total_tokens": 2123890688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353744, "loss/out": 5.234501361846924, "created_at": "2025-01-15T00:25:18.372720+00:00"} {"global_step": 8103, "acc_step": 0, "speed/wps": 12980.897197054612, "speed/FLOPS": 203882828101740.62, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.131924629211426, "optim/lr": 0.0029765154486003356, "optim/total_tokens": 2124152832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428688, "loss/out": 5.273562431335449, "created_at": "2025-01-15T00:25:23.427600+00:00"} {"global_step": 8104, "acc_step": 0, "speed/wps": 12977.407183030966, "speed/FLOPS": 203828012635717.44, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4096705913543701, "optim/lr": 0.0029765003442715145, "optim/total_tokens": 2124414976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 5.229959011077881, "created_at": "2025-01-15T00:25:28.486679+00:00"} {"global_step": 8105, "acc_step": 0, "speed/wps": 12972.331706938608, "speed/FLOPS": 203748295309251.94, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4856067895889282, "optim/lr": 0.0029764852351253507, "optim/total_tokens": 2124677120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443545, "loss/out": 5.145483016967773, "created_at": "2025-01-15T00:25:33.544564+00:00"} {"global_step": 8106, "acc_step": 0, "speed/wps": 12975.624811504618, "speed/FLOPS": 203800018041661.0, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4356080293655396, "optim/lr": 0.002976470121161895, "optim/total_tokens": 2124939264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447554, "loss/out": 5.14573860168457, "created_at": "2025-01-15T00:25:38.602776+00:00"} {"global_step": 8107, "acc_step": 0, "speed/wps": 12978.13494319129, "speed/FLOPS": 203839443109082.6, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.4444353580474854, "optim/lr": 0.0029764550023811957, "optim/total_tokens": 2125201408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 5.238959312438965, "created_at": "2025-01-15T00:25:43.659937+00:00"} {"global_step": 8108, "acc_step": 0, "speed/wps": 12971.268381182646, "speed/FLOPS": 203731594317091.84, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.817700743675232, "optim/lr": 0.002976439878783302, "optim/total_tokens": 2125463552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456045, "loss/out": 5.321167469024658, "created_at": "2025-01-15T00:25:48.720297+00:00"} {"global_step": 8109, "acc_step": 0, "speed/wps": 12971.764390703787, "speed/FLOPS": 203739384828209.88, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.625591516494751, "optim/lr": 0.002976424750368264, "optim/total_tokens": 2125725696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320435, "loss/out": 5.247017860412598, "created_at": "2025-01-15T00:25:53.776230+00:00"} {"global_step": 8110, "acc_step": 0, "speed/wps": 12975.281694942665, "speed/FLOPS": 203794628924563.97, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.434116244316101, "optim/lr": 0.002976409617136131, "optim/total_tokens": 2125987840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348536, "loss/out": 5.18808650970459, "created_at": "2025-01-15T00:25:58.834329+00:00"} {"global_step": 8111, "acc_step": 0, "speed/wps": 12972.54441309874, "speed/FLOPS": 203751636151785.7, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1541451215744019, "optim/lr": 0.0029763944790869514, "optim/total_tokens": 2126249984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451523, "loss/out": 5.21885347366333, "created_at": "2025-01-15T00:26:03.890006+00:00"} {"global_step": 8112, "acc_step": 0, "speed/wps": 12971.692085601451, "speed/FLOPS": 203738249177220.06, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2888522148132324, "optim/lr": 0.0029763793362207755, "optim/total_tokens": 2126512128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453393, "loss/out": 5.274181365966797, "created_at": "2025-01-15T00:26:08.954163+00:00"} {"global_step": 8113, "acc_step": 0, "speed/wps": 12971.254618426523, "speed/FLOPS": 203731378154094.94, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2724609375, "optim/lr": 0.002976364188537652, "optim/total_tokens": 2126774272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428118, "loss/out": 5.236440658569336, "created_at": "2025-01-15T00:26:14.014915+00:00"} {"global_step": 8114, "acc_step": 0, "speed/wps": 12949.053607428008, "speed/FLOPS": 203382680768977.2, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6841354370117188, "optim/lr": 0.0029763490360376313, "optim/total_tokens": 2127036416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448597, "loss/out": 5.204732894897461, "created_at": "2025-01-15T00:26:19.079832+00:00"} {"global_step": 8115, "acc_step": 0, "speed/wps": 12974.916766062519, "speed/FLOPS": 203788897215036.44, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.3554623126983643, "optim/lr": 0.0029763338787207622, "optim/total_tokens": 2127298560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422265, "loss/out": 5.243582248687744, "created_at": "2025-01-15T00:26:24.141889+00:00"} {"global_step": 8116, "acc_step": 0, "speed/wps": 12971.30760854529, "speed/FLOPS": 203732210436726.22, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2900426387786865, "optim/lr": 0.0029763187165870947, "optim/total_tokens": 2127560704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 5.22558069229126, "created_at": "2025-01-15T00:26:29.197796+00:00"} {"global_step": 8117, "acc_step": 0, "speed/wps": 12972.363811863559, "speed/FLOPS": 203748799561215.1, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.427997350692749, "optim/lr": 0.0029763035496366765, "optim/total_tokens": 2127822848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445404, "loss/out": 5.192841529846191, "created_at": "2025-01-15T00:26:34.257786+00:00"} {"global_step": 8118, "acc_step": 0, "speed/wps": 12967.49368914914, "speed/FLOPS": 203672307591735.22, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3798490762710571, "optim/lr": 0.002976288377869559, "optim/total_tokens": 2128084992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291186, "loss/out": 5.235651016235352, "created_at": "2025-01-15T00:26:39.318815+00:00"} {"global_step": 8119, "acc_step": 0, "speed/wps": 12972.62149162775, "speed/FLOPS": 203752846776000.8, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.9806931018829346, "optim/lr": 0.0029762732012857916, "optim/total_tokens": 2128347136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417683, "loss/out": 5.193830490112305, "created_at": "2025-01-15T00:26:44.389709+00:00"} {"global_step": 8120, "acc_step": 0, "speed/wps": 12971.179614788154, "speed/FLOPS": 203730200118888.53, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3507050275802612, "optim/lr": 0.0029762580198854226, "optim/total_tokens": 2128609280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 5.175182819366455, "created_at": "2025-01-15T00:26:49.450209+00:00"} {"global_step": 8121, "acc_step": 0, "speed/wps": 12976.544357375587, "speed/FLOPS": 203814460773153.03, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2089513540267944, "optim/lr": 0.0029762428336685036, "optim/total_tokens": 2128871424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438251, "loss/out": 5.249732971191406, "created_at": "2025-01-15T00:26:54.506625+00:00"} {"global_step": 8122, "acc_step": 0, "speed/wps": 12977.254406510167, "speed/FLOPS": 203825613070521.34, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1405470371246338, "optim/lr": 0.002976227642635081, "optim/total_tokens": 2129133568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 5.281187057495117, "created_at": "2025-01-15T00:26:59.560427+00:00"} {"global_step": 8123, "acc_step": 0, "speed/wps": 12974.196165144107, "speed/FLOPS": 203777579187403.34, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.126665472984314, "optim/lr": 0.002976212446785207, "optim/total_tokens": 2129395712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 5.3840436935424805, "created_at": "2025-01-15T00:27:04.620183+00:00"} {"global_step": 8124, "acc_step": 0, "speed/wps": 12979.261049284234, "speed/FLOPS": 203857130152704.22, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.9579051733016968, "optim/lr": 0.0029761972461189307, "optim/total_tokens": 2129657856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481474, "loss/out": 5.175559997558594, "created_at": "2025-01-15T00:27:09.673727+00:00"} {"global_step": 8125, "acc_step": 0, "speed/wps": 12974.185816885147, "speed/FLOPS": 203777416653776.38, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2643424272537231, "optim/lr": 0.0029761820406363007, "optim/total_tokens": 2129920000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 5.270733833312988, "created_at": "2025-01-15T00:27:14.728669+00:00"} {"global_step": 8126, "acc_step": 0, "speed/wps": 12973.003578373497, "speed/FLOPS": 203758847973384.97, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.001, "optim/grad_norm": 1.9593206644058228, "optim/lr": 0.0029761668303373676, "optim/total_tokens": 2130182144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 5.243729591369629, "created_at": "2025-01-15T00:27:19.789887+00:00"} {"global_step": 8127, "acc_step": 0, "speed/wps": 12978.889770692975, "speed/FLOPS": 203851298712238.22, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6066783666610718, "optim/lr": 0.0029761516152221807, "optim/total_tokens": 2130444288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456330, "loss/out": 5.266417503356934, "created_at": "2025-01-15T00:27:24.850042+00:00"} {"global_step": 8128, "acc_step": 0, "speed/wps": 12979.01361005334, "speed/FLOPS": 203853243779565.88, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3171658515930176, "optim/lr": 0.0029761363952907896, "optim/total_tokens": 2130706432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 5.169533729553223, "created_at": "2025-01-15T00:27:29.907892+00:00"} {"global_step": 8129, "acc_step": 0, "speed/wps": 12975.17192032794, "speed/FLOPS": 203792904763393.6, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.353743314743042, "optim/lr": 0.0029761211705432436, "optim/total_tokens": 2130968576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320571, "loss/out": 5.0635786056518555, "created_at": "2025-01-15T00:27:34.965789+00:00"} {"global_step": 8130, "acc_step": 0, "speed/wps": 12981.492718476353, "speed/FLOPS": 203892181584001.25, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8689237833023071, "optim/lr": 0.0029761059409795934, "optim/total_tokens": 2131230720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431521, "loss/out": 5.196727275848389, "created_at": "2025-01-15T00:27:40.017928+00:00"} {"global_step": 8131, "acc_step": 0, "speed/wps": 12974.806074181206, "speed/FLOPS": 203787158646933.84, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2741724252700806, "optim/lr": 0.0029760907065998874, "optim/total_tokens": 2131492864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442778, "loss/out": 5.228516578674316, "created_at": "2025-01-15T00:27:45.073647+00:00"} {"global_step": 8132, "acc_step": 0, "speed/wps": 12966.806145228835, "speed/FLOPS": 203661508769682.47, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.6106975078582764, "optim/lr": 0.002976075467404176, "optim/total_tokens": 2131755008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 5.202539443969727, "created_at": "2025-01-15T00:27:50.134482+00:00"} {"global_step": 8133, "acc_step": 0, "speed/wps": 12976.620964162828, "speed/FLOPS": 203815663988015.4, "speed/curr_iter_time": 1.2723, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1999040842056274, "optim/lr": 0.0029760602233925093, "optim/total_tokens": 2132017152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425201, "loss/out": 5.151880264282227, "created_at": "2025-01-15T00:27:55.192151+00:00"} {"global_step": 8134, "acc_step": 0, "speed/wps": 12966.480867216651, "speed/FLOPS": 203656399831522.3, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 296.32989501953125, "optim/lr": 0.0029760449745649363, "optim/total_tokens": 2132279296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500037, "loss/out": 5.236234664916992, "created_at": "2025-01-15T00:28:00.251421+00:00"} {"global_step": 8135, "acc_step": 0, "speed/wps": 12974.65567940752, "speed/FLOPS": 203784796490349.03, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6211036443710327, "optim/lr": 0.0029760297209215068, "optim/total_tokens": 2132541440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462672, "loss/out": 5.219003677368164, "created_at": "2025-01-15T00:28:05.307189+00:00"} {"global_step": 8136, "acc_step": 0, "speed/wps": 12975.419954792753, "speed/FLOPS": 203796800485498.53, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7367944717407227, "optim/lr": 0.002976014462462271, "optim/total_tokens": 2132803584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320196, "loss/out": 5.128913879394531, "created_at": "2025-01-15T00:28:10.362451+00:00"} {"global_step": 8137, "acc_step": 0, "speed/wps": 12976.188115273128, "speed/FLOPS": 203808865501405.94, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.463179111480713, "optim/lr": 0.0029759991991872787, "optim/total_tokens": 2133065728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310371, "loss/out": 5.183581352233887, "created_at": "2025-01-15T00:28:15.420355+00:00"} {"global_step": 8138, "acc_step": 0, "speed/wps": 12974.006050215354, "speed/FLOPS": 203774593171202.0, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7443907260894775, "optim/lr": 0.0029759839310965794, "optim/total_tokens": 2133327872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 400357, "loss/out": 5.234251022338867, "created_at": "2025-01-15T00:28:20.476495+00:00"} {"global_step": 8139, "acc_step": 0, "speed/wps": 12972.828213925193, "speed/FLOPS": 203756093633748.34, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.1475813388824463, "optim/lr": 0.002975968658190223, "optim/total_tokens": 2133590016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365644, "loss/out": 5.194864273071289, "created_at": "2025-01-15T00:28:25.536778+00:00"} {"global_step": 8140, "acc_step": 0, "speed/wps": 12979.239438881843, "speed/FLOPS": 203856790731638.84, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6336463689804077, "optim/lr": 0.0029759533804682593, "optim/total_tokens": 2133852160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 387212, "loss/out": 5.194935321807861, "created_at": "2025-01-15T00:28:30.593715+00:00"} {"global_step": 8141, "acc_step": 0, "speed/wps": 12974.235481980759, "speed/FLOPS": 203778196712349.97, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4423681497573853, "optim/lr": 0.002975938097930738, "optim/total_tokens": 2134114304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435329, "loss/out": 5.171329975128174, "created_at": "2025-01-15T00:28:35.648982+00:00"} {"global_step": 8142, "acc_step": 0, "speed/wps": 12978.641012299582, "speed/FLOPS": 203847391619840.16, "speed/curr_iter_time": 1.2721, "speed/data_load_time": 0.0008, "optim/grad_norm": 11.213980674743652, "optim/lr": 0.0029759228105777098, "optim/total_tokens": 2134376448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459528, "loss/out": 5.342397689819336, "created_at": "2025-01-15T00:28:40.708366+00:00"} {"global_step": 8143, "acc_step": 0, "speed/wps": 12975.643447778575, "speed/FLOPS": 203800310749952.44, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.747406959533691, "optim/lr": 0.0029759075184092234, "optim/total_tokens": 2134638592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 5.2620930671691895, "created_at": "2025-01-15T00:28:45.770835+00:00"} {"global_step": 8144, "acc_step": 0, "speed/wps": 12972.484523020503, "speed/FLOPS": 203750695495809.4, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.818624258041382, "optim/lr": 0.0029758922214253294, "optim/total_tokens": 2134900736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433993, "loss/out": 5.425753593444824, "created_at": "2025-01-15T00:28:50.834086+00:00"} {"global_step": 8145, "acc_step": 0, "speed/wps": 12972.483621433084, "speed/FLOPS": 203750681335140.1, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7150659561157227, "optim/lr": 0.0029758769196260772, "optim/total_tokens": 2135162880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 5.250207901000977, "created_at": "2025-01-15T00:28:55.893892+00:00"} {"global_step": 8146, "acc_step": 0, "speed/wps": 12979.066175228361, "speed/FLOPS": 203854069387871.72, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2744724750518799, "optim/lr": 0.0029758616130115176, "optim/total_tokens": 2135425024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429232, "loss/out": 5.20415735244751, "created_at": "2025-01-15T00:29:00.947050+00:00"} {"global_step": 8147, "acc_step": 0, "speed/wps": 12965.43741312166, "speed/FLOPS": 203640010951104.28, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.306235671043396, "optim/lr": 0.0029758463015816996, "optim/total_tokens": 2135687168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291186, "loss/out": 5.2292633056640625, "created_at": "2025-01-15T00:29:06.007642+00:00"} {"global_step": 8148, "acc_step": 0, "speed/wps": 12970.622098751353, "speed/FLOPS": 203721443563423.28, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0012, "optim/grad_norm": 3.173766851425171, "optim/lr": 0.002975830985336673, "optim/total_tokens": 2135949312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 391766, "loss/out": 5.272984504699707, "created_at": "2025-01-15T00:29:11.065065+00:00"} {"global_step": 8149, "acc_step": 0, "speed/wps": 12972.104781664062, "speed/FLOPS": 203744731135987.2, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.6939308643341064, "optim/lr": 0.0029758156642764894, "optim/total_tokens": 2136211456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436381, "loss/out": 5.315843105316162, "created_at": "2025-01-15T00:29:16.124713+00:00"} {"global_step": 8150, "acc_step": 0, "speed/wps": 12976.278464922432, "speed/FLOPS": 203810284566801.7, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.8373037576675415, "optim/lr": 0.0029758003384011975, "optim/total_tokens": 2136473600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 5.277371406555176, "created_at": "2025-01-15T00:29:21.180409+00:00"} {"global_step": 8151, "acc_step": 0, "speed/wps": 12970.116298591738, "speed/FLOPS": 203713499276874.25, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.237405300140381, "optim/lr": 0.0029757850077108475, "optim/total_tokens": 2136735744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437723, "loss/out": 5.238243103027344, "created_at": "2025-01-15T00:29:26.240023+00:00"} {"global_step": 8152, "acc_step": 0, "speed/wps": 12977.981194526123, "speed/FLOPS": 203837028275023.3, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.58298659324646, "optim/lr": 0.002975769672205489, "optim/total_tokens": 2136997888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438240, "loss/out": 5.252264976501465, "created_at": "2025-01-15T00:29:31.297315+00:00"} {"global_step": 8153, "acc_step": 0, "speed/wps": 12976.112176406503, "speed/FLOPS": 203807672777154.66, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5401034355163574, "optim/lr": 0.0029757543318851727, "optim/total_tokens": 2137260032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435053, "loss/out": 5.207942485809326, "created_at": "2025-01-15T00:29:36.360587+00:00"} {"global_step": 8154, "acc_step": 0, "speed/wps": 12971.068859941599, "speed/FLOPS": 203728460561830.28, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1672747135162354, "optim/lr": 0.002975738986749948, "optim/total_tokens": 2137522176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 5.198221206665039, "created_at": "2025-01-15T00:29:41.420542+00:00"} {"global_step": 8155, "acc_step": 0, "speed/wps": 12968.789526275426, "speed/FLOPS": 203692660494470.38, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7086358070373535, "optim/lr": 0.0029757236367998663, "optim/total_tokens": 2137784320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453945, "loss/out": 5.115839958190918, "created_at": "2025-01-15T00:29:46.485856+00:00"} {"global_step": 8156, "acc_step": 0, "speed/wps": 12971.597983528094, "speed/FLOPS": 203736771174848.2, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8305743932724, "optim/lr": 0.002975708282034976, "optim/total_tokens": 2138046464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324848, "loss/out": 5.115107536315918, "created_at": "2025-01-15T00:29:51.544922+00:00"} {"global_step": 8157, "acc_step": 0, "speed/wps": 12974.241091514781, "speed/FLOPS": 203778284817790.1, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6912035942077637, "optim/lr": 0.0029756929224553285, "optim/total_tokens": 2138308608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440390, "loss/out": 5.1945061683654785, "created_at": "2025-01-15T00:29:56.605223+00:00"} {"global_step": 8158, "acc_step": 0, "speed/wps": 12976.210890406446, "speed/FLOPS": 203809223216171.03, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1136202812194824, "optim/lr": 0.002975677558060973, "optim/total_tokens": 2138570752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438769, "loss/out": 5.1480631828308105, "created_at": "2025-01-15T00:30:01.663886+00:00"} {"global_step": 8159, "acc_step": 0, "speed/wps": 12974.814796639603, "speed/FLOPS": 203787295645128.88, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0933915376663208, "optim/lr": 0.0029756621888519597, "optim/total_tokens": 2138832896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 5.1677656173706055, "created_at": "2025-01-15T00:30:06.720031+00:00"} {"global_step": 8160, "acc_step": 0, "speed/wps": 12973.770773452576, "speed/FLOPS": 203770897826338.4, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.302527904510498, "optim/lr": 0.0029756468148283395, "optim/total_tokens": 2139095040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 5.176122665405273, "created_at": "2025-01-15T00:30:11.776226+00:00"} {"global_step": 8161, "acc_step": 0, "speed/wps": 12974.569591415868, "speed/FLOPS": 203783444360142.84, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1483886241912842, "optim/lr": 0.0029756314359901622, "optim/total_tokens": 2139357184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436388, "loss/out": 5.105523109436035, "created_at": "2025-01-15T00:30:16.832944+00:00"} {"global_step": 8162, "acc_step": 0, "speed/wps": 12980.019529739508, "speed/FLOPS": 203869043130515.62, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6662929058074951, "optim/lr": 0.002975616052337478, "optim/total_tokens": 2139619328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418803, "loss/out": 5.258249282836914, "created_at": "2025-01-15T00:30:21.886043+00:00"} {"global_step": 8163, "acc_step": 0, "speed/wps": 12981.520177325026, "speed/FLOPS": 203892612862952.1, "speed/curr_iter_time": 1.2723, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2209686040878296, "optim/lr": 0.002975600663870337, "optim/total_tokens": 2139881472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431605, "loss/out": 5.166130542755127, "created_at": "2025-01-15T00:30:26.943875+00:00"} {"global_step": 8164, "acc_step": 0, "speed/wps": 12979.283811265886, "speed/FLOPS": 203857487660904.28, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.9840598106384277, "optim/lr": 0.002975585270588789, "optim/total_tokens": 2140143616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464535, "loss/out": 5.2451934814453125, "created_at": "2025-01-15T00:30:31.999231+00:00"} {"global_step": 8165, "acc_step": 0, "speed/wps": 12975.65819126497, "speed/FLOPS": 203800542316665.94, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0117855072021484, "optim/lr": 0.0029755698724928843, "optim/total_tokens": 2140405760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 5.275249004364014, "created_at": "2025-01-15T00:30:37.057260+00:00"} {"global_step": 8166, "acc_step": 0, "speed/wps": 12977.036082286972, "speed/FLOPS": 203822183988587.16, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.513538122177124, "optim/lr": 0.0029755544695826744, "optim/total_tokens": 2140667904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 5.2935566902160645, "created_at": "2025-01-15T00:30:42.114920+00:00"} {"global_step": 8167, "acc_step": 0, "speed/wps": 12982.084349943474, "speed/FLOPS": 203901473969179.22, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.765978217124939, "optim/lr": 0.002975539061858208, "optim/total_tokens": 2140930048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 5.214587211608887, "created_at": "2025-01-15T00:30:47.172599+00:00"} {"global_step": 8168, "acc_step": 0, "speed/wps": 12972.66370225595, "speed/FLOPS": 203753509751920.16, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.375614881515503, "optim/lr": 0.002975523649319536, "optim/total_tokens": 2141192192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451275, "loss/out": 5.23414421081543, "created_at": "2025-01-15T00:30:52.228586+00:00"} {"global_step": 8169, "acc_step": 0, "speed/wps": 12977.179199451088, "speed/FLOPS": 203824431840313.2, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.705146074295044, "optim/lr": 0.0029755082319667093, "optim/total_tokens": 2141454336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 5.306427955627441, "created_at": "2025-01-15T00:30:57.283358+00:00"} {"global_step": 8170, "acc_step": 0, "speed/wps": 12978.668866024627, "speed/FLOPS": 203847829100867.0, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0009, "optim/grad_norm": 2.717674732208252, "optim/lr": 0.0029754928097997764, "optim/total_tokens": 2141716480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309312, "loss/out": 5.195048809051514, "created_at": "2025-01-15T00:31:02.346224+00:00"} {"global_step": 8171, "acc_step": 0, "speed/wps": 12966.907749867383, "speed/FLOPS": 203663104610150.28, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7795734405517578, "optim/lr": 0.002975477382818789, "optim/total_tokens": 2141978624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 5.122446060180664, "created_at": "2025-01-15T00:31:07.409131+00:00"} {"global_step": 8172, "acc_step": 0, "speed/wps": 12972.809425500012, "speed/FLOPS": 203755798535712.88, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1810109615325928, "optim/lr": 0.0029754619510237983, "optim/total_tokens": 2142240768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453945, "loss/out": 5.186644554138184, "created_at": "2025-01-15T00:31:12.465496+00:00"} {"global_step": 8173, "acc_step": 0, "speed/wps": 12967.754126791935, "speed/FLOPS": 203676398122794.16, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.2017247676849365, "optim/lr": 0.002975446514414853, "optim/total_tokens": 2142502912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 5.211842060089111, "created_at": "2025-01-15T00:31:17.526693+00:00"} {"global_step": 8174, "acc_step": 0, "speed/wps": 12975.418509685984, "speed/FLOPS": 203796777788110.94, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3909406661987305, "optim/lr": 0.0029754310729920037, "optim/total_tokens": 2142765056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447026, "loss/out": 5.31202507019043, "created_at": "2025-01-15T00:31:22.584428+00:00"} {"global_step": 8175, "acc_step": 0, "speed/wps": 12972.965907287047, "speed/FLOPS": 203758256297206.78, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7843290567398071, "optim/lr": 0.0029754156267553012, "optim/total_tokens": 2143027200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497130, "loss/out": 5.209366798400879, "created_at": "2025-01-15T00:31:27.641576+00:00"} {"global_step": 8176, "acc_step": 0, "speed/wps": 12975.926965487162, "speed/FLOPS": 203804763785159.94, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7837235927581787, "optim/lr": 0.002975400175704796, "optim/total_tokens": 2143289344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 5.252703666687012, "created_at": "2025-01-15T00:31:32.696002+00:00"} {"global_step": 8177, "acc_step": 0, "speed/wps": 12974.039567562872, "speed/FLOPS": 203775119607202.44, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0803076028823853, "optim/lr": 0.002975384719840538, "optim/total_tokens": 2143551488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440399, "loss/out": 5.151905536651611, "created_at": "2025-01-15T00:31:37.755581+00:00"} {"global_step": 8178, "acc_step": 0, "speed/wps": 12974.76406731117, "speed/FLOPS": 203786498871316.44, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.042823314666748, "optim/lr": 0.0029753692591625784, "optim/total_tokens": 2143813632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306381, "loss/out": 5.121002197265625, "created_at": "2025-01-15T00:31:42.815463+00:00"} {"global_step": 8179, "acc_step": 0, "speed/wps": 12976.412216203324, "speed/FLOPS": 203812385314461.62, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0671956539154053, "optim/lr": 0.002975353793670967, "optim/total_tokens": 2144075776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312942, "loss/out": 5.278195381164551, "created_at": "2025-01-15T00:31:47.870963+00:00"} {"global_step": 8180, "acc_step": 0, "speed/wps": 12978.101667970855, "speed/FLOPS": 203838920476020.72, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.0300323963165283, "optim/lr": 0.0029753383233657544, "optim/total_tokens": 2144337920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301284, "loss/out": 5.220749378204346, "created_at": "2025-01-15T00:31:52.924286+00:00"} {"global_step": 8181, "acc_step": 0, "speed/wps": 12975.190495489072, "speed/FLOPS": 203793196511824.12, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.3573148250579834, "optim/lr": 0.002975322848246991, "optim/total_tokens": 2144600064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 284929, "loss/out": 5.188977241516113, "created_at": "2025-01-15T00:31:57.979489+00:00"} {"global_step": 8182, "acc_step": 0, "speed/wps": 12983.295535674933, "speed/FLOPS": 203920497305435.4, "speed/curr_iter_time": 1.2722, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.162774920463562, "optim/lr": 0.0029753073683147276, "optim/total_tokens": 2144862208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352401, "loss/out": 5.169160842895508, "created_at": "2025-01-15T00:32:03.035592+00:00"} {"global_step": 8183, "acc_step": 0, "speed/wps": 12975.918468844138, "speed/FLOPS": 203804630333705.66, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3797039985656738, "optim/lr": 0.0029752918835690145, "optim/total_tokens": 2145124352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434505, "loss/out": 5.131771564483643, "created_at": "2025-01-15T00:32:08.094382+00:00"} {"global_step": 8184, "acc_step": 0, "speed/wps": 12978.08495622476, "speed/FLOPS": 203838657995083.94, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.203182220458984, "optim/lr": 0.0029752763940099026, "optim/total_tokens": 2145386496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448361, "loss/out": 5.175548076629639, "created_at": "2025-01-15T00:32:13.148369+00:00"} {"global_step": 8185, "acc_step": 0, "speed/wps": 12979.66350620349, "speed/FLOPS": 203863451291654.88, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.311318039894104, "optim/lr": 0.0029752608996374413, "optim/total_tokens": 2145648640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 5.27497673034668, "created_at": "2025-01-15T00:32:18.201386+00:00"} {"global_step": 8186, "acc_step": 0, "speed/wps": 12970.14170360765, "speed/FLOPS": 203713898297559.28, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.3474199771881104, "optim/lr": 0.0029752454004516827, "optim/total_tokens": 2145910784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 5.096800327301025, "created_at": "2025-01-15T00:32:23.276177+00:00"} {"global_step": 8187, "acc_step": 0, "speed/wps": 12977.540598579717, "speed/FLOPS": 203830108110242.75, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2751243114471436, "optim/lr": 0.002975229896452676, "optim/total_tokens": 2146172928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 5.133386135101318, "created_at": "2025-01-15T00:32:28.329991+00:00"} {"global_step": 8188, "acc_step": 0, "speed/wps": 12980.040264044237, "speed/FLOPS": 203869368791263.34, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.418821096420288, "optim/lr": 0.002975214387640473, "optim/total_tokens": 2146435072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 5.224116802215576, "created_at": "2025-01-15T00:32:33.393096+00:00"} {"global_step": 8189, "acc_step": 0, "speed/wps": 12980.805770182815, "speed/FLOPS": 203881392117085.38, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.579836845397949, "optim/lr": 0.002975198874015123, "optim/total_tokens": 2146697216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431842, "loss/out": 5.1129045486450195, "created_at": "2025-01-15T00:32:38.446702+00:00"} {"global_step": 8190, "acc_step": 0, "speed/wps": 12979.51629064977, "speed/FLOPS": 203861139069088.34, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.389265775680542, "optim/lr": 0.0029751833555766776, "optim/total_tokens": 2146959360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319377, "loss/out": 5.11924409866333, "created_at": "2025-01-15T00:32:43.500753+00:00"} {"global_step": 8191, "acc_step": 0, "speed/wps": 12979.371810010267, "speed/FLOPS": 203858869802107.75, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3522365093231201, "optim/lr": 0.0029751678323251874, "optim/total_tokens": 2147221504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297001, "loss/out": 5.133600234985352, "created_at": "2025-01-15T00:32:48.553463+00:00"} {"global_step": 8192, "acc_step": 0, "speed/wps": 12979.955706326049, "speed/FLOPS": 203868040696104.72, "speed/curr_iter_time": 1.2715, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.436629295349121, "optim/lr": 0.002975152304260702, "optim/total_tokens": 2147483648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315105, "loss/out": 5.131933212280273, "created_at": "2025-01-15T00:32:53.606097+00:00"} {"global_step": 8193, "acc_step": 0, "speed/wps": 12975.461674015542, "speed/FLOPS": 203797455743220.66, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1859617233276367, "optim/lr": 0.002975136771383274, "optim/total_tokens": 2147745792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310665, "loss/out": 5.287703514099121, "created_at": "2025-01-15T00:32:58.666015+00:00"} {"global_step": 8194, "acc_step": 0, "speed/wps": 12974.43409592947, "speed/FLOPS": 203781316217338.62, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.598125696182251, "optim/lr": 0.0029751212336929516, "optim/total_tokens": 2148007936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333311, "loss/out": 5.139394760131836, "created_at": "2025-01-15T00:33:03.721430+00:00"} {"global_step": 8195, "acc_step": 0, "speed/wps": 12977.345618226725, "speed/FLOPS": 203827045675867.94, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.168772578239441, "optim/lr": 0.0029751056911897877, "optim/total_tokens": 2148270080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365740, "loss/out": 5.10965633392334, "created_at": "2025-01-15T00:33:08.776139+00:00"} {"global_step": 8196, "acc_step": 0, "speed/wps": 12975.9683314486, "speed/FLOPS": 203805413494427.1, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.116429328918457, "optim/lr": 0.002975090143873832, "optim/total_tokens": 2148532224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441971, "loss/out": 5.03163480758667, "created_at": "2025-01-15T00:33:13.833792+00:00"} {"global_step": 8197, "acc_step": 0, "speed/wps": 12973.048960472513, "speed/FLOPS": 203759560761611.56, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0014, "optim/grad_norm": 1.1349419355392456, "optim/lr": 0.002975074591745135, "optim/total_tokens": 2148794368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 5.170112133026123, "created_at": "2025-01-15T00:33:18.889366+00:00"} {"global_step": 8198, "acc_step": 0, "speed/wps": 12974.139905356487, "speed/FLOPS": 203776695550129.75, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.058376431465149, "optim/lr": 0.002975059034803749, "optim/total_tokens": 2149056512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 5.1116414070129395, "created_at": "2025-01-15T00:33:23.945051+00:00"} {"global_step": 8199, "acc_step": 0, "speed/wps": 12979.500298268227, "speed/FLOPS": 203860887886760.38, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.454660654067993, "optim/lr": 0.002975043473049722, "optim/total_tokens": 2149318656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469353, "loss/out": 5.228137493133545, "created_at": "2025-01-15T00:33:29.030371+00:00"} {"global_step": 8200, "acc_step": 0, "speed/wps": 12970.82633299146, "speed/FLOPS": 203724651342811.4, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.407779574394226, "optim/lr": 0.0029750279064831063, "optim/total_tokens": 2149580800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 5.157748222351074, "created_at": "2025-01-15T00:33:34.087347+00:00"} {"global_step": 8201, "acc_step": 0, "speed/wps": 12982.267772707924, "speed/FLOPS": 203904354875744.22, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.091030120849609, "optim/lr": 0.0029750123351039537, "optim/total_tokens": 2149842944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323502, "loss/out": 5.231036186218262, "created_at": "2025-01-15T00:33:39.143277+00:00"} {"global_step": 8202, "acc_step": 0, "speed/wps": 12979.390842487193, "speed/FLOPS": 203859168733311.22, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9807984828948975, "optim/lr": 0.002974996758912314, "optim/total_tokens": 2150105088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289017, "loss/out": 5.192925453186035, "created_at": "2025-01-15T00:33:44.200955+00:00"} {"global_step": 8203, "acc_step": 0, "speed/wps": 12977.053408768099, "speed/FLOPS": 203822456124782.34, "speed/curr_iter_time": 1.2721, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0406240224838257, "optim/lr": 0.0029749811779082374, "optim/total_tokens": 2150367232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368718, "loss/out": 5.153895854949951, "created_at": "2025-01-15T00:33:49.259957+00:00"} {"global_step": 8204, "acc_step": 0, "speed/wps": 12973.886334514786, "speed/FLOPS": 203772712871617.97, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5274710655212402, "optim/lr": 0.0029749655920917758, "optim/total_tokens": 2150629376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436916, "loss/out": 5.111319541931152, "created_at": "2025-01-15T00:33:54.319049+00:00"} {"global_step": 8205, "acc_step": 0, "speed/wps": 12973.48250646726, "speed/FLOPS": 203766370197213.7, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8607907295227051, "optim/lr": 0.00297495000146298, "optim/total_tokens": 2150891520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444359, "loss/out": 5.225164413452148, "created_at": "2025-01-15T00:33:59.377607+00:00"} {"global_step": 8206, "acc_step": 0, "speed/wps": 12983.184779834844, "speed/FLOPS": 203918757732772.34, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.271800994873047, "optim/lr": 0.0029749344060218995, "optim/total_tokens": 2151153664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428688, "loss/out": 5.158285140991211, "created_at": "2025-01-15T00:34:04.432937+00:00"} {"global_step": 8207, "acc_step": 0, "speed/wps": 12979.654054199538, "speed/FLOPS": 203863302834944.4, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.205559492111206, "optim/lr": 0.002974918805768587, "optim/total_tokens": 2151415808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454466, "loss/out": 5.091520309448242, "created_at": "2025-01-15T00:34:09.489641+00:00"} {"global_step": 8208, "acc_step": 0, "speed/wps": 12970.465824117855, "speed/FLOPS": 203718989055560.34, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.001, "optim/grad_norm": 1.4309107065200806, "optim/lr": 0.002974903200703092, "optim/total_tokens": 2151677952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 5.220754623413086, "created_at": "2025-01-15T00:34:14.550762+00:00"} {"global_step": 8209, "acc_step": 0, "speed/wps": 12965.890673778335, "speed/FLOPS": 203647130032562.78, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.160233736038208, "optim/lr": 0.002974887590825466, "optim/total_tokens": 2151940096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 5.224231719970703, "created_at": "2025-01-15T00:34:19.617528+00:00"} {"global_step": 8210, "acc_step": 0, "speed/wps": 12969.548583289778, "speed/FLOPS": 203704582527934.4, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1833667755126953, "optim/lr": 0.0029748719761357604, "optim/total_tokens": 2152202240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 5.004088401794434, "created_at": "2025-01-15T00:34:24.676889+00:00"} {"global_step": 8211, "acc_step": 0, "speed/wps": 12972.321359262209, "speed/FLOPS": 203748132784774.88, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1112333536148071, "optim/lr": 0.0029748563566340254, "optim/total_tokens": 2152464384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448890, "loss/out": 5.075312614440918, "created_at": "2025-01-15T00:34:29.732387+00:00"} {"global_step": 8212, "acc_step": 0, "speed/wps": 12975.663287609344, "speed/FLOPS": 203800622361757.44, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.83693528175354, "optim/lr": 0.0029748407323203123, "optim/total_tokens": 2152726528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299406, "loss/out": 5.021744251251221, "created_at": "2025-01-15T00:34:34.794895+00:00"} {"global_step": 8213, "acc_step": 0, "speed/wps": 12975.568386198895, "speed/FLOPS": 203799131804698.12, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2488943338394165, "optim/lr": 0.002974825103194672, "optim/total_tokens": 2152988672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 409734, "loss/out": 5.100748538970947, "created_at": "2025-01-15T00:34:39.853363+00:00"} {"global_step": 8214, "acc_step": 0, "speed/wps": 12973.592300335373, "speed/FLOPS": 203768094660784.97, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 55.0800895690918, "optim/lr": 0.0029748094692571554, "optim/total_tokens": 2153250816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455808, "loss/out": 5.262358665466309, "created_at": "2025-01-15T00:34:44.909480+00:00"} {"global_step": 8215, "acc_step": 0, "speed/wps": 12969.282128890633, "speed/FLOPS": 203700397495449.1, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.945265531539917, "optim/lr": 0.0029747938305078134, "optim/total_tokens": 2153512960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 5.268942356109619, "created_at": "2025-01-15T00:34:49.966204+00:00"} {"global_step": 8216, "acc_step": 0, "speed/wps": 12970.465903012095, "speed/FLOPS": 203718990294702.78, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4214030504226685, "optim/lr": 0.0029747781869466977, "optim/total_tokens": 2153775104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454466, "loss/out": 5.1840620040893555, "created_at": "2025-01-15T00:34:55.026611+00:00"} {"global_step": 8217, "acc_step": 0, "speed/wps": 12967.19978455644, "speed/FLOPS": 203667691416239.66, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 15.141417503356934, "optim/lr": 0.0029747625385738584, "optim/total_tokens": 2154037248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 5.169233322143555, "created_at": "2025-01-15T00:35:00.088606+00:00"} {"global_step": 8218, "acc_step": 0, "speed/wps": 12972.028730364691, "speed/FLOPS": 203743536645825.12, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.446316957473755, "optim/lr": 0.0029747468853893472, "optim/total_tokens": 2154299392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 5.483384609222412, "created_at": "2025-01-15T00:35:05.144207+00:00"} {"global_step": 8219, "acc_step": 0, "speed/wps": 12973.648677361225, "speed/FLOPS": 203768980139446.16, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.471623659133911, "optim/lr": 0.002974731227393215, "optim/total_tokens": 2154561536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 5.263784885406494, "created_at": "2025-01-15T00:35:10.203963+00:00"} {"global_step": 8220, "acc_step": 0, "speed/wps": 12972.441268489005, "speed/FLOPS": 203750016123954.28, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4808154106140137, "optim/lr": 0.0029747155645855135, "optim/total_tokens": 2154823680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500794, "loss/out": 5.272342205047607, "created_at": "2025-01-15T00:35:15.264892+00:00"} {"global_step": 8221, "acc_step": 0, "speed/wps": 12971.793326839519, "speed/FLOPS": 203739839309983.97, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2251373529434204, "optim/lr": 0.0029746998969662924, "optim/total_tokens": 2155085824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455791, "loss/out": 5.128861427307129, "created_at": "2025-01-15T00:35:20.320664+00:00"} {"global_step": 8222, "acc_step": 0, "speed/wps": 12968.365311191017, "speed/FLOPS": 203685997613637.1, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7931941747665405, "optim/lr": 0.0029746842245356032, "optim/total_tokens": 2155347968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322837, "loss/out": 5.159679412841797, "created_at": "2025-01-15T00:35:25.384522+00:00"} {"global_step": 8223, "acc_step": 0, "speed/wps": 12968.043420219741, "speed/FLOPS": 203680941873608.66, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.253974437713623, "optim/lr": 0.0029746685472934985, "optim/total_tokens": 2155610112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353744, "loss/out": 5.310544967651367, "created_at": "2025-01-15T00:35:30.445591+00:00"} {"global_step": 8224, "acc_step": 0, "speed/wps": 12967.657359976842, "speed/FLOPS": 203674878266991.56, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0966824293136597, "optim/lr": 0.0029746528652400276, "optim/total_tokens": 2155872256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416992, "loss/out": 5.191306114196777, "created_at": "2025-01-15T00:35:35.508056+00:00"} {"global_step": 8225, "acc_step": 0, "speed/wps": 12956.423777371345, "speed/FLOPS": 203498439415612.1, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.001, "optim/grad_norm": 3.2662160396575928, "optim/lr": 0.002974637178375243, "optim/total_tokens": 2156134400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 5.123452186584473, "created_at": "2025-01-15T00:35:40.570898+00:00"} {"global_step": 8226, "acc_step": 0, "speed/wps": 12973.395972422795, "speed/FLOPS": 203765011061135.12, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.018271446228027, "optim/lr": 0.002974621486699195, "optim/total_tokens": 2156396544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435567, "loss/out": 5.303491592407227, "created_at": "2025-01-15T00:35:45.629608+00:00"} {"global_step": 8227, "acc_step": 0, "speed/wps": 12968.277201126477, "speed/FLOPS": 203684613723997.6, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3718093633651733, "optim/lr": 0.002974605790211935, "optim/total_tokens": 2156658688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 5.172492980957031, "created_at": "2025-01-15T00:35:50.686810+00:00"} {"global_step": 8228, "acc_step": 0, "speed/wps": 12973.839169316961, "speed/FLOPS": 203771972077374.38, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.3912097215652466, "optim/lr": 0.002974590088913515, "optim/total_tokens": 2156920832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 5.0450639724731445, "created_at": "2025-01-15T00:35:55.747535+00:00"} {"global_step": 8229, "acc_step": 0, "speed/wps": 12973.159902206962, "speed/FLOPS": 203761303254001.0, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8457969427108765, "optim/lr": 0.0029745743828039854, "optim/total_tokens": 2157182976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434003, "loss/out": 5.339552402496338, "created_at": "2025-01-15T00:36:00.806096+00:00"} {"global_step": 8230, "acc_step": 0, "speed/wps": 12967.182713772028, "speed/FLOPS": 203667423296112.75, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.0872926712036133, "optim/lr": 0.0029745586718833974, "optim/total_tokens": 2157445120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332493, "loss/out": 5.263747692108154, "created_at": "2025-01-15T00:36:05.863560+00:00"} {"global_step": 8231, "acc_step": 0, "speed/wps": 12979.178917210671, "speed/FLOPS": 203855840155627.6, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.9985969066619873, "optim/lr": 0.0029745429561518024, "optim/total_tokens": 2157707264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351583, "loss/out": 5.121973514556885, "created_at": "2025-01-15T00:36:10.919387+00:00"} {"global_step": 8232, "acc_step": 0, "speed/wps": 12976.372776098953, "speed/FLOPS": 203811765853426.1, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4823497533798218, "optim/lr": 0.002974527235609252, "optim/total_tokens": 2157969408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 382664, "loss/out": 5.117306709289551, "created_at": "2025-01-15T00:36:15.976021+00:00"} {"global_step": 8233, "acc_step": 0, "speed/wps": 12972.667870719859, "speed/FLOPS": 203753575223373.94, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.4387739896774292, "optim/lr": 0.002974511510255797, "optim/total_tokens": 2158231552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415046, "loss/out": 5.20937442779541, "created_at": "2025-01-15T00:36:21.033605+00:00"} {"global_step": 8234, "acc_step": 0, "speed/wps": 12969.625573633006, "speed/FLOPS": 203705791767070.62, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0534766912460327, "optim/lr": 0.002974495780091489, "optim/total_tokens": 2158493696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437717, "loss/out": 5.037546634674072, "created_at": "2025-01-15T00:36:26.097988+00:00"} {"global_step": 8235, "acc_step": 0, "speed/wps": 12977.067674431833, "speed/FLOPS": 203822680186634.34, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.488437294960022, "optim/lr": 0.002974480045116379, "optim/total_tokens": 2158755840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 5.131009578704834, "created_at": "2025-01-15T00:36:31.152895+00:00"} {"global_step": 8236, "acc_step": 0, "speed/wps": 12970.103922438142, "speed/FLOPS": 203713304892375.4, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0009, "optim/grad_norm": 7.115095138549805, "optim/lr": 0.0029744643053305187, "optim/total_tokens": 2159017984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439302, "loss/out": 5.224987983703613, "created_at": "2025-01-15T00:36:36.212394+00:00"} {"global_step": 8237, "acc_step": 0, "speed/wps": 12973.281706869455, "speed/FLOPS": 203763216363602.6, "speed/curr_iter_time": 1.2723, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.328566074371338, "optim/lr": 0.00297444856073396, "optim/total_tokens": 2159280128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 5.1995344161987305, "created_at": "2025-01-15T00:36:41.273633+00:00"} {"global_step": 8238, "acc_step": 0, "speed/wps": 12967.667853154879, "speed/FLOPS": 203675043076771.78, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.148396372795105, "optim/lr": 0.002974432811326753, "optim/total_tokens": 2159542272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456045, "loss/out": 5.105262279510498, "created_at": "2025-01-15T00:36:46.336462+00:00"} {"global_step": 8239, "acc_step": 0, "speed/wps": 12966.867856183528, "speed/FLOPS": 203662478025025.66, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1201395988464355, "optim/lr": 0.0029744170571089495, "optim/total_tokens": 2159804416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317059, "loss/out": 5.143786430358887, "created_at": "2025-01-15T00:36:51.394002+00:00"} {"global_step": 8240, "acc_step": 0, "speed/wps": 12966.875848759246, "speed/FLOPS": 203662603559410.28, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2520020008087158, "optim/lr": 0.0029744012980806016, "optim/total_tokens": 2160066560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356143, "loss/out": 5.084406852722168, "created_at": "2025-01-15T00:36:56.452044+00:00"} {"global_step": 8241, "acc_step": 0, "speed/wps": 12969.982947198914, "speed/FLOPS": 203711404810005.94, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1141715049743652, "optim/lr": 0.0029743855342417606, "optim/total_tokens": 2160328704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436679, "loss/out": 5.045605182647705, "created_at": "2025-01-15T00:37:01.511097+00:00"} {"global_step": 8242, "acc_step": 0, "speed/wps": 12971.593354259725, "speed/FLOPS": 203736698465827.2, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.023991584777832, "optim/lr": 0.0029743697655924767, "optim/total_tokens": 2160590848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474937, "loss/out": 5.082877159118652, "created_at": "2025-01-15T00:37:06.569167+00:00"} {"global_step": 8243, "acc_step": 0, "speed/wps": 12977.836073106484, "speed/FLOPS": 203834748943710.2, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.411174774169922, "optim/lr": 0.002974353992132803, "optim/total_tokens": 2160852992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458410, "loss/out": 5.1727705001831055, "created_at": "2025-01-15T00:37:11.624231+00:00"} {"global_step": 8244, "acc_step": 0, "speed/wps": 12971.410825989304, "speed/FLOPS": 203733831608520.1, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1141289472579956, "optim/lr": 0.0029743382138627894, "optim/total_tokens": 2161115136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440921, "loss/out": 5.1899285316467285, "created_at": "2025-01-15T00:37:16.681360+00:00"} {"global_step": 8245, "acc_step": 0, "speed/wps": 12977.40826001338, "speed/FLOPS": 203828029551206.2, "speed/curr_iter_time": 1.2722, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7088934183120728, "optim/lr": 0.002974322430782489, "optim/total_tokens": 2161377280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448890, "loss/out": 5.014966011047363, "created_at": "2025-01-15T00:37:21.738786+00:00"} {"global_step": 8246, "acc_step": 0, "speed/wps": 12972.75833159048, "speed/FLOPS": 203754996035653.28, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1532081365585327, "optim/lr": 0.0029743066428919514, "optim/total_tokens": 2161639424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445407, "loss/out": 5.292307376861572, "created_at": "2025-01-15T00:37:26.799112+00:00"} {"global_step": 8247, "acc_step": 0, "speed/wps": 12972.396763519384, "speed/FLOPS": 203749317112250.12, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.55810809135437, "optim/lr": 0.00297429085019123, "optim/total_tokens": 2161901568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 5.045647144317627, "created_at": "2025-01-15T00:37:31.863332+00:00"} {"global_step": 8248, "acc_step": 0, "speed/wps": 12973.840453809387, "speed/FLOPS": 203771992252093.0, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.14772367477417, "optim/lr": 0.0029742750526803754, "optim/total_tokens": 2162163712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453646, "loss/out": 5.183857440948486, "created_at": "2025-01-15T00:37:36.918306+00:00"} {"global_step": 8249, "acc_step": 0, "speed/wps": 12969.272049011173, "speed/FLOPS": 203700239177091.0, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.519921898841858, "optim/lr": 0.002974259250359439, "optim/total_tokens": 2162425856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321540, "loss/out": 5.159011363983154, "created_at": "2025-01-15T00:37:41.975886+00:00"} {"global_step": 8250, "acc_step": 0, "speed/wps": 12966.404086229111, "speed/FLOPS": 203655193880604.88, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5109179019927979, "optim/lr": 0.0029742434432284727, "optim/total_tokens": 2162688000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319225, "loss/out": 5.1495890617370605, "created_at": "2025-01-15T00:37:47.034252+00:00"} {"global_step": 8251, "acc_step": 0, "speed/wps": 12973.101522787138, "speed/FLOPS": 203760386324990.94, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.206573486328125, "optim/lr": 0.0029742276312875278, "optim/total_tokens": 2162950144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 405621, "loss/out": 5.095584869384766, "created_at": "2025-01-15T00:37:52.094783+00:00"} {"global_step": 8252, "acc_step": 0, "speed/wps": 12971.504530812826, "speed/FLOPS": 203735303371537.44, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0978598594665527, "optim/lr": 0.002974211814536656, "optim/total_tokens": 2163212288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448890, "loss/out": 5.111080169677734, "created_at": "2025-01-15T00:37:57.151680+00:00"} {"global_step": 8253, "acc_step": 0, "speed/wps": 12972.187948175178, "speed/FLOPS": 203746037380327.03, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3168463706970215, "optim/lr": 0.0029741959929759096, "optim/total_tokens": 2163474432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 5.0695881843566895, "created_at": "2025-01-15T00:38:02.209616+00:00"} {"global_step": 8254, "acc_step": 0, "speed/wps": 12969.458064391905, "speed/FLOPS": 203703160804257.3, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3450055122375488, "optim/lr": 0.002974180166605339, "optim/total_tokens": 2163736576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459528, "loss/out": 5.043270111083984, "created_at": "2025-01-15T00:38:07.271701+00:00"} {"global_step": 8255, "acc_step": 0, "speed/wps": 12970.825998270035, "speed/FLOPS": 203724646085551.47, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.385352373123169, "optim/lr": 0.0029741643354249964, "optim/total_tokens": 2163998720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468018, "loss/out": 4.983375072479248, "created_at": "2025-01-15T00:38:12.331984+00:00"} {"global_step": 8256, "acc_step": 0, "speed/wps": 12965.413404910443, "speed/FLOPS": 203639633869156.38, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5743367671966553, "optim/lr": 0.0029741484994349338, "optim/total_tokens": 2164260864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460840, "loss/out": 5.1319661140441895, "created_at": "2025-01-15T00:38:17.401112+00:00"} {"global_step": 8257, "acc_step": 0, "speed/wps": 12971.060673310685, "speed/FLOPS": 203728331979542.2, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.646623134613037, "optim/lr": 0.0029741326586352024, "optim/total_tokens": 2164523008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295368, "loss/out": 5.198981285095215, "created_at": "2025-01-15T00:38:22.458734+00:00"} {"global_step": 8258, "acc_step": 0, "speed/wps": 12973.256827697985, "speed/FLOPS": 203762825602027.1, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5290437936782837, "optim/lr": 0.0029741168130258543, "optim/total_tokens": 2164785152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327253, "loss/out": 5.160698413848877, "created_at": "2025-01-15T00:38:27.518472+00:00"} {"global_step": 8259, "acc_step": 0, "speed/wps": 12971.218707717233, "speed/FLOPS": 203730814127059.03, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2756694555282593, "optim/lr": 0.0029741009626069403, "optim/total_tokens": 2165047296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332731, "loss/out": 5.001773357391357, "created_at": "2025-01-15T00:38:32.575427+00:00"} {"global_step": 8260, "acc_step": 0, "speed/wps": 12976.286232208284, "speed/FLOPS": 203810406562699.38, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.193135380744934, "optim/lr": 0.002974085107378513, "optim/total_tokens": 2165309440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322156, "loss/out": 5.13065767288208, "created_at": "2025-01-15T00:38:37.634671+00:00"} {"global_step": 8261, "acc_step": 0, "speed/wps": 12968.71418067828, "speed/FLOPS": 203691477088331.47, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1108325719833374, "optim/lr": 0.002974069247340624, "optim/total_tokens": 2165571584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 5.135430335998535, "created_at": "2025-01-15T00:38:42.692034+00:00"} {"global_step": 8262, "acc_step": 0, "speed/wps": 12975.307555636477, "speed/FLOPS": 203795035102296.6, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.138725757598877, "optim/lr": 0.0029740533824933247, "optim/total_tokens": 2165833728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431584, "loss/out": 5.090594291687012, "created_at": "2025-01-15T00:38:47.747301+00:00"} {"global_step": 8263, "acc_step": 0, "speed/wps": 12973.375182801958, "speed/FLOPS": 203764684531571.94, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3851118087768555, "optim/lr": 0.0029740375128366674, "optim/total_tokens": 2166095872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 5.12732458114624, "created_at": "2025-01-15T00:38:52.805826+00:00"} {"global_step": 8264, "acc_step": 0, "speed/wps": 12971.782834157582, "speed/FLOPS": 203739674507995.7, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 41.9465446472168, "optim/lr": 0.002974021638370703, "optim/total_tokens": 2166358016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459507, "loss/out": 5.145728588104248, "created_at": "2025-01-15T00:38:57.861684+00:00"} {"global_step": 8265, "acc_step": 0, "speed/wps": 12973.085003293314, "speed/FLOPS": 203760126863640.25, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1268298625946045, "optim/lr": 0.002974005759095484, "optim/total_tokens": 2166620160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 5.133123397827148, "created_at": "2025-01-15T00:39:02.921082+00:00"} {"global_step": 8266, "acc_step": 0, "speed/wps": 12971.203838030386, "speed/FLOPS": 203730580578194.0, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1698098182678223, "optim/lr": 0.0029739898750110625, "optim/total_tokens": 2166882304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 5.049654483795166, "created_at": "2025-01-15T00:39:07.977332+00:00"} {"global_step": 8267, "acc_step": 0, "speed/wps": 12969.870159334272, "speed/FLOPS": 203709633321604.97, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.569333553314209, "optim/lr": 0.0029739739861174897, "optim/total_tokens": 2167144448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 5.178476333618164, "created_at": "2025-01-15T00:39:13.034156+00:00"} {"global_step": 8268, "acc_step": 0, "speed/wps": 12969.384496607143, "speed/FLOPS": 203702005321104.88, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5210264921188354, "optim/lr": 0.0029739580924148176, "optim/total_tokens": 2167406592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310665, "loss/out": 5.052375793457031, "created_at": "2025-01-15T00:39:18.092555+00:00"} {"global_step": 8269, "acc_step": 0, "speed/wps": 12965.112633752531, "speed/FLOPS": 203634909844818.5, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.6915442943573, "optim/lr": 0.0029739421939030983, "optim/total_tokens": 2167668736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351344, "loss/out": 5.184266090393066, "created_at": "2025-01-15T00:39:23.154224+00:00"} {"global_step": 8270, "acc_step": 0, "speed/wps": 12965.706401943153, "speed/FLOPS": 203644235790175.22, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4445399045944214, "optim/lr": 0.002973926290582383, "optim/total_tokens": 2167930880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433464, "loss/out": 5.134713172912598, "created_at": "2025-01-15T00:39:28.212436+00:00"} {"global_step": 8271, "acc_step": 0, "speed/wps": 12965.245698687779, "speed/FLOPS": 203636999812476.94, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0012, "optim/grad_norm": 11.861339569091797, "optim/lr": 0.002973910382452724, "optim/total_tokens": 2168193024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469590, "loss/out": 5.0329132080078125, "created_at": "2025-01-15T00:39:33.271699+00:00"} {"global_step": 8272, "acc_step": 0, "speed/wps": 12976.314214614627, "speed/FLOPS": 203810846064842.84, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.6406636238098145, "optim/lr": 0.002973894469514173, "optim/total_tokens": 2168455168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445166, "loss/out": 5.190802574157715, "created_at": "2025-01-15T00:39:38.332403+00:00"} {"global_step": 8273, "acc_step": 0, "speed/wps": 12965.844488863224, "speed/FLOPS": 203646404635006.25, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.030886173248291, "optim/lr": 0.002973878551766783, "optim/total_tokens": 2168717312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457095, "loss/out": 5.2660722732543945, "created_at": "2025-01-15T00:39:43.395515+00:00"} {"global_step": 8274, "acc_step": 0, "speed/wps": 12967.342221263343, "speed/FLOPS": 203669928580450.1, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.7263388633728027, "optim/lr": 0.0029738626292106045, "optim/total_tokens": 2168979456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 5.093420028686523, "created_at": "2025-01-15T00:39:48.459004+00:00"} {"global_step": 8275, "acc_step": 0, "speed/wps": 12974.989597561087, "speed/FLOPS": 203790041133802.44, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.2641353607177734, "optim/lr": 0.00297384670184569, "optim/total_tokens": 2169241600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433464, "loss/out": 5.143718719482422, "created_at": "2025-01-15T00:39:53.517830+00:00"} {"global_step": 8276, "acc_step": 0, "speed/wps": 12963.61744390506, "speed/FLOPS": 203611425833657.78, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9446377754211426, "optim/lr": 0.002973830769672092, "optim/total_tokens": 2169503744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 5.147905349731445, "created_at": "2025-01-15T00:39:58.577794+00:00"} {"global_step": 8277, "acc_step": 0, "speed/wps": 12968.35762269357, "speed/FLOPS": 203685876855219.6, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.415879011154175, "optim/lr": 0.0029738148326898615, "optim/total_tokens": 2169765888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452840, "loss/out": 5.007928848266602, "created_at": "2025-01-15T00:40:03.636221+00:00"} {"global_step": 8278, "acc_step": 0, "speed/wps": 12970.411738757028, "speed/FLOPS": 203718139570647.47, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6000840663909912, "optim/lr": 0.0029737988908990513, "optim/total_tokens": 2170028032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 339699, "loss/out": 5.116457939147949, "created_at": "2025-01-15T00:40:08.693064+00:00"} {"global_step": 8279, "acc_step": 0, "speed/wps": 12958.713320697916, "speed/FLOPS": 203534399839718.12, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.98900842666626, "optim/lr": 0.002973782944299713, "optim/total_tokens": 2170290176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354800, "loss/out": 5.09903621673584, "created_at": "2025-01-15T00:40:13.758270+00:00"} {"global_step": 8280, "acc_step": 0, "speed/wps": 12971.317990470197, "speed/FLOPS": 203732373499123.16, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0987623929977417, "optim/lr": 0.0029737669928918986, "optim/total_tokens": 2170552320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438240, "loss/out": 5.10397481918335, "created_at": "2025-01-15T00:40:18.816783+00:00"} {"global_step": 8281, "acc_step": 0, "speed/wps": 12968.755808988573, "speed/FLOPS": 203692130918147.94, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.318063259124756, "optim/lr": 0.0029737510366756605, "optim/total_tokens": 2170814464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 5.035317420959473, "created_at": "2025-01-15T00:40:23.874922+00:00"} {"global_step": 8282, "acc_step": 0, "speed/wps": 12965.848700726547, "speed/FLOPS": 203646470788107.44, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.5170769691467285, "optim/lr": 0.0029737350756510504, "optim/total_tokens": 2171076608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448077, "loss/out": 5.033061981201172, "created_at": "2025-01-15T00:40:28.936227+00:00"} {"global_step": 8283, "acc_step": 0, "speed/wps": 12971.597907011233, "speed/FLOPS": 203736769973045.75, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8570351600646973, "optim/lr": 0.0029737191098181204, "optim/total_tokens": 2171338752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435852, "loss/out": 5.155329704284668, "created_at": "2025-01-15T00:40:33.992224+00:00"} {"global_step": 8284, "acc_step": 0, "speed/wps": 12964.118349822269, "speed/FLOPS": 203619293249404.44, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.570298433303833, "optim/lr": 0.0029737031391769224, "optim/total_tokens": 2171600896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 5.066647529602051, "created_at": "2025-01-15T00:40:39.056639+00:00"} {"global_step": 8285, "acc_step": 0, "speed/wps": 12959.02442694821, "speed/FLOPS": 203539286190883.78, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.512825846672058, "optim/lr": 0.0029736871637275094, "optim/total_tokens": 2171863040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459237, "loss/out": 5.124750137329102, "created_at": "2025-01-15T00:40:44.117604+00:00"} {"global_step": 8286, "acc_step": 0, "speed/wps": 12963.48009937146, "speed/FLOPS": 203609268649026.44, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.5444753170013428, "optim/lr": 0.0029736711834699326, "optim/total_tokens": 2172125184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 5.215114593505859, "created_at": "2025-01-15T00:40:49.184433+00:00"} {"global_step": 8287, "acc_step": 0, "speed/wps": 12968.80556158735, "speed/FLOPS": 203692912351078.97, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.794132232666016, "optim/lr": 0.002973655198404245, "optim/total_tokens": 2172387328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324468, "loss/out": 5.163681983947754, "created_at": "2025-01-15T00:40:54.241904+00:00"} {"global_step": 8288, "acc_step": 0, "speed/wps": 12972.614874269753, "speed/FLOPS": 203752742841300.25, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 17.247299194335938, "optim/lr": 0.0029736392085304977, "optim/total_tokens": 2172649472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299938, "loss/out": 5.289878845214844, "created_at": "2025-01-15T00:40:59.297807+00:00"} {"global_step": 8289, "acc_step": 0, "speed/wps": 12968.902988491987, "speed/FLOPS": 203694442574494.88, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.2870891094207764, "optim/lr": 0.0029736232138487433, "optim/total_tokens": 2172911616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360416, "loss/out": 5.157052516937256, "created_at": "2025-01-15T00:41:04.355929+00:00"} {"global_step": 8290, "acc_step": 0, "speed/wps": 12969.638228553966, "speed/FLOPS": 203705990529993.97, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.249395370483398, "optim/lr": 0.002973607214359034, "optim/total_tokens": 2173173760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456868, "loss/out": 5.114521026611328, "created_at": "2025-01-15T00:41:09.416468+00:00"} {"global_step": 8291, "acc_step": 0, "speed/wps": 12957.147081611403, "speed/FLOPS": 203509799902626.4, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.774395227432251, "optim/lr": 0.0029735912100614225, "optim/total_tokens": 2173435904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 5.082425117492676, "created_at": "2025-01-15T00:41:14.479425+00:00"} {"global_step": 8292, "acc_step": 0, "speed/wps": 12971.900442598137, "speed/FLOPS": 203741521710165.53, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8058125972747803, "optim/lr": 0.00297357520095596, "optim/total_tokens": 2173698048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 5.108051300048828, "created_at": "2025-01-15T00:41:19.540489+00:00"} {"global_step": 8293, "acc_step": 0, "speed/wps": 12965.798872862528, "speed/FLOPS": 203645688173032.34, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.222973108291626, "optim/lr": 0.0029735591870427005, "optim/total_tokens": 2173960192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450756, "loss/out": 5.055779457092285, "created_at": "2025-01-15T00:41:24.603920+00:00"} {"global_step": 8294, "acc_step": 0, "speed/wps": 12969.940875931825, "speed/FLOPS": 203710744022944.3, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.987086296081543, "optim/lr": 0.0029735431683216938, "optim/total_tokens": 2174222336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 5.151203155517578, "created_at": "2025-01-15T00:41:29.661695+00:00"} {"global_step": 8295, "acc_step": 0, "speed/wps": 12967.973316986061, "speed/FLOPS": 203679840805991.78, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7862374782562256, "optim/lr": 0.002973527144792994, "optim/total_tokens": 2174484480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427880, "loss/out": 5.158045291900635, "created_at": "2025-01-15T00:41:34.724742+00:00"} {"global_step": 8296, "acc_step": 0, "speed/wps": 12972.660851535187, "speed/FLOPS": 203753464977433.28, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5176079273223877, "optim/lr": 0.002973511116456653, "optim/total_tokens": 2174746624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322689, "loss/out": 5.220224380493164, "created_at": "2025-01-15T00:41:39.780233+00:00"} {"global_step": 8297, "acc_step": 0, "speed/wps": 12967.616642018636, "speed/FLOPS": 203674238735505.62, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.67643928527832, "optim/lr": 0.0029734950833127227, "optim/total_tokens": 2175008768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 383400, "loss/out": 5.028343200683594, "created_at": "2025-01-15T00:41:44.838403+00:00"} {"global_step": 8298, "acc_step": 0, "speed/wps": 12965.45395614481, "speed/FLOPS": 203640270782015.34, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4558008909225464, "optim/lr": 0.0029734790453612553, "optim/total_tokens": 2175270912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440399, "loss/out": 5.163733005523682, "created_at": "2025-01-15T00:41:49.903916+00:00"} {"global_step": 8299, "acc_step": 0, "speed/wps": 12970.617286088644, "speed/FLOPS": 203721367973942.12, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6310055255889893, "optim/lr": 0.002973463002602304, "optim/total_tokens": 2175533056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430506, "loss/out": 5.146564483642578, "created_at": "2025-01-15T00:41:54.960239+00:00"} {"global_step": 8300, "acc_step": 0, "speed/wps": 12967.967916004754, "speed/FLOPS": 203679755976158.62, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.346890926361084, "optim/lr": 0.00297344695503592, "optim/total_tokens": 2175795200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449696, "loss/out": 5.1374664306640625, "created_at": "2025-01-15T00:42:00.019708+00:00"} {"global_step": 8301, "acc_step": 0, "speed/wps": 12968.817211054808, "speed/FLOPS": 203693095321973.4, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1904027462005615, "optim/lr": 0.0029734309026621573, "optim/total_tokens": 2176057344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 5.067935943603516, "created_at": "2025-01-15T00:42:05.076553+00:00"} {"global_step": 8302, "acc_step": 0, "speed/wps": 12970.265378473046, "speed/FLOPS": 203715840781267.44, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3206186294555664, "optim/lr": 0.002973414845481066, "optim/total_tokens": 2176319488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453375, "loss/out": 5.061861515045166, "created_at": "2025-01-15T00:42:10.134455+00:00"} {"global_step": 8303, "acc_step": 0, "speed/wps": 12971.185264718017, "speed/FLOPS": 203730288858800.84, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8846687078475952, "optim/lr": 0.0029733987834927008, "optim/total_tokens": 2176581632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433179, "loss/out": 5.143992900848389, "created_at": "2025-01-15T00:42:15.194586+00:00"} {"global_step": 8304, "acc_step": 0, "speed/wps": 12967.980051490294, "speed/FLOPS": 203679946580634.94, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.122109889984131, "optim/lr": 0.002973382716697112, "optim/total_tokens": 2176843776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 5.137784004211426, "created_at": "2025-01-15T00:42:20.252639+00:00"} {"global_step": 8305, "acc_step": 0, "speed/wps": 12965.397342152903, "speed/FLOPS": 203639381581476.66, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.679161548614502, "optim/lr": 0.0029733666450943537, "optim/total_tokens": 2177105920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444878, "loss/out": 5.128037452697754, "created_at": "2025-01-15T00:42:25.325090+00:00"} {"global_step": 8306, "acc_step": 0, "speed/wps": 12970.072548265474, "speed/FLOPS": 203712812117881.03, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7354350090026855, "optim/lr": 0.0029733505686844776, "optim/total_tokens": 2177368064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302697, "loss/out": 5.072117805480957, "created_at": "2025-01-15T00:42:30.384480+00:00"} {"global_step": 8307, "acc_step": 0, "speed/wps": 12968.739532399764, "speed/FLOPS": 203691875271954.3, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6502790451049805, "optim/lr": 0.0029733344874675357, "optim/total_tokens": 2177630208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 403224, "loss/out": 5.159526348114014, "created_at": "2025-01-15T00:42:35.445275+00:00"} {"global_step": 8308, "acc_step": 0, "speed/wps": 12971.109958228246, "speed/FLOPS": 203729106066897.03, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.884125232696533, "optim/lr": 0.0029733184014435815, "optim/total_tokens": 2177892352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433179, "loss/out": 5.1604766845703125, "created_at": "2025-01-15T00:42:40.506403+00:00"} {"global_step": 8309, "acc_step": 0, "speed/wps": 12971.101934131342, "speed/FLOPS": 203728980037428.9, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4363471269607544, "optim/lr": 0.002973302310612667, "optim/total_tokens": 2178154496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470925, "loss/out": 5.190073013305664, "created_at": "2025-01-15T00:42:45.562462+00:00"} {"global_step": 8310, "acc_step": 0, "speed/wps": 12973.276172309888, "speed/FLOPS": 203763129435739.3, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.825051188468933, "optim/lr": 0.0029732862149748442, "optim/total_tokens": 2178416640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 5.063272476196289, "created_at": "2025-01-15T00:42:50.647187+00:00"} {"global_step": 8311, "acc_step": 0, "speed/wps": 12967.469236019786, "speed/FLOPS": 203671923521736.6, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 17.456920623779297, "optim/lr": 0.0029732701145301667, "optim/total_tokens": 2178678784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 5.111993312835693, "created_at": "2025-01-15T00:42:55.709282+00:00"} {"global_step": 8312, "acc_step": 0, "speed/wps": 12968.214955757321, "speed/FLOPS": 203683636074940.9, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3693853616714478, "optim/lr": 0.002973254009278686, "optim/total_tokens": 2178940928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456851, "loss/out": 5.157063961029053, "created_at": "2025-01-15T00:43:00.766505+00:00"} {"global_step": 8313, "acc_step": 0, "speed/wps": 12960.076835834829, "speed/FLOPS": 203555815718609.5, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1609903573989868, "optim/lr": 0.002973237899220455, "optim/total_tokens": 2179203072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447555, "loss/out": 4.983604907989502, "created_at": "2025-01-15T00:43:05.826898+00:00"} {"global_step": 8314, "acc_step": 0, "speed/wps": 12957.229493377043, "speed/FLOPS": 203511094292651.4, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0422903299331665, "optim/lr": 0.002973221784355527, "optim/total_tokens": 2179465216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312537, "loss/out": 5.086115837097168, "created_at": "2025-01-15T00:43:10.889239+00:00"} {"global_step": 8315, "acc_step": 0, "speed/wps": 12956.319586456062, "speed/FLOPS": 203496802954114.28, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.244586944580078, "optim/lr": 0.0029732056646839537, "optim/total_tokens": 2179727360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 374134, "loss/out": 5.141722679138184, "created_at": "2025-01-15T00:43:15.953146+00:00"} {"global_step": 8316, "acc_step": 0, "speed/wps": 12958.68690803254, "speed/FLOPS": 203533984992513.6, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3776240348815918, "optim/lr": 0.0029731895402057874, "optim/total_tokens": 2179989504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 5.212372779846191, "created_at": "2025-01-15T00:43:21.028650+00:00"} {"global_step": 8317, "acc_step": 0, "speed/wps": 12955.15507192783, "speed/FLOPS": 203478512653237.72, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.3259127140045166, "optim/lr": 0.0029731734109210814, "optim/total_tokens": 2180251648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446742, "loss/out": 5.246587753295898, "created_at": "2025-01-15T00:43:26.094827+00:00"} {"global_step": 8318, "acc_step": 0, "speed/wps": 12960.558999276727, "speed/FLOPS": 203563388758026.5, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4107166528701782, "optim/lr": 0.002973157276829888, "optim/total_tokens": 2180513792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 5.1233038902282715, "created_at": "2025-01-15T00:43:31.159790+00:00"} {"global_step": 8319, "acc_step": 0, "speed/wps": 12961.154250734964, "speed/FLOPS": 203572738000139.34, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5616319179534912, "optim/lr": 0.0029731411379322606, "optim/total_tokens": 2180775936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302630, "loss/out": 5.018223762512207, "created_at": "2025-01-15T00:43:36.234691+00:00"} {"global_step": 8320, "acc_step": 0, "speed/wps": 12954.618071278042, "speed/FLOPS": 203470078320116.6, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.228779673576355, "optim/lr": 0.0029731249942282502, "optim/total_tokens": 2181038080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453131, "loss/out": 5.183611869812012, "created_at": "2025-01-15T00:43:41.299848+00:00"} {"global_step": 8321, "acc_step": 0, "speed/wps": 12954.696887797218, "speed/FLOPS": 203471316241855.5, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.059004306793213, "optim/lr": 0.002973108845717912, "optim/total_tokens": 2181300224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 5.150447368621826, "created_at": "2025-01-15T00:43:46.367012+00:00"} {"global_step": 8322, "acc_step": 0, "speed/wps": 12959.181573941512, "speed/FLOPS": 203541754400356.2, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2039077281951904, "optim/lr": 0.0029730926924012957, "optim/total_tokens": 2181562368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 5.235367774963379, "created_at": "2025-01-15T00:43:51.431491+00:00"} {"global_step": 8323, "acc_step": 0, "speed/wps": 12965.778310648531, "speed/FLOPS": 203645365215206.1, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4149134159088135, "optim/lr": 0.0029730765342784564, "optim/total_tokens": 2181824512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437195, "loss/out": 5.085696220397949, "created_at": "2025-01-15T00:43:56.492067+00:00"} {"global_step": 8324, "acc_step": 0, "speed/wps": 12957.868791576755, "speed/FLOPS": 203521135349365.66, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1578625440597534, "optim/lr": 0.0029730603713494456, "optim/total_tokens": 2182086656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323215, "loss/out": 5.194341659545898, "created_at": "2025-01-15T00:44:01.556797+00:00"} {"global_step": 8325, "acc_step": 0, "speed/wps": 12960.23031094697, "speed/FLOPS": 203558226256142.56, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.7627363204956055, "optim/lr": 0.002973044203614316, "optim/total_tokens": 2182348800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 399541, "loss/out": 5.209782600402832, "created_at": "2025-01-15T00:44:06.617107+00:00"} {"global_step": 8326, "acc_step": 0, "speed/wps": 12960.109961403707, "speed/FLOPS": 203556336001188.0, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1109471321105957, "optim/lr": 0.002973028031073122, "optim/total_tokens": 2182610944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 387212, "loss/out": 5.097306251525879, "created_at": "2025-01-15T00:44:11.680832+00:00"} {"global_step": 8327, "acc_step": 0, "speed/wps": 12958.809622359637, "speed/FLOPS": 203535912389647.8, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3246651887893677, "optim/lr": 0.002973011853725914, "optim/total_tokens": 2182873088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450746, "loss/out": 5.085503578186035, "created_at": "2025-01-15T00:44:16.743372+00:00"} {"global_step": 8328, "acc_step": 0, "speed/wps": 12953.71104440342, "speed/FLOPS": 203455832216665.12, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2395718097686768, "optim/lr": 0.0029729956715727457, "optim/total_tokens": 2183135232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439302, "loss/out": 5.02456521987915, "created_at": "2025-01-15T00:44:21.820371+00:00"} {"global_step": 8329, "acc_step": 0, "speed/wps": 12958.898911606706, "speed/FLOPS": 203537314799969.3, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3773349523544312, "optim/lr": 0.0029729794846136705, "optim/total_tokens": 2183397376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446505, "loss/out": 5.111789703369141, "created_at": "2025-01-15T00:44:26.882774+00:00"} {"global_step": 8330, "acc_step": 0, "speed/wps": 12957.943180837281, "speed/FLOPS": 203522303734923.78, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6370407342910767, "optim/lr": 0.0029729632928487404, "optim/total_tokens": 2183659520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312011, "loss/out": 5.077842712402344, "created_at": "2025-01-15T00:44:31.943944+00:00"} {"global_step": 8331, "acc_step": 0, "speed/wps": 12962.467264350867, "speed/FLOPS": 203593360683246.88, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.644754648208618, "optim/lr": 0.0029729470962780087, "optim/total_tokens": 2183921664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350526, "loss/out": 4.941149711608887, "created_at": "2025-01-15T00:44:37.003288+00:00"} {"global_step": 8332, "acc_step": 0, "speed/wps": 12956.600005915143, "speed/FLOPS": 203501207327055.84, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.1655092239379883, "optim/lr": 0.0029729308949015287, "optim/total_tokens": 2184183808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444119, "loss/out": 5.038640022277832, "created_at": "2025-01-15T00:44:42.070007+00:00"} {"global_step": 8333, "acc_step": 0, "speed/wps": 12959.15099242316, "speed/FLOPS": 203541274075587.06, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1409533023834229, "optim/lr": 0.002972914688719352, "optim/total_tokens": 2184445952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 5.161050319671631, "created_at": "2025-01-15T00:44:47.137382+00:00"} {"global_step": 8334, "acc_step": 0, "speed/wps": 12967.060077373766, "speed/FLOPS": 203665497122957.97, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0801827907562256, "optim/lr": 0.002972898477731532, "optim/total_tokens": 2184708096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449696, "loss/out": 5.169390678405762, "created_at": "2025-01-15T00:44:52.194961+00:00"} {"global_step": 8335, "acc_step": 0, "speed/wps": 12962.66292255347, "speed/FLOPS": 203596433764181.3, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1981552839279175, "optim/lr": 0.002972882261938122, "optim/total_tokens": 2184970240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 5.270400047302246, "created_at": "2025-01-15T00:44:57.254475+00:00"} {"global_step": 8336, "acc_step": 0, "speed/wps": 12964.120403842391, "speed/FLOPS": 203619325510612.97, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2044188976287842, "optim/lr": 0.002972866041339175, "optim/total_tokens": 2185232384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433993, "loss/out": 5.12244176864624, "created_at": "2025-01-15T00:45:02.314365+00:00"} {"global_step": 8337, "acc_step": 0, "speed/wps": 12963.116634006852, "speed/FLOPS": 203603559926021.53, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3660242557525635, "optim/lr": 0.0029728498159347434, "optim/total_tokens": 2185494528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 343042, "loss/out": 5.032844066619873, "created_at": "2025-01-15T00:45:07.374470+00:00"} {"global_step": 8338, "acc_step": 0, "speed/wps": 12962.231281291075, "speed/FLOPS": 203589654245018.12, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.8397469520568848, "optim/lr": 0.0029728335857248807, "optim/total_tokens": 2185756672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449420, "loss/out": 5.105416297912598, "created_at": "2025-01-15T00:45:12.436806+00:00"} {"global_step": 8339, "acc_step": 0, "speed/wps": 12965.462310088931, "speed/FLOPS": 203640401992187.3, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3705554008483887, "optim/lr": 0.002972817350709639, "optim/total_tokens": 2186018816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434294, "loss/out": 5.118649005889893, "created_at": "2025-01-15T00:45:17.502747+00:00"} {"global_step": 8340, "acc_step": 0, "speed/wps": 12964.09876978872, "speed/FLOPS": 203618985718071.75, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.445810317993164, "optim/lr": 0.002972801110889072, "optim/total_tokens": 2186280960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 5.132840156555176, "created_at": "2025-01-15T00:45:22.562249+00:00"} {"global_step": 8341, "acc_step": 0, "speed/wps": 12964.542021969226, "speed/FLOPS": 203625947602658.94, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.386297702789307, "optim/lr": 0.002972784866263232, "optim/total_tokens": 2186543104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420954, "loss/out": 5.085589408874512, "created_at": "2025-01-15T00:45:27.624657+00:00"} {"global_step": 8342, "acc_step": 0, "speed/wps": 12969.041449455965, "speed/FLOPS": 203696617294199.44, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2960162162780762, "optim/lr": 0.0029727686168321728, "optim/total_tokens": 2186805248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 5.113617420196533, "created_at": "2025-01-15T00:45:32.684513+00:00"} {"global_step": 8343, "acc_step": 0, "speed/wps": 12967.767853443997, "speed/FLOPS": 203676613718727.16, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.1908700466156006, "optim/lr": 0.002972752362595947, "optim/total_tokens": 2187067392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464535, "loss/out": 5.097846984863281, "created_at": "2025-01-15T00:45:37.745908+00:00"} {"global_step": 8344, "acc_step": 0, "speed/wps": 12972.183701104548, "speed/FLOPS": 203745970674246.7, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3568881750106812, "optim/lr": 0.0029727361035546077, "optim/total_tokens": 2187329536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296714, "loss/out": 5.015657901763916, "created_at": "2025-01-15T00:45:42.806334+00:00"} {"global_step": 8345, "acc_step": 0, "speed/wps": 12975.706144913478, "speed/FLOPS": 203801295494610.97, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5333253145217896, "optim/lr": 0.0029727198397082075, "optim/total_tokens": 2187591680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351058, "loss/out": 4.991883277893066, "created_at": "2025-01-15T00:45:47.863963+00:00"} {"global_step": 8346, "acc_step": 0, "speed/wps": 12970.253881796423, "speed/FLOPS": 203715660210162.97, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3769162893295288, "optim/lr": 0.0029727035710568005, "optim/total_tokens": 2187853824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431605, "loss/out": 5.1252007484436035, "created_at": "2025-01-15T00:45:52.923871+00:00"} {"global_step": 8347, "acc_step": 0, "speed/wps": 12975.23559382828, "speed/FLOPS": 203793904843212.97, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2179127931594849, "optim/lr": 0.0029726872976004385, "optim/total_tokens": 2188115968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 5.050687789916992, "created_at": "2025-01-15T00:45:57.983077+00:00"} {"global_step": 8348, "acc_step": 0, "speed/wps": 12969.073026589667, "speed/FLOPS": 203697113256475.78, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.582616806030273, "optim/lr": 0.0029726710193391757, "optim/total_tokens": 2188378112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456851, "loss/out": 5.018750190734863, "created_at": "2025-01-15T00:46:03.040363+00:00"} {"global_step": 8349, "acc_step": 0, "speed/wps": 12969.879544502588, "speed/FLOPS": 203709780728569.94, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.3102426528930664, "optim/lr": 0.002972654736273065, "optim/total_tokens": 2188640256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450471, "loss/out": 5.040096282958984, "created_at": "2025-01-15T00:46:08.098001+00:00"} {"global_step": 8350, "acc_step": 0, "speed/wps": 12981.436165492318, "speed/FLOPS": 203891293341675.03, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.3775556087493896, "optim/lr": 0.0029726384484021593, "optim/total_tokens": 2188902400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466392, "loss/out": 4.967009544372559, "created_at": "2025-01-15T00:46:13.150611+00:00"} {"global_step": 8351, "acc_step": 0, "speed/wps": 12966.272309179929, "speed/FLOPS": 203653124140966.8, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1660044193267822, "optim/lr": 0.002972622155726511, "optim/total_tokens": 2189164544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 4.965877532958984, "created_at": "2025-01-15T00:46:18.209757+00:00"} {"global_step": 8352, "acc_step": 0, "speed/wps": 12970.487068219325, "speed/FLOPS": 203719322723366.38, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0009, "optim/grad_norm": 6.359485626220703, "optim/lr": 0.0029726058582461748, "optim/total_tokens": 2189426688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 5.028765678405762, "created_at": "2025-01-15T00:46:23.266104+00:00"} {"global_step": 8353, "acc_step": 0, "speed/wps": 12945.276556940527, "speed/FLOPS": 203323357000860.03, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.8176295757293701, "optim/lr": 0.002972589555961203, "optim/total_tokens": 2189688832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442778, "loss/out": 5.073953628540039, "created_at": "2025-01-15T00:46:28.333384+00:00"} {"global_step": 8354, "acc_step": 0, "speed/wps": 12967.250731939366, "speed/FLOPS": 203668491614897.53, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.993225574493408, "optim/lr": 0.002972573248871648, "optim/total_tokens": 2189950976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311430, "loss/out": 4.988636016845703, "created_at": "2025-01-15T00:46:33.394876+00:00"} {"global_step": 8355, "acc_step": 0, "speed/wps": 12970.097157040618, "speed/FLOPS": 203713198632510.7, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.505021333694458, "optim/lr": 0.0029725569369775653, "optim/total_tokens": 2190213120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 398956, "loss/out": 5.147904396057129, "created_at": "2025-01-15T00:46:38.453458+00:00"} {"global_step": 8356, "acc_step": 0, "speed/wps": 12975.739089624729, "speed/FLOPS": 203801812936571.94, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1856462955474854, "optim/lr": 0.0029725406202790053, "optim/total_tokens": 2190475264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421239, "loss/out": 5.09602689743042, "created_at": "2025-01-15T00:46:43.507777+00:00"} {"global_step": 8357, "acc_step": 0, "speed/wps": 12975.552374833167, "speed/FLOPS": 203798880324197.44, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3104690313339233, "optim/lr": 0.0029725242987760233, "optim/total_tokens": 2190737408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475696, "loss/out": 4.972701072692871, "created_at": "2025-01-15T00:46:48.567482+00:00"} {"global_step": 8358, "acc_step": 0, "speed/wps": 12980.92923423939, "speed/FLOPS": 203883331289751.25, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2834229469299316, "optim/lr": 0.002972507972468672, "optim/total_tokens": 2190999552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465083, "loss/out": 5.103237152099609, "created_at": "2025-01-15T00:46:53.620218+00:00"} {"global_step": 8359, "acc_step": 0, "speed/wps": 12975.035766307596, "speed/FLOPS": 203790766277408.84, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.001, "optim/grad_norm": 1.3332929611206055, "optim/lr": 0.0029724916413570047, "optim/total_tokens": 2191261696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439286, "loss/out": 5.119647979736328, "created_at": "2025-01-15T00:46:58.674704+00:00"} {"global_step": 8360, "acc_step": 0, "speed/wps": 12976.099987411058, "speed/FLOPS": 203807481332231.78, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3469964265823364, "optim/lr": 0.002972475305441074, "optim/total_tokens": 2191523840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462150, "loss/out": 4.995355606079102, "created_at": "2025-01-15T00:47:03.730506+00:00"} {"global_step": 8361, "acc_step": 0, "speed/wps": 12980.353182913721, "speed/FLOPS": 203874283612104.16, "speed/curr_iter_time": 1.2725, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2373615503311157, "optim/lr": 0.0029724589647209337, "optim/total_tokens": 2191785984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 5.094699382781982, "created_at": "2025-01-15T00:47:08.787741+00:00"} {"global_step": 8362, "acc_step": 0, "speed/wps": 12973.795984686089, "speed/FLOPS": 203771293803404.6, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3049546480178833, "optim/lr": 0.0029724426191966376, "optim/total_tokens": 2192048128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436666, "loss/out": 5.029720783233643, "created_at": "2025-01-15T00:47:13.842756+00:00"} {"global_step": 8363, "acc_step": 0, "speed/wps": 12980.600900699334, "speed/FLOPS": 203878174360327.12, "speed/curr_iter_time": 1.2721, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0406428575515747, "optim/lr": 0.0029724262688682382, "optim/total_tokens": 2192310272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314573, "loss/out": 5.025521278381348, "created_at": "2025-01-15T00:47:18.894991+00:00"} {"global_step": 8364, "acc_step": 0, "speed/wps": 12974.698900861596, "speed/FLOPS": 203785475342677.8, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2451481819152832, "optim/lr": 0.00297240991373579, "optim/total_tokens": 2192572416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318405, "loss/out": 5.038581848144531, "created_at": "2025-01-15T00:47:23.953814+00:00"} {"global_step": 8365, "acc_step": 0, "speed/wps": 12971.097090029973, "speed/FLOPS": 203728903954160.4, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0009, "optim/grad_norm": 3.502335786819458, "optim/lr": 0.0029723935537993443, "optim/total_tokens": 2192834560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355325, "loss/out": 5.134328842163086, "created_at": "2025-01-15T00:47:29.029310+00:00"} {"global_step": 8366, "acc_step": 0, "speed/wps": 12971.761243867952, "speed/FLOPS": 203739335402828.88, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.115593194961548, "optim/lr": 0.002972377189058957, "optim/total_tokens": 2193096704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437432, "loss/out": 5.144885063171387, "created_at": "2025-01-15T00:47:34.091673+00:00"} {"global_step": 8367, "acc_step": 0, "speed/wps": 12968.963125156086, "speed/FLOPS": 203695387103440.97, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4751602411270142, "optim/lr": 0.0029723608195146795, "optim/total_tokens": 2193358848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 5.111992359161377, "created_at": "2025-01-15T00:47:39.153171+00:00"} {"global_step": 8368, "acc_step": 0, "speed/wps": 12966.260738348748, "speed/FLOPS": 203652942405163.12, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2322635650634766, "optim/lr": 0.0029723444451665664, "optim/total_tokens": 2193620992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441679, "loss/out": 5.127202033996582, "created_at": "2025-01-15T00:47:44.211481+00:00"} {"global_step": 8369, "acc_step": 0, "speed/wps": 12959.453360580013, "speed/FLOPS": 203546023182987.84, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.021960496902466, "optim/lr": 0.00297232806601467, "optim/total_tokens": 2193883136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 5.031716346740723, "created_at": "2025-01-15T00:47:49.276963+00:00"} {"global_step": 8370, "acc_step": 0, "speed/wps": 12967.696921569142, "speed/FLOPS": 203675499636161.97, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4351277351379395, "optim/lr": 0.0029723116820590447, "optim/total_tokens": 2194145280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453375, "loss/out": 5.145489692687988, "created_at": "2025-01-15T00:47:54.338071+00:00"} {"global_step": 8371, "acc_step": 0, "speed/wps": 12966.2183304614, "speed/FLOPS": 203652276331017.78, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5867393016815186, "optim/lr": 0.002972295293299744, "optim/total_tokens": 2194407424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295714, "loss/out": 5.156716346740723, "created_at": "2025-01-15T00:47:59.400717+00:00"} {"global_step": 8372, "acc_step": 0, "speed/wps": 12962.894417665852, "speed/FLOPS": 203600069713028.88, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.079930305480957, "optim/lr": 0.002972278899736821, "optim/total_tokens": 2194669568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453666, "loss/out": 5.142467975616455, "created_at": "2025-01-15T00:48:04.460754+00:00"} {"global_step": 8373, "acc_step": 0, "speed/wps": 12964.162872048442, "speed/FLOPS": 203619992532146.44, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.31632137298584, "optim/lr": 0.002972262501370329, "optim/total_tokens": 2194931712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457665, "loss/out": 4.977115631103516, "created_at": "2025-01-15T00:48:09.520998+00:00"} {"global_step": 8374, "acc_step": 0, "speed/wps": 12972.566194924671, "speed/FLOPS": 203751978265293.3, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.9115688800811768, "optim/lr": 0.0029722460982003216, "optim/total_tokens": 2195193856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 5.178985595703125, "created_at": "2025-01-15T00:48:14.577591+00:00"} {"global_step": 8375, "acc_step": 0, "speed/wps": 12963.412629427825, "speed/FLOPS": 203608208940848.25, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.5938602685928345, "optim/lr": 0.0029722296902268534, "optim/total_tokens": 2195456000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413790, "loss/out": 5.113739490509033, "created_at": "2025-01-15T00:48:19.652569+00:00"} {"global_step": 8376, "acc_step": 0, "speed/wps": 12969.108466494352, "speed/FLOPS": 203697669888878.38, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6318691968917847, "optim/lr": 0.0029722132774499758, "optim/total_tokens": 2195718144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457652, "loss/out": 4.992125511169434, "created_at": "2025-01-15T00:48:24.713949+00:00"} {"global_step": 8377, "acc_step": 0, "speed/wps": 12974.656208101265, "speed/FLOPS": 203784804794210.78, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5644845962524414, "optim/lr": 0.0029721968598697443, "optim/total_tokens": 2195980288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449657, "loss/out": 5.051087379455566, "created_at": "2025-01-15T00:48:29.776471+00:00"} {"global_step": 8378, "acc_step": 0, "speed/wps": 12970.461599691387, "speed/FLOPS": 203718922705137.7, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0012, "optim/grad_norm": 4.0393171310424805, "optim/lr": 0.002972180437486212, "optim/total_tokens": 2196242432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324713, "loss/out": 5.213550567626953, "created_at": "2025-01-15T00:48:34.833458+00:00"} {"global_step": 8379, "acc_step": 0, "speed/wps": 12968.446809346453, "speed/FLOPS": 203687277654158.44, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2701352834701538, "optim/lr": 0.0029721640102994312, "optim/total_tokens": 2196504576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329371, "loss/out": 5.088688850402832, "created_at": "2025-01-15T00:48:39.893256+00:00"} {"global_step": 8380, "acc_step": 0, "speed/wps": 12968.204229873472, "speed/FLOPS": 203683467610196.12, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.597771167755127, "optim/lr": 0.0029721475783094572, "optim/total_tokens": 2196766720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 280447, "loss/out": 5.087107181549072, "created_at": "2025-01-15T00:48:44.950803+00:00"} {"global_step": 8381, "acc_step": 0, "speed/wps": 12967.680752785309, "speed/FLOPS": 203675245683193.62, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5410454273223877, "optim/lr": 0.0029721311415163423, "optim/total_tokens": 2197028864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375557, "loss/out": 5.140473365783691, "created_at": "2025-01-15T00:48:50.013049+00:00"} {"global_step": 8382, "acc_step": 0, "speed/wps": 12973.16885205836, "speed/FLOPS": 203761443823715.56, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.001, "optim/grad_norm": 1.4105689525604248, "optim/lr": 0.0029721146999201415, "optim/total_tokens": 2197291008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 5.066074371337891, "created_at": "2025-01-15T00:48:55.068372+00:00"} {"global_step": 8383, "acc_step": 0, "speed/wps": 12969.690856461455, "speed/FLOPS": 203706817123599.4, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2424463033676147, "optim/lr": 0.0029720982535209072, "optim/total_tokens": 2197553152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 5.100220680236816, "created_at": "2025-01-15T00:49:00.131103+00:00"} {"global_step": 8384, "acc_step": 0, "speed/wps": 12974.280738739177, "speed/FLOPS": 203778907531930.22, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.2600884437561035, "optim/lr": 0.002972081802318694, "optim/total_tokens": 2197815296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450462, "loss/out": 5.101258754730225, "created_at": "2025-01-15T00:49:05.185971+00:00"} {"global_step": 8385, "acc_step": 0, "speed/wps": 12973.27606468021, "speed/FLOPS": 203763127745267.34, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5469273328781128, "optim/lr": 0.002972065346313554, "optim/total_tokens": 2198077440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436666, "loss/out": 5.045086860656738, "created_at": "2025-01-15T00:49:10.244739+00:00"} {"global_step": 8386, "acc_step": 0, "speed/wps": 12976.107017496332, "speed/FLOPS": 203807591749381.34, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0155738592147827, "optim/lr": 0.0029720488855055427, "optim/total_tokens": 2198339584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 5.056235313415527, "created_at": "2025-01-15T00:49:15.300248+00:00"} {"global_step": 8387, "acc_step": 0, "speed/wps": 12972.849300744694, "speed/FLOPS": 203756424831224.94, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.712362289428711, "optim/lr": 0.0029720324198947135, "optim/total_tokens": 2198601728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461913, "loss/out": 5.049773693084717, "created_at": "2025-01-15T00:49:20.360842+00:00"} {"global_step": 8388, "acc_step": 0, "speed/wps": 12978.246476981194, "speed/FLOPS": 203841194900516.53, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.094858169555664, "optim/lr": 0.002972015949481119, "optim/total_tokens": 2198863872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318931, "loss/out": 4.944890022277832, "created_at": "2025-01-15T00:49:25.414812+00:00"} {"global_step": 8389, "acc_step": 0, "speed/wps": 12974.726153946187, "speed/FLOPS": 203785903389820.97, "speed/curr_iter_time": 1.2724, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.080432415008545, "optim/lr": 0.0029719994742648144, "optim/total_tokens": 2199126016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354398, "loss/out": 4.954328536987305, "created_at": "2025-01-15T00:49:30.475669+00:00"} {"global_step": 8390, "acc_step": 0, "speed/wps": 12976.029641514357, "speed/FLOPS": 203806376453258.7, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8399503231048584, "optim/lr": 0.002971982994245852, "optim/total_tokens": 2199388160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450471, "loss/out": 4.971426010131836, "created_at": "2025-01-15T00:49:35.531324+00:00"} {"global_step": 8391, "acc_step": 0, "speed/wps": 12965.944432363614, "speed/FLOPS": 203647974385016.62, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3675450086593628, "optim/lr": 0.0029719665094242865, "optim/total_tokens": 2199650304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437717, "loss/out": 4.929263591766357, "created_at": "2025-01-15T00:49:40.622768+00:00"} {"global_step": 8392, "acc_step": 0, "speed/wps": 12972.44365756076, "speed/FLOPS": 203750053647609.1, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.3252637386322021, "optim/lr": 0.0029719500198001716, "optim/total_tokens": 2199912448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430776, "loss/out": 5.101056098937988, "created_at": "2025-01-15T00:49:45.680024+00:00"} {"global_step": 8393, "acc_step": 0, "speed/wps": 12970.352145689209, "speed/FLOPS": 203717203579627.9, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2164974212646484, "optim/lr": 0.002971933525373561, "optim/total_tokens": 2200174592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424678, "loss/out": 5.034792900085449, "created_at": "2025-01-15T00:49:50.737197+00:00"} {"global_step": 8394, "acc_step": 0, "speed/wps": 12974.106626698218, "speed/FLOPS": 203776172863071.94, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.175692081451416, "optim/lr": 0.0029719170261445086, "optim/total_tokens": 2200436736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457106, "loss/out": 5.033184051513672, "created_at": "2025-01-15T00:49:55.820115+00:00"} {"global_step": 8395, "acc_step": 0, "speed/wps": 12974.53447810131, "speed/FLOPS": 203782892857286.94, "speed/curr_iter_time": 1.2726, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1338727474212646, "optim/lr": 0.002971900522113068, "optim/total_tokens": 2200698880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314942, "loss/out": 5.066265106201172, "created_at": "2025-01-15T00:50:00.879346+00:00"} {"global_step": 8396, "acc_step": 0, "speed/wps": 12968.26948166448, "speed/FLOPS": 203684492479239.28, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.342104434967041, "optim/lr": 0.0029718840132792928, "optim/total_tokens": 2200961024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 386926, "loss/out": 5.028689861297607, "created_at": "2025-01-15T00:50:05.936859+00:00"} {"global_step": 8397, "acc_step": 0, "speed/wps": 12972.879327518194, "speed/FLOPS": 203756896442963.88, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.067525863647461, "optim/lr": 0.002971867499643238, "optim/total_tokens": 2201223168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453911, "loss/out": 5.065907955169678, "created_at": "2025-01-15T00:50:10.992325+00:00"} {"global_step": 8398, "acc_step": 0, "speed/wps": 12975.803845407916, "speed/FLOPS": 203802830015130.66, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9836186170578003, "optim/lr": 0.002971850981204956, "optim/total_tokens": 2201485312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 4.9911065101623535, "created_at": "2025-01-15T00:50:16.050025+00:00"} {"global_step": 8399, "acc_step": 0, "speed/wps": 12975.39245714581, "speed/FLOPS": 203796368597167.6, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5077422857284546, "optim/lr": 0.0029718344579645014, "optim/total_tokens": 2201747456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 4.988293647766113, "created_at": "2025-01-15T00:50:21.105802+00:00"} {"global_step": 8400, "acc_step": 0, "speed/wps": 12972.283716051761, "speed/FLOPS": 203747541546427.62, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.809633255004883, "optim/lr": 0.0029718179299219284, "optim/total_tokens": 2202009600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417484, "loss/out": 5.127300262451172, "created_at": "2025-01-15T00:50:26.161467+00:00"} {"global_step": 8401, "acc_step": 0, "speed/wps": 12967.377278932288, "speed/FLOPS": 203670479209315.16, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.0474328994750977, "optim/lr": 0.0029718013970772904, "optim/total_tokens": 2202271744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 4.983954429626465, "created_at": "2025-01-15T00:50:31.219232+00:00"} {"global_step": 8402, "acc_step": 0, "speed/wps": 12969.934569687799, "speed/FLOPS": 203710644974716.22, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9313583374023438, "optim/lr": 0.002971784859430642, "optim/total_tokens": 2202533888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444119, "loss/out": 5.025910377502441, "created_at": "2025-01-15T00:50:36.280653+00:00"} {"global_step": 8403, "acc_step": 0, "speed/wps": 12969.979779712346, "speed/FLOPS": 203711355060276.8, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.001, "optim/grad_norm": 2.2338969707489014, "optim/lr": 0.002971768316982036, "optim/total_tokens": 2202796032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313070, "loss/out": 4.9798479080200195, "created_at": "2025-01-15T00:50:41.338302+00:00"} {"global_step": 8404, "acc_step": 0, "speed/wps": 12960.816742614847, "speed/FLOPS": 203567436971328.12, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.653032064437866, "optim/lr": 0.002971751769731528, "optim/total_tokens": 2203058176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355086, "loss/out": 5.068986892700195, "created_at": "2025-01-15T00:50:46.403342+00:00"} {"global_step": 8405, "acc_step": 0, "speed/wps": 12976.444848632069, "speed/FLOPS": 203812897851596.72, "speed/curr_iter_time": 1.2717, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1371999979019165, "optim/lr": 0.0029717352176791704, "optim/total_tokens": 2203320320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421450, "loss/out": 5.087919235229492, "created_at": "2025-01-15T00:50:51.461763+00:00"} {"global_step": 8406, "acc_step": 0, "speed/wps": 12973.01334356989, "speed/FLOPS": 203759001349213.1, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5665756464004517, "optim/lr": 0.0029717186608250183, "optim/total_tokens": 2203582464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446220, "loss/out": 5.06872034072876, "created_at": "2025-01-15T00:50:56.522122+00:00"} {"global_step": 8407, "acc_step": 0, "speed/wps": 12971.207489098431, "speed/FLOPS": 203730637923234.75, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.203163743019104, "optim/lr": 0.0029717020991691253, "optim/total_tokens": 2203844608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 4.978427410125732, "created_at": "2025-01-15T00:51:01.579386+00:00"} {"global_step": 8408, "acc_step": 0, "speed/wps": 12972.807722682091, "speed/FLOPS": 203755771790617.53, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2997137308120728, "optim/lr": 0.002971685532711546, "optim/total_tokens": 2204106752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451797, "loss/out": 5.111270904541016, "created_at": "2025-01-15T00:51:06.636224+00:00"} {"global_step": 8409, "acc_step": 0, "speed/wps": 12968.619428690077, "speed/FLOPS": 203689988878153.88, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0009, "optim/grad_norm": 3.03538179397583, "optim/lr": 0.0029716689614523325, "optim/total_tokens": 2204368896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422813, "loss/out": 5.079751014709473, "created_at": "2025-01-15T00:51:11.699037+00:00"} {"global_step": 8410, "acc_step": 0, "speed/wps": 12971.266618993513, "speed/FLOPS": 203731566639490.03, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0844756364822388, "optim/lr": 0.002971652385391541, "optim/total_tokens": 2204631040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 4.991591453552246, "created_at": "2025-01-15T00:51:16.760447+00:00"} {"global_step": 8411, "acc_step": 0, "speed/wps": 12963.00413507685, "speed/FLOPS": 203601792975736.12, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.5791358947753906, "optim/lr": 0.002971635804529225, "optim/total_tokens": 2204893184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425201, "loss/out": 5.05477237701416, "created_at": "2025-01-15T00:51:21.820836+00:00"} {"global_step": 8412, "acc_step": 0, "speed/wps": 12974.883220258884, "speed/FLOPS": 203788370332093.6, "speed/curr_iter_time": 1.2719, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.718554973602295, "optim/lr": 0.0029716192188654384, "optim/total_tokens": 2205155328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309025, "loss/out": 5.047797203063965, "created_at": "2025-01-15T00:51:26.881387+00:00"} {"global_step": 8413, "acc_step": 0, "speed/wps": 12973.791660019715, "speed/FLOPS": 203771225878576.5, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.40989089012146, "optim/lr": 0.0029716026284002353, "optim/total_tokens": 2205417472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355857, "loss/out": 5.069981575012207, "created_at": "2025-01-15T00:51:31.943749+00:00"} {"global_step": 8414, "acc_step": 0, "speed/wps": 12971.013959023941, "speed/FLOPS": 203727598267476.72, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2565511465072632, "optim/lr": 0.00297158603313367, "optim/total_tokens": 2205679616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 5.103306770324707, "created_at": "2025-01-15T00:51:37.002408+00:00"} {"global_step": 8415, "acc_step": 0, "speed/wps": 12962.36471449524, "speed/FLOPS": 203591749996845.97, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3231745958328247, "optim/lr": 0.002971569433065796, "optim/total_tokens": 2205941760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446742, "loss/out": 4.968774795532227, "created_at": "2025-01-15T00:51:42.065206+00:00"} {"global_step": 8416, "acc_step": 0, "speed/wps": 12970.820618820915, "speed/FLOPS": 203724561593910.9, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.341516375541687, "optim/lr": 0.002971552828196669, "optim/total_tokens": 2206203904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460578, "loss/out": 5.089720726013184, "created_at": "2025-01-15T00:51:47.132394+00:00"} {"global_step": 8417, "acc_step": 0, "speed/wps": 12973.474575108461, "speed/FLOPS": 203766245624324.9, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.195455551147461, "optim/lr": 0.0029715362185263414, "optim/total_tokens": 2206466048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456045, "loss/out": 4.998177528381348, "created_at": "2025-01-15T00:51:52.188061+00:00"} {"global_step": 8418, "acc_step": 0, "speed/wps": 12969.004006933297, "speed/FLOPS": 203696029207929.53, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3931461572647095, "optim/lr": 0.0029715196040548683, "optim/total_tokens": 2206728192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436381, "loss/out": 5.060733318328857, "created_at": "2025-01-15T00:51:57.245010+00:00"} {"global_step": 8419, "acc_step": 0, "speed/wps": 12971.339335396575, "speed/FLOPS": 203732708750522.9, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7808177471160889, "optim/lr": 0.0029715029847823043, "optim/total_tokens": 2206990336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430811, "loss/out": 5.028836727142334, "created_at": "2025-01-15T00:52:02.302528+00:00"} {"global_step": 8420, "acc_step": 0, "speed/wps": 12973.315100885253, "speed/FLOPS": 203763740862508.88, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2086387872695923, "optim/lr": 0.0029714863607087027, "optim/total_tokens": 2207252480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329804, "loss/out": 5.0563249588012695, "created_at": "2025-01-15T00:52:07.363349+00:00"} {"global_step": 8421, "acc_step": 0, "speed/wps": 12972.226269982706, "speed/FLOPS": 203746639276973.94, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.214094638824463, "optim/lr": 0.0029714697318341176, "optim/total_tokens": 2207514624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352926, "loss/out": 5.033190727233887, "created_at": "2025-01-15T00:52:12.419996+00:00"} {"global_step": 8422, "acc_step": 0, "speed/wps": 12968.34396410153, "speed/FLOPS": 203685662328262.75, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0014, "optim/grad_norm": 1.3152575492858887, "optim/lr": 0.0029714530981586047, "optim/total_tokens": 2207776768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447269, "loss/out": 5.05414342880249, "created_at": "2025-01-15T00:52:17.480744+00:00"} {"global_step": 8423, "acc_step": 0, "speed/wps": 12980.76343950563, "speed/FLOPS": 203880727255631.75, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2483155727386475, "optim/lr": 0.0029714364596822164, "optim/total_tokens": 2208038912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443545, "loss/out": 4.909740447998047, "created_at": "2025-01-15T00:52:22.533795+00:00"} {"global_step": 8424, "acc_step": 0, "speed/wps": 12969.239796848591, "speed/FLOPS": 203699732612558.53, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1889389753341675, "optim/lr": 0.002971419816405009, "optim/total_tokens": 2208301056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 4.925194263458252, "created_at": "2025-01-15T00:52:27.595826+00:00"} {"global_step": 8425, "acc_step": 0, "speed/wps": 12962.70278797993, "speed/FLOPS": 203597059905484.7, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3505266904830933, "optim/lr": 0.002971403168327035, "optim/total_tokens": 2208563200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440390, "loss/out": 4.986085891723633, "created_at": "2025-01-15T00:52:32.661053+00:00"} {"global_step": 8426, "acc_step": 0, "speed/wps": 12967.885257653903, "speed/FLOPS": 203678457713174.03, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.398585319519043, "optim/lr": 0.00297138651544835, "optim/total_tokens": 2208825344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 4.991647720336914, "created_at": "2025-01-15T00:52:37.718355+00:00"} {"global_step": 8427, "acc_step": 0, "speed/wps": 12975.733139035192, "speed/FLOPS": 203801719474386.28, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0008, "optim/grad_norm": 22.916221618652344, "optim/lr": 0.0029713698577690075, "optim/total_tokens": 2209087488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454182, "loss/out": 5.154187202453613, "created_at": "2025-01-15T00:52:42.773919+00:00"} {"global_step": 8428, "acc_step": 0, "speed/wps": 12968.322029156008, "speed/FLOPS": 203685317809801.53, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.0228114128112793, "optim/lr": 0.0029713531952890618, "optim/total_tokens": 2209349632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321343, "loss/out": 5.140220642089844, "created_at": "2025-01-15T00:52:47.833865+00:00"} {"global_step": 8429, "acc_step": 0, "speed/wps": 12940.154875543609, "speed/FLOPS": 203242913956593.7, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2991721630096436, "optim/lr": 0.0029713365280085685, "optim/total_tokens": 2209611776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456045, "loss/out": 5.074443340301514, "created_at": "2025-01-15T00:52:52.907190+00:00"} {"global_step": 8430, "acc_step": 0, "speed/wps": 12977.306354335771, "speed/FLOPS": 203826428982506.25, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5674738883972168, "optim/lr": 0.0029713198559275806, "optim/total_tokens": 2209873920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463485, "loss/out": 5.021807670593262, "created_at": "2025-01-15T00:52:57.966625+00:00"} {"global_step": 8431, "acc_step": 0, "speed/wps": 12976.558222292315, "speed/FLOPS": 203814678540722.62, "speed/curr_iter_time": 1.2718, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.7941442728042603, "optim/lr": 0.0029713031790461527, "optim/total_tokens": 2210136064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 5.054513454437256, "created_at": "2025-01-15T00:53:03.049490+00:00"} {"global_step": 8432, "acc_step": 0, "speed/wps": 12976.784915729491, "speed/FLOPS": 203818239072662.9, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.293177127838135, "optim/lr": 0.00297128649736434, "optim/total_tokens": 2210398208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 5.066640853881836, "created_at": "2025-01-15T00:53:08.105886+00:00"} {"global_step": 8433, "acc_step": 0, "speed/wps": 12981.036345281793, "speed/FLOPS": 203885013615858.22, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5478404760360718, "optim/lr": 0.002971269810882197, "optim/total_tokens": 2210660352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 5.147593975067139, "created_at": "2025-01-15T00:53:13.160897+00:00"} {"global_step": 8434, "acc_step": 0, "speed/wps": 12972.78901802195, "speed/FLOPS": 203755478008227.1, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3684732913970947, "optim/lr": 0.0029712531195997764, "optim/total_tokens": 2210922496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449135, "loss/out": 5.045431137084961, "created_at": "2025-01-15T00:53:18.216606+00:00"} {"global_step": 8435, "acc_step": 0, "speed/wps": 12974.293374040442, "speed/FLOPS": 203779105986699.3, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.277457594871521, "optim/lr": 0.002971236423517134, "optim/total_tokens": 2211184640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316287, "loss/out": 4.93149995803833, "created_at": "2025-01-15T00:53:23.276148+00:00"} {"global_step": 8436, "acc_step": 0, "speed/wps": 12987.676781416934, "speed/FLOPS": 203989310790276.06, "speed/curr_iter_time": 1.2719, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1838767528533936, "optim/lr": 0.002971219722634324, "optim/total_tokens": 2211446784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297348, "loss/out": 5.0346150398254395, "created_at": "2025-01-15T00:53:28.328059+00:00"} {"global_step": 8437, "acc_step": 0, "speed/wps": 12973.97065290776, "speed/FLOPS": 203774037207845.22, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.386369228363037, "optim/lr": 0.0029712030169514016, "optim/total_tokens": 2211708928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360703, "loss/out": 5.038667678833008, "created_at": "2025-01-15T00:53:33.384469+00:00"} {"global_step": 8438, "acc_step": 0, "speed/wps": 12968.012875726563, "speed/FLOPS": 203680462130370.78, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.001, "optim/grad_norm": 2.827080249786377, "optim/lr": 0.0029711863064684207, "optim/total_tokens": 2211971072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453382, "loss/out": 5.037360191345215, "created_at": "2025-01-15T00:53:38.447016+00:00"} {"global_step": 8439, "acc_step": 0, "speed/wps": 12989.782294072407, "speed/FLOPS": 204022380759807.88, "speed/curr_iter_time": 1.2722, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3073409795761108, "optim/lr": 0.0029711695911854353, "optim/total_tokens": 2212233216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435044, "loss/out": 5.103531837463379, "created_at": "2025-01-15T00:53:43.500804+00:00"} {"global_step": 8440, "acc_step": 0, "speed/wps": 12977.948741062413, "speed/FLOPS": 203836518548780.03, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6484302282333374, "optim/lr": 0.002971152871102501, "optim/total_tokens": 2212495360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 4.931021690368652, "created_at": "2025-01-15T00:53:48.555612+00:00"} {"global_step": 8441, "acc_step": 0, "speed/wps": 12983.105767204259, "speed/FLOPS": 203917516730834.25, "speed/curr_iter_time": 1.272, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6706074476242065, "optim/lr": 0.0029711361462196714, "optim/total_tokens": 2212757504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 5.071642875671387, "created_at": "2025-01-15T00:53:53.636894+00:00"} {"global_step": 8442, "acc_step": 0, "speed/wps": 12972.639978175017, "speed/FLOPS": 203753137132628.8, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.189764976501465, "optim/lr": 0.002971119416537002, "optim/total_tokens": 2213019648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461889, "loss/out": 5.110626697540283, "created_at": "2025-01-15T00:53:58.694696+00:00"} {"global_step": 8443, "acc_step": 0, "speed/wps": 12971.535572373414, "speed/FLOPS": 203735790921902.03, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7855637073516846, "optim/lr": 0.002971102682054546, "optim/total_tokens": 2213281792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435567, "loss/out": 5.097506523132324, "created_at": "2025-01-15T00:54:03.756269+00:00"} {"global_step": 8444, "acc_step": 0, "speed/wps": 12973.715031619578, "speed/FLOPS": 203770022324253.88, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.001, "optim/grad_norm": 1.510978102684021, "optim/lr": 0.00297108594277236, "optim/total_tokens": 2213543936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320571, "loss/out": 5.019858360290527, "created_at": "2025-01-15T00:54:08.812298+00:00"} {"global_step": 8445, "acc_step": 0, "speed/wps": 12972.995334291882, "speed/FLOPS": 203758718488754.66, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7289434671401978, "optim/lr": 0.002971069198690496, "optim/total_tokens": 2213806080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338116, "loss/out": 5.012592792510986, "created_at": "2025-01-15T00:54:13.867718+00:00"} {"global_step": 8446, "acc_step": 0, "speed/wps": 12967.160139974812, "speed/FLOPS": 203667068743607.8, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1851379871368408, "optim/lr": 0.0029710524498090113, "optim/total_tokens": 2214068224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308860, "loss/out": 4.925861358642578, "created_at": "2025-01-15T00:54:18.925408+00:00"} {"global_step": 8447, "acc_step": 0, "speed/wps": 12969.770936991103, "speed/FLOPS": 203708074898359.72, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.913061141967773, "optim/lr": 0.0029710356961279586, "optim/total_tokens": 2214330368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362914, "loss/out": 4.954805374145508, "created_at": "2025-01-15T00:54:23.987995+00:00"} {"global_step": 8448, "acc_step": 0, "speed/wps": 12976.595110198301, "speed/FLOPS": 203815257915975.44, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.554958820343018, "optim/lr": 0.002971018937647394, "optim/total_tokens": 2214592512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436381, "loss/out": 5.035065174102783, "created_at": "2025-01-15T00:54:29.041968+00:00"} {"global_step": 8449, "acc_step": 0, "speed/wps": 12974.092886623435, "speed/FLOPS": 203775957056316.66, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7205580472946167, "optim/lr": 0.0029710021743673716, "optim/total_tokens": 2214854656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463222, "loss/out": 5.064707279205322, "created_at": "2025-01-15T00:54:34.097187+00:00"} {"global_step": 8450, "acc_step": 0, "speed/wps": 12966.21205889597, "speed/FLOPS": 203652177827464.66, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.7161740064620972, "optim/lr": 0.002970985406287945, "optim/total_tokens": 2215116800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 5.0186309814453125, "created_at": "2025-01-15T00:54:39.163142+00:00"} {"global_step": 8451, "acc_step": 0, "speed/wps": 12976.164471620332, "speed/FLOPS": 203808494145349.0, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5028610229492188, "optim/lr": 0.0029709686334091706, "optim/total_tokens": 2215378944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429438, "loss/out": 4.975065231323242, "created_at": "2025-01-15T00:54:44.222125+00:00"} {"global_step": 8452, "acc_step": 0, "speed/wps": 12969.180683040417, "speed/FLOPS": 203698804148970.25, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1928435564041138, "optim/lr": 0.002970951855731102, "optim/total_tokens": 2215641088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 4.968697547912598, "created_at": "2025-01-15T00:54:49.283774+00:00"} {"global_step": 8453, "acc_step": 0, "speed/wps": 12967.943922319173, "speed/FLOPS": 203679379122355.8, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.7226601839065552, "optim/lr": 0.002970935073253795, "optim/total_tokens": 2215903232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443307, "loss/out": 5.088381767272949, "created_at": "2025-01-15T00:54:54.344736+00:00"} {"global_step": 8454, "acc_step": 0, "speed/wps": 12959.17815636495, "speed/FLOPS": 203541700722620.03, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.956740140914917, "optim/lr": 0.002970918285977303, "optim/total_tokens": 2216165376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360279, "loss/out": 4.897011756896973, "created_at": "2025-01-15T00:54:59.406178+00:00"} {"global_step": 8455, "acc_step": 0, "speed/wps": 12962.86369444911, "speed/FLOPS": 203599587162691.8, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4092379808425903, "optim/lr": 0.002970901493901682, "optim/total_tokens": 2216427520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422813, "loss/out": 5.070782661437988, "created_at": "2025-01-15T00:55:04.472207+00:00"} {"global_step": 8456, "acc_step": 0, "speed/wps": 12967.192140452433, "speed/FLOPS": 203667571355082.1, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1151272058486938, "optim/lr": 0.002970884697026986, "optim/total_tokens": 2216689664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 5.09090518951416, "created_at": "2025-01-15T00:55:09.531428+00:00"} {"global_step": 8457, "acc_step": 0, "speed/wps": 12966.054612337972, "speed/FLOPS": 203649704912917.06, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5915292501449585, "optim/lr": 0.0029708678953532706, "optim/total_tokens": 2216951808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428403, "loss/out": 5.056920051574707, "created_at": "2025-01-15T00:55:14.590610+00:00"} {"global_step": 8458, "acc_step": 0, "speed/wps": 12963.413993053104, "speed/FLOPS": 203608230358457.06, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5958797931671143, "optim/lr": 0.0029708510888805896, "optim/total_tokens": 2217213952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 5.023384094238281, "created_at": "2025-01-15T00:55:19.653723+00:00"} {"global_step": 8459, "acc_step": 0, "speed/wps": 12958.45678019, "speed/FLOPS": 203530370518517.7, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.13527512550354, "optim/lr": 0.002970834277608998, "optim/total_tokens": 2217476096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 5.111355304718018, "created_at": "2025-01-15T00:55:24.715445+00:00"} {"global_step": 8460, "acc_step": 0, "speed/wps": 12968.274034473, "speed/FLOPS": 203684563987353.28, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3485684394836426, "optim/lr": 0.002970817461538552, "optim/total_tokens": 2217738240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 5.0785441398620605, "created_at": "2025-01-15T00:55:29.774036+00:00"} {"global_step": 8461, "acc_step": 0, "speed/wps": 12967.48156891167, "speed/FLOPS": 203672117226750.8, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0472139120101929, "optim/lr": 0.0029708006406693045, "optim/total_tokens": 2218000384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425730, "loss/out": 5.068662643432617, "created_at": "2025-01-15T00:55:34.831357+00:00"} {"global_step": 8462, "acc_step": 0, "speed/wps": 12963.09741281383, "speed/FLOPS": 203603258030772.44, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1581804752349854, "optim/lr": 0.0029707838150013115, "optim/total_tokens": 2218262528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305919, "loss/out": 5.110563278198242, "created_at": "2025-01-15T00:55:39.891329+00:00"} {"global_step": 8463, "acc_step": 0, "speed/wps": 12967.100349780441, "speed/FLOPS": 203666129656444.88, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2200162410736084, "optim/lr": 0.0029707669845346276, "optim/total_tokens": 2218524672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444644, "loss/out": 5.032683372497559, "created_at": "2025-01-15T00:55:44.953830+00:00"} {"global_step": 8464, "acc_step": 0, "speed/wps": 12959.17448104021, "speed/FLOPS": 203541642996594.6, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1285442113876343, "optim/lr": 0.0029707501492693086, "optim/total_tokens": 2218786816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 5.125638008117676, "created_at": "2025-01-15T00:55:50.018444+00:00"} {"global_step": 8465, "acc_step": 0, "speed/wps": 12968.329875347763, "speed/FLOPS": 203685441045024.84, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.110269546508789, "optim/lr": 0.0029707333092054084, "optim/total_tokens": 2219048960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432656, "loss/out": 5.010787487030029, "created_at": "2025-01-15T00:55:55.078124+00:00"} {"global_step": 8466, "acc_step": 0, "speed/wps": 12965.192289882294, "speed/FLOPS": 203636160953795.4, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0009, "optim/grad_norm": 5.499515056610107, "optim/lr": 0.0029707164643429816, "optim/total_tokens": 2219311104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 4.99500846862793, "created_at": "2025-01-15T00:56:00.137638+00:00"} {"global_step": 8467, "acc_step": 0, "speed/wps": 12960.974140769873, "speed/FLOPS": 203569909125640.56, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.193886637687683, "optim/lr": 0.0029706996146820846, "optim/total_tokens": 2219573248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466369, "loss/out": 4.958428859710693, "created_at": "2025-01-15T00:56:05.202472+00:00"} {"global_step": 8468, "acc_step": 0, "speed/wps": 12959.359903773733, "speed/FLOPS": 203544555315422.1, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3493164777755737, "optim/lr": 0.0029706827602227707, "optim/total_tokens": 2219835392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359360, "loss/out": 4.983283996582031, "created_at": "2025-01-15T00:56:10.266780+00:00"} {"global_step": 8469, "acc_step": 0, "speed/wps": 12964.604122209885, "speed/FLOPS": 203626922972273.16, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1850351095199585, "optim/lr": 0.0029706659009650963, "optim/total_tokens": 2220097536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456851, "loss/out": 4.995793342590332, "created_at": "2025-01-15T00:56:15.326829+00:00"} {"global_step": 8470, "acc_step": 0, "speed/wps": 12958.159340698212, "speed/FLOPS": 203525698822574.47, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1114064455032349, "optim/lr": 0.002970649036909115, "optim/total_tokens": 2220359680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 4.946499347686768, "created_at": "2025-01-15T00:56:20.391688+00:00"} {"global_step": 8471, "acc_step": 0, "speed/wps": 12947.840613379769, "speed/FLOPS": 203363629030621.1, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.059358835220337, "optim/lr": 0.0029706321680548838, "optim/total_tokens": 2220621824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431842, "loss/out": 5.0076189041137695, "created_at": "2025-01-15T00:56:25.456927+00:00"} {"global_step": 8472, "acc_step": 0, "speed/wps": 12965.779562488207, "speed/FLOPS": 203645384877068.4, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2500457763671875, "optim/lr": 0.002970615294402456, "optim/total_tokens": 2220883968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442778, "loss/out": 4.988217353820801, "created_at": "2025-01-15T00:56:30.518443+00:00"} {"global_step": 8473, "acc_step": 0, "speed/wps": 12963.879395295864, "speed/FLOPS": 203615540140209.3, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3461419343948364, "optim/lr": 0.0029705984159518876, "optim/total_tokens": 2221146112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 5.0575432777404785, "created_at": "2025-01-15T00:56:35.577976+00:00"} {"global_step": 8474, "acc_step": 0, "speed/wps": 12969.821957170829, "speed/FLOPS": 203708876240391.38, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.4929611682891846, "optim/lr": 0.0029705815327032326, "optim/total_tokens": 2221408256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459237, "loss/out": 4.936009883880615, "created_at": "2025-01-15T00:56:40.640047+00:00"} {"global_step": 8475, "acc_step": 0, "speed/wps": 12964.291255215616, "speed/FLOPS": 203622008966203.84, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5287468433380127, "optim/lr": 0.002970564644656548, "optim/total_tokens": 2221670400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 286904, "loss/out": 4.996394157409668, "created_at": "2025-01-15T00:56:45.698653+00:00"} {"global_step": 8476, "acc_step": 0, "speed/wps": 12963.453944296503, "speed/FLOPS": 203608857847633.4, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2719478607177734, "optim/lr": 0.002970547751811886, "optim/total_tokens": 2221932544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 389966, "loss/out": 4.976359844207764, "created_at": "2025-01-15T00:56:50.778561+00:00"} {"global_step": 8477, "acc_step": 0, "speed/wps": 12957.837342948142, "speed/FLOPS": 203520641405438.34, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7220237851142883, "optim/lr": 0.002970530854169305, "optim/total_tokens": 2222194688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433718, "loss/out": 4.906900405883789, "created_at": "2025-01-15T00:56:55.846749+00:00"} {"global_step": 8478, "acc_step": 0, "speed/wps": 12963.253170833264, "speed/FLOPS": 203605704424501.6, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6033461093902588, "optim/lr": 0.0029705139517288578, "optim/total_tokens": 2222456832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483617, "loss/out": 4.939236640930176, "created_at": "2025-01-15T00:57:00.905801+00:00"} {"global_step": 8479, "acc_step": 0, "speed/wps": 12955.231991653844, "speed/FLOPS": 203479720783233.56, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2193515300750732, "optim/lr": 0.0029704970444906003, "optim/total_tokens": 2222718976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332731, "loss/out": 5.115546226501465, "created_at": "2025-01-15T00:57:05.970491+00:00"} {"global_step": 8480, "acc_step": 0, "speed/wps": 12954.381317449519, "speed/FLOPS": 203466359775905.78, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0966625213623047, "optim/lr": 0.0029704801324545883, "optim/total_tokens": 2222981120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375475, "loss/out": 5.160735607147217, "created_at": "2025-01-15T00:57:11.035534+00:00"} {"global_step": 8481, "acc_step": 0, "speed/wps": 12951.480804281353, "speed/FLOPS": 203420803230876.97, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7437799572944641, "optim/lr": 0.0029704632156208758, "optim/total_tokens": 2223243264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466392, "loss/out": 5.139291763305664, "created_at": "2025-01-15T00:57:16.100833+00:00"} {"global_step": 8482, "acc_step": 0, "speed/wps": 12960.0097116064, "speed/FLOPS": 203554761440363.66, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.001, "optim/grad_norm": 1.5549908876419067, "optim/lr": 0.0029704462939895183, "optim/total_tokens": 2223505408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440921, "loss/out": 4.940260887145996, "created_at": "2025-01-15T00:57:21.165624+00:00"} {"global_step": 8483, "acc_step": 0, "speed/wps": 12951.297125345965, "speed/FLOPS": 203417918300795.4, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.144874572753906, "optim/lr": 0.0029704293675605715, "optim/total_tokens": 2223767552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 5.114477157592773, "created_at": "2025-01-15T00:57:26.235331+00:00"} {"global_step": 8484, "acc_step": 0, "speed/wps": 12960.747999393048, "speed/FLOPS": 203566357264566.7, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6464060544967651, "optim/lr": 0.0029704124363340904, "optim/total_tokens": 2224029696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428932, "loss/out": 4.985603332519531, "created_at": "2025-01-15T00:57:31.303116+00:00"} {"global_step": 8485, "acc_step": 0, "speed/wps": 12961.055164090956, "speed/FLOPS": 203571181708235.88, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.150976538658142, "optim/lr": 0.0029703955003101304, "optim/total_tokens": 2224291840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294941, "loss/out": 5.033524990081787, "created_at": "2025-01-15T00:57:36.364423+00:00"} {"global_step": 8486, "acc_step": 0, "speed/wps": 12963.585298670767, "speed/FLOPS": 203610920948581.06, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3872184753417969, "optim/lr": 0.0029703785594887466, "optim/total_tokens": 2224553984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 347191, "loss/out": 5.040534973144531, "created_at": "2025-01-15T00:57:41.428630+00:00"} {"global_step": 8487, "acc_step": 0, "speed/wps": 12959.700792061747, "speed/FLOPS": 203549909434414.94, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.294693112373352, "optim/lr": 0.0029703616138699934, "optim/total_tokens": 2224816128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430811, "loss/out": 5.055793762207031, "created_at": "2025-01-15T00:57:46.490646+00:00"} {"global_step": 8488, "acc_step": 0, "speed/wps": 12958.442202241717, "speed/FLOPS": 203530141551807.72, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3771690130233765, "optim/lr": 0.002970344663453928, "optim/total_tokens": 2225078272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 5.078042984008789, "created_at": "2025-01-15T00:57:51.555813+00:00"} {"global_step": 8489, "acc_step": 0, "speed/wps": 12958.744200838959, "speed/FLOPS": 203534884854766.97, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3461307287216187, "optim/lr": 0.002970327708240604, "optim/total_tokens": 2225340416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430506, "loss/out": 5.020623207092285, "created_at": "2025-01-15T00:57:56.620522+00:00"} {"global_step": 8490, "acc_step": 0, "speed/wps": 12966.070545322835, "speed/FLOPS": 203649955162338.56, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8118864297866821, "optim/lr": 0.0029703107482300776, "optim/total_tokens": 2225602560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 4.937889575958252, "created_at": "2025-01-15T00:58:01.683672+00:00"} {"global_step": 8491, "acc_step": 0, "speed/wps": 12964.244935999193, "speed/FLOPS": 203621281459260.66, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7244338989257812, "optim/lr": 0.0029702937834224035, "optim/total_tokens": 2225864704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432656, "loss/out": 5.024890422821045, "created_at": "2025-01-15T00:58:06.750603+00:00"} {"global_step": 8492, "acc_step": 0, "speed/wps": 12966.788783976559, "speed/FLOPS": 203661236087358.62, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3744908571243286, "optim/lr": 0.002970276813817638, "optim/total_tokens": 2226126848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330623, "loss/out": 5.061729431152344, "created_at": "2025-01-15T00:58:11.813996+00:00"} {"global_step": 8493, "acc_step": 0, "speed/wps": 12954.320927219618, "speed/FLOPS": 203465411264360.3, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3787956237792969, "optim/lr": 0.0029702598394158357, "optim/total_tokens": 2226388992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419618, "loss/out": 5.120062828063965, "created_at": "2025-01-15T00:58:16.879152+00:00"} {"global_step": 8494, "acc_step": 0, "speed/wps": 12968.633080692198, "speed/FLOPS": 203690203301607.0, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1549268960952759, "optim/lr": 0.002970242860217052, "optim/total_tokens": 2226651136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502417, "loss/out": 4.880080223083496, "created_at": "2025-01-15T00:58:21.938620+00:00"} {"global_step": 8495, "acc_step": 0, "speed/wps": 12965.480964960405, "speed/FLOPS": 203640694992578.28, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0814720392227173, "optim/lr": 0.002970225876221343, "optim/total_tokens": 2226913280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450227, "loss/out": 4.98489236831665, "created_at": "2025-01-15T00:58:26.998738+00:00"} {"global_step": 8496, "acc_step": 0, "speed/wps": 12957.175766474751, "speed/FLOPS": 203510250437822.4, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.388798713684082, "optim/lr": 0.002970208887428763, "optim/total_tokens": 2227175424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 5.016423225402832, "created_at": "2025-01-15T00:58:32.062720+00:00"} {"global_step": 8497, "acc_step": 0, "speed/wps": 12959.351328574254, "speed/FLOPS": 203544420630130.72, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3879179954528809, "optim/lr": 0.002970191893839368, "optim/total_tokens": 2227437568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445933, "loss/out": 4.910634994506836, "created_at": "2025-01-15T00:58:37.123547+00:00"} {"global_step": 8498, "acc_step": 0, "speed/wps": 12955.557626226959, "speed/FLOPS": 203484835321673.12, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1109936237335205, "optim/lr": 0.002970174895453214, "optim/total_tokens": 2227699712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325525, "loss/out": 4.991647243499756, "created_at": "2025-01-15T00:58:42.189931+00:00"} {"global_step": 8499, "acc_step": 0, "speed/wps": 12955.768449258321, "speed/FLOPS": 203488146587081.88, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 10.736001014709473, "optim/lr": 0.0029701578922703554, "optim/total_tokens": 2227961856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297061, "loss/out": 5.035396575927734, "created_at": "2025-01-15T00:58:47.256138+00:00"} {"global_step": 8500, "acc_step": 0, "speed/wps": 12948.851757176097, "speed/FLOPS": 203379510433399.62, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8950583934783936, "optim/lr": 0.0029701408842908485, "optim/total_tokens": 2228224000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432431, "loss/out": 5.078192710876465, "created_at": "2025-01-15T00:58:52.321403+00:00"} {"global_step": 8501, "acc_step": 0, "speed/wps": 12954.620527721285, "speed/FLOPS": 203470116901933.22, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.0879889726638794, "optim/lr": 0.0029701238715147482, "optim/total_tokens": 2228486144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 4.984208106994629, "created_at": "2025-01-15T00:58:57.385054+00:00"} {"global_step": 8502, "acc_step": 0, "speed/wps": 12953.197680484185, "speed/FLOPS": 203447769130877.6, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5739942789077759, "optim/lr": 0.0029701068539421107, "optim/total_tokens": 2228748288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437432, "loss/out": 4.94194221496582, "created_at": "2025-01-15T00:59:02.452702+00:00"} {"global_step": 8503, "acc_step": 0, "speed/wps": 12942.977872619558, "speed/FLOPS": 203287253004875.97, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0165280103683472, "optim/lr": 0.002970089831572991, "optim/total_tokens": 2229010432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448605, "loss/out": 4.846045970916748, "created_at": "2025-01-15T00:59:07.520823+00:00"} {"global_step": 8504, "acc_step": 0, "speed/wps": 12960.221137855693, "speed/FLOPS": 203558082180138.9, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6878818869590759, "optim/lr": 0.0029700728044074446, "optim/total_tokens": 2229272576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326058, "loss/out": 4.981151580810547, "created_at": "2025-01-15T00:59:12.590799+00:00"} {"global_step": 8505, "acc_step": 0, "speed/wps": 12949.774177729472, "speed/FLOPS": 203393998315729.03, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9258214831352234, "optim/lr": 0.0029700557724455278, "optim/total_tokens": 2229534720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438531, "loss/out": 4.922336578369141, "created_at": "2025-01-15T00:59:17.660059+00:00"} {"global_step": 8506, "acc_step": 0, "speed/wps": 12956.791533933396, "speed/FLOPS": 203504215537769.6, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.038037657737732, "optim/lr": 0.0029700387356872953, "optim/total_tokens": 2229796864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 5.028273582458496, "created_at": "2025-01-15T00:59:22.726709+00:00"} {"global_step": 8507, "acc_step": 0, "speed/wps": 12948.461942440945, "speed/FLOPS": 203373387857322.47, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.219771146774292, "optim/lr": 0.0029700216941328027, "optim/total_tokens": 2230059008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 5.010003089904785, "created_at": "2025-01-15T00:59:27.816655+00:00"} {"global_step": 8508, "acc_step": 0, "speed/wps": 12964.052719445257, "speed/FLOPS": 203618262434147.88, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2234042882919312, "optim/lr": 0.0029700046477821058, "optim/total_tokens": 2230321152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345442, "loss/out": 4.99611759185791, "created_at": "2025-01-15T00:59:32.878935+00:00"} {"global_step": 8509, "acc_step": 0, "speed/wps": 12958.446667018268, "speed/FLOPS": 203530211677258.72, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6011230945587158, "optim/lr": 0.0029699875966352606, "optim/total_tokens": 2230583296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426015, "loss/out": 4.961495399475098, "created_at": "2025-01-15T00:59:37.939868+00:00"} {"global_step": 8510, "acc_step": 0, "speed/wps": 12954.900289312423, "speed/FLOPS": 203474510942154.78, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1285076141357422, "optim/lr": 0.0029699705406923222, "optim/total_tokens": 2230845440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495864, "loss/out": 5.0378217697143555, "created_at": "2025-01-15T00:59:43.002773+00:00"} {"global_step": 8511, "acc_step": 0, "speed/wps": 12955.840350342367, "speed/FLOPS": 203489275892409.25, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0542371273040771, "optim/lr": 0.002969953479953347, "optim/total_tokens": 2231107584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449411, "loss/out": 4.857694149017334, "created_at": "2025-01-15T00:59:48.071529+00:00"} {"global_step": 8512, "acc_step": 0, "speed/wps": 12963.997842078717, "speed/FLOPS": 203617400509696.97, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.143333911895752, "optim/lr": 0.002969936414418389, "optim/total_tokens": 2231369728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428403, "loss/out": 4.886655807495117, "created_at": "2025-01-15T00:59:53.133372+00:00"} {"global_step": 8513, "acc_step": 0, "speed/wps": 12958.210424668021, "speed/FLOPS": 203526501166517.06, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.058598279953003, "optim/lr": 0.002969919344087506, "optim/total_tokens": 2231631872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 4.991092205047607, "created_at": "2025-01-15T00:59:58.199456+00:00"} {"global_step": 8514, "acc_step": 0, "speed/wps": 12955.174462768002, "speed/FLOPS": 203478817213028.44, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0022125244140625, "optim/lr": 0.002969902268960753, "optim/total_tokens": 2231894016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428932, "loss/out": 5.028429985046387, "created_at": "2025-01-15T01:00:03.271027+00:00"} {"global_step": 8515, "acc_step": 0, "speed/wps": 12952.965501561177, "speed/FLOPS": 203444122441844.47, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.118335485458374, "optim/lr": 0.0029698851890381846, "optim/total_tokens": 2232156160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376285, "loss/out": 4.886592864990234, "created_at": "2025-01-15T01:00:08.338494+00:00"} {"global_step": 8516, "acc_step": 0, "speed/wps": 12957.305819793972, "speed/FLOPS": 203512293103912.75, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1800787448883057, "optim/lr": 0.002969868104319858, "optim/total_tokens": 2232418304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462435, "loss/out": 4.854089736938477, "created_at": "2025-01-15T01:00:13.401530+00:00"} {"global_step": 8517, "acc_step": 0, "speed/wps": 12950.951854704794, "speed/FLOPS": 203412495350924.22, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0332063436508179, "optim/lr": 0.002969851014805828, "optim/total_tokens": 2232680448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448890, "loss/out": 5.045477390289307, "created_at": "2025-01-15T01:00:18.466289+00:00"} {"global_step": 8518, "acc_step": 0, "speed/wps": 12949.746183364548, "speed/FLOPS": 203393558625759.62, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.297412395477295, "optim/lr": 0.0029698339204961505, "optim/total_tokens": 2232942592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495289, "loss/out": 4.901298522949219, "created_at": "2025-01-15T01:00:23.535147+00:00"} {"global_step": 8519, "acc_step": 0, "speed/wps": 12955.728600233693, "speed/FLOPS": 203487520703391.78, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2872577905654907, "optim/lr": 0.002969816821390881, "optim/total_tokens": 2233204736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447554, "loss/out": 4.888547420501709, "created_at": "2025-01-15T01:00:28.602116+00:00"} {"global_step": 8520, "acc_step": 0, "speed/wps": 12957.02344207072, "speed/FLOPS": 203507857973741.44, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9907621145248413, "optim/lr": 0.0029697997174900764, "optim/total_tokens": 2233466880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390956, "loss/out": 5.008254051208496, "created_at": "2025-01-15T01:00:33.664964+00:00"} {"global_step": 8521, "acc_step": 0, "speed/wps": 12956.781883799382, "speed/FLOPS": 203504063969154.22, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1361753940582275, "optim/lr": 0.002969782608793792, "optim/total_tokens": 2233729024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427880, "loss/out": 4.956122875213623, "created_at": "2025-01-15T01:00:38.756062+00:00"} {"global_step": 8522, "acc_step": 0, "speed/wps": 12952.811144629579, "speed/FLOPS": 203441698054125.06, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1658611297607422, "optim/lr": 0.002969765495302083, "optim/total_tokens": 2233991168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 4.856686115264893, "created_at": "2025-01-15T01:00:43.822141+00:00"} {"global_step": 8523, "acc_step": 0, "speed/wps": 12951.627873231191, "speed/FLOPS": 203423113150828.1, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.5082752704620361, "optim/lr": 0.0029697483770150055, "optim/total_tokens": 2234253312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437710, "loss/out": 4.967544078826904, "created_at": "2025-01-15T01:00:48.890021+00:00"} {"global_step": 8524, "acc_step": 0, "speed/wps": 12950.709879230502, "speed/FLOPS": 203408694793591.3, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1842454671859741, "optim/lr": 0.002969731253932616, "optim/total_tokens": 2234515456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304572, "loss/out": 4.88768196105957, "created_at": "2025-01-15T01:00:53.955341+00:00"} {"global_step": 8525, "acc_step": 0, "speed/wps": 12949.248617842197, "speed/FLOPS": 203385743675504.4, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2540578842163086, "optim/lr": 0.0029697141260549694, "optim/total_tokens": 2234777600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352687, "loss/out": 4.960437297821045, "created_at": "2025-01-15T01:00:59.028809+00:00"} {"global_step": 8526, "acc_step": 0, "speed/wps": 12955.47248796393, "speed/FLOPS": 203483498108260.03, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1284321546554565, "optim/lr": 0.0029696969933821226, "optim/total_tokens": 2235039744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464007, "loss/out": 4.924432754516602, "created_at": "2025-01-15T01:01:04.090874+00:00"} {"global_step": 8527, "acc_step": 0, "speed/wps": 12954.36610715731, "speed/FLOPS": 203466120877365.5, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2341352701187134, "optim/lr": 0.0029696798559141306, "optim/total_tokens": 2235301888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 4.838652610778809, "created_at": "2025-01-15T01:01:09.154510+00:00"} {"global_step": 8528, "acc_step": 0, "speed/wps": 12949.46215714636, "speed/FLOPS": 203389097603710.03, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1358299255371094, "optim/lr": 0.00296966271365105, "optim/total_tokens": 2235564032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450756, "loss/out": 4.939035415649414, "created_at": "2025-01-15T01:01:14.219045+00:00"} {"global_step": 8529, "acc_step": 0, "speed/wps": 12955.446241160778, "speed/FLOPS": 203483085866149.62, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.304240345954895, "optim/lr": 0.002969645566592936, "optim/total_tokens": 2235826176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 5.01261568069458, "created_at": "2025-01-15T01:01:19.286114+00:00"} {"global_step": 8530, "acc_step": 0, "speed/wps": 12950.22393677029, "speed/FLOPS": 203401062399498.78, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7461462020874023, "optim/lr": 0.002969628414739845, "optim/total_tokens": 2236088320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295474, "loss/out": 4.954343795776367, "created_at": "2025-01-15T01:01:24.354330+00:00"} {"global_step": 8531, "acc_step": 0, "speed/wps": 12954.976256771848, "speed/FLOPS": 203475704115495.25, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2148020267486572, "optim/lr": 0.002969611258091833, "optim/total_tokens": 2236350464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440108, "loss/out": 5.0475969314575195, "created_at": "2025-01-15T01:01:29.422808+00:00"} {"global_step": 8532, "acc_step": 0, "speed/wps": 12950.316439611144, "speed/FLOPS": 203402515283726.62, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.2816373109817505, "optim/lr": 0.002969594096648956, "optim/total_tokens": 2236612608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 5.042959213256836, "created_at": "2025-01-15T01:01:34.491952+00:00"} {"global_step": 8533, "acc_step": 0, "speed/wps": 12950.159762553896, "speed/FLOPS": 203400054455245.7, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.238826036453247, "optim/lr": 0.00296957693041127, "optim/total_tokens": 2236874752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456045, "loss/out": 4.95144510269165, "created_at": "2025-01-15T01:01:39.560205+00:00"} {"global_step": 8534, "acc_step": 0, "speed/wps": 12951.896320340362, "speed/FLOPS": 203427329481562.6, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0412236452102661, "optim/lr": 0.0029695597593788307, "optim/total_tokens": 2237136896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433464, "loss/out": 4.904006481170654, "created_at": "2025-01-15T01:01:44.631488+00:00"} {"global_step": 8535, "acc_step": 0, "speed/wps": 12950.961391280227, "speed/FLOPS": 203412645135946.03, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1393035650253296, "optim/lr": 0.0029695425835516947, "optim/total_tokens": 2237399040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356429, "loss/out": 4.876547813415527, "created_at": "2025-01-15T01:01:49.695588+00:00"} {"global_step": 8536, "acc_step": 0, "speed/wps": 12947.716391411179, "speed/FLOPS": 203361677953905.53, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3685365915298462, "optim/lr": 0.002969525402929917, "optim/total_tokens": 2237661184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462150, "loss/out": 4.917470932006836, "created_at": "2025-01-15T01:01:54.760726+00:00"} {"global_step": 8537, "acc_step": 0, "speed/wps": 12955.29154126716, "speed/FLOPS": 203480656091740.4, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3206413984298706, "optim/lr": 0.002969508217513555, "optim/total_tokens": 2237923328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 5.07136869430542, "created_at": "2025-01-15T01:01:59.824310+00:00"} {"global_step": 8538, "acc_step": 0, "speed/wps": 12951.722062426474, "speed/FLOPS": 203424592521569.53, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9736409783363342, "optim/lr": 0.002969491027302664, "optim/total_tokens": 2238185472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436916, "loss/out": 4.958756446838379, "created_at": "2025-01-15T01:02:04.893189+00:00"} {"global_step": 8539, "acc_step": 0, "speed/wps": 12951.385953541032, "speed/FLOPS": 203419313469661.66, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.331866979598999, "optim/lr": 0.0029694738322973, "optim/total_tokens": 2238447616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452325, "loss/out": 4.9759721755981445, "created_at": "2025-01-15T01:02:09.961053+00:00"} {"global_step": 8540, "acc_step": 0, "speed/wps": 12952.060756275869, "speed/FLOPS": 203429912173888.38, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.5754592418670654, "optim/lr": 0.002969456632497519, "optim/total_tokens": 2238709760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 4.888603687286377, "created_at": "2025-01-15T01:02:15.029570+00:00"} {"global_step": 8541, "acc_step": 0, "speed/wps": 12949.58236620871, "speed/FLOPS": 203390985652218.8, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.143155097961426, "optim/lr": 0.0029694394279033776, "optim/total_tokens": 2238971904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338641, "loss/out": 4.99127721786499, "created_at": "2025-01-15T01:02:20.093995+00:00"} {"global_step": 8542, "acc_step": 0, "speed/wps": 12952.10823051795, "speed/FLOPS": 203430657822096.72, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9463608264923096, "optim/lr": 0.0029694222185149316, "optim/total_tokens": 2239234048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429739, "loss/out": 5.067086219787598, "created_at": "2025-01-15T01:02:25.158749+00:00"} {"global_step": 8543, "acc_step": 0, "speed/wps": 12955.365488304684, "speed/FLOPS": 203481817531578.62, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4207834005355835, "optim/lr": 0.0029694050043322373, "optim/total_tokens": 2239496192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 4.953554153442383, "created_at": "2025-01-15T01:02:30.226251+00:00"} {"global_step": 8544, "acc_step": 0, "speed/wps": 12957.155591862444, "speed/FLOPS": 203509933567813.3, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4394234418869019, "optim/lr": 0.002969387785355351, "optim/total_tokens": 2239758336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441442, "loss/out": 4.98995304107666, "created_at": "2025-01-15T01:02:35.290066+00:00"} {"global_step": 8545, "acc_step": 0, "speed/wps": 12953.84605255442, "speed/FLOPS": 203457952705197.2, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.2773300409317017, "optim/lr": 0.0029693705615843288, "optim/total_tokens": 2240020480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 5.039911270141602, "created_at": "2025-01-15T01:02:40.353906+00:00"} {"global_step": 8546, "acc_step": 0, "speed/wps": 12951.020162732544, "speed/FLOPS": 203413568222365.1, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.553297996520996, "optim/lr": 0.002969353333019227, "optim/total_tokens": 2240282624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430268, "loss/out": 4.934059143066406, "created_at": "2025-01-15T01:02:45.418773+00:00"} {"global_step": 8547, "acc_step": 0, "speed/wps": 12953.101214241788, "speed/FLOPS": 203446253995982.22, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0544886589050293, "optim/lr": 0.002969336099660101, "optim/total_tokens": 2240544768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353744, "loss/out": 4.880912780761719, "created_at": "2025-01-15T01:02:50.482692+00:00"} {"global_step": 8548, "acc_step": 0, "speed/wps": 12951.084290653944, "speed/FLOPS": 203414575439491.7, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8847664594650269, "optim/lr": 0.002969318861507008, "optim/total_tokens": 2240806912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431605, "loss/out": 4.918540954589844, "created_at": "2025-01-15T01:02:55.551002+00:00"} {"global_step": 8549, "acc_step": 0, "speed/wps": 12953.8623657033, "speed/FLOPS": 203458208925617.0, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.4878177642822266, "optim/lr": 0.0029693016185600037, "optim/total_tokens": 2241069056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445976, "loss/out": 4.994950771331787, "created_at": "2025-01-15T01:03:00.620412+00:00"} {"global_step": 8550, "acc_step": 0, "speed/wps": 12948.011047836151, "speed/FLOPS": 203366305937960.06, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8481382131576538, "optim/lr": 0.002969284370819145, "optim/total_tokens": 2241331200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 5.032268524169922, "created_at": "2025-01-15T01:03:05.686543+00:00"} {"global_step": 8551, "acc_step": 0, "speed/wps": 12957.241654104138, "speed/FLOPS": 203511285293581.03, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2991567850112915, "optim/lr": 0.0029692671182844875, "optim/total_tokens": 2241593344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452325, "loss/out": 5.085630416870117, "created_at": "2025-01-15T01:03:10.748425+00:00"} {"global_step": 8552, "acc_step": 0, "speed/wps": 12953.742387161525, "speed/FLOPS": 203456324497750.6, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6075619459152222, "optim/lr": 0.0029692498609560874, "optim/total_tokens": 2241855488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 269784, "loss/out": 4.791617393493652, "created_at": "2025-01-15T01:03:15.819706+00:00"} {"global_step": 8553, "acc_step": 0, "speed/wps": 12954.76222791476, "speed/FLOPS": 203472342498188.22, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5807535648345947, "optim/lr": 0.0029692325988340015, "optim/total_tokens": 2242117632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370061, "loss/out": 5.010385513305664, "created_at": "2025-01-15T01:03:20.883594+00:00"} {"global_step": 8554, "acc_step": 0, "speed/wps": 12952.713604104058, "speed/FLOPS": 203440166046137.03, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1685287952423096, "optim/lr": 0.0029692153319182857, "optim/total_tokens": 2242379776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442778, "loss/out": 4.943125247955322, "created_at": "2025-01-15T01:03:25.948014+00:00"} {"global_step": 8555, "acc_step": 0, "speed/wps": 12949.468798564934, "speed/FLOPS": 203389201916315.03, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.375544548034668, "optim/lr": 0.0029691980602089972, "optim/total_tokens": 2242641920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 4.825962066650391, "created_at": "2025-01-15T01:03:31.013137+00:00"} {"global_step": 8556, "acc_step": 0, "speed/wps": 12957.299454254291, "speed/FLOPS": 203512193124364.9, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.777571201324463, "optim/lr": 0.002969180783706191, "optim/total_tokens": 2242904064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 4.958645343780518, "created_at": "2025-01-15T01:03:36.075025+00:00"} {"global_step": 8557, "acc_step": 0, "speed/wps": 12956.085134398229, "speed/FLOPS": 203493120562375.1, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.147233009338379, "optim/lr": 0.002969163502409925, "optim/total_tokens": 2243166208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458430, "loss/out": 4.890944004058838, "created_at": "2025-01-15T01:03:41.143603+00:00"} {"global_step": 8558, "acc_step": 0, "speed/wps": 12958.998337579824, "speed/FLOPS": 203538876421502.72, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.130471706390381, "optim/lr": 0.0029691462163202544, "optim/total_tokens": 2243428352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314942, "loss/out": 5.002866744995117, "created_at": "2025-01-15T01:03:46.208380+00:00"} {"global_step": 8559, "acc_step": 0, "speed/wps": 12960.081401987993, "speed/FLOPS": 203555887436319.47, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7126247882843018, "optim/lr": 0.002969128925437236, "optim/total_tokens": 2243690496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 405668, "loss/out": 5.024083137512207, "created_at": "2025-01-15T01:03:51.268672+00:00"} {"global_step": 8560, "acc_step": 0, "speed/wps": 12956.244135206689, "speed/FLOPS": 203495617888561.97, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.35668420791626, "optim/lr": 0.0029691116297609265, "optim/total_tokens": 2243952640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457902, "loss/out": 5.032317638397217, "created_at": "2025-01-15T01:03:56.331201+00:00"} {"global_step": 8561, "acc_step": 0, "speed/wps": 12954.192395657263, "speed/FLOPS": 203463392499552.66, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.001, "optim/grad_norm": 2.3242902755737305, "optim/lr": 0.0029690943292913813, "optim/total_tokens": 2244214784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 4.965127944946289, "created_at": "2025-01-15T01:04:01.395053+00:00"} {"global_step": 8562, "acc_step": 0, "speed/wps": 12959.596531727751, "speed/FLOPS": 203548271882600.84, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.9818196296691895, "optim/lr": 0.002969077024028658, "optim/total_tokens": 2244476928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461913, "loss/out": 4.983104705810547, "created_at": "2025-01-15T01:04:06.463722+00:00"} {"global_step": 8563, "acc_step": 0, "speed/wps": 12946.987481716538, "speed/FLOPS": 203350229425524.3, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7868462800979614, "optim/lr": 0.0029690597139728128, "optim/total_tokens": 2244739072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433718, "loss/out": 5.04454231262207, "created_at": "2025-01-15T01:04:11.533941+00:00"} {"global_step": 8564, "acc_step": 0, "speed/wps": 12955.561504631365, "speed/FLOPS": 203484896237343.8, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2515090703964233, "optim/lr": 0.0029690423991239018, "optim/total_tokens": 2245001216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292485, "loss/out": 4.940745830535889, "created_at": "2025-01-15T01:04:16.598203+00:00"} {"global_step": 8565, "acc_step": 0, "speed/wps": 12953.01805862321, "speed/FLOPS": 203444947922724.7, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3131043910980225, "optim/lr": 0.0029690250794819815, "optim/total_tokens": 2245263360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348373, "loss/out": 4.922650337219238, "created_at": "2025-01-15T01:04:21.661518+00:00"} {"global_step": 8566, "acc_step": 0, "speed/wps": 12952.308863931588, "speed/FLOPS": 203433809045557.12, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9868147373199463, "optim/lr": 0.002969007755047109, "optim/total_tokens": 2245525504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 395218, "loss/out": 4.857701301574707, "created_at": "2025-01-15T01:04:26.728583+00:00"} {"global_step": 8567, "acc_step": 0, "speed/wps": 12950.685551329083, "speed/FLOPS": 203408312690469.5, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.197797179222107, "optim/lr": 0.00296899042581934, "optim/total_tokens": 2245787648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445407, "loss/out": 4.945621490478516, "created_at": "2025-01-15T01:04:31.793248+00:00"} {"global_step": 8568, "acc_step": 0, "speed/wps": 12954.79650926206, "speed/FLOPS": 203472880933855.06, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1365631818771362, "optim/lr": 0.002968973091798732, "optim/total_tokens": 2246049792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 5.122166633605957, "created_at": "2025-01-15T01:04:36.861511+00:00"} {"global_step": 8569, "acc_step": 0, "speed/wps": 12957.430462043088, "speed/FLOPS": 203514250781711.53, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2012500762939453, "optim/lr": 0.002968955752985341, "optim/total_tokens": 2246311936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467734, "loss/out": 5.046411991119385, "created_at": "2025-01-15T01:04:41.925853+00:00"} {"global_step": 8570, "acc_step": 0, "speed/wps": 12957.11027317467, "speed/FLOPS": 203509221775547.4, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.12623131275177, "optim/lr": 0.002968938409379223, "optim/total_tokens": 2246574080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 4.9848222732543945, "created_at": "2025-01-15T01:04:46.992618+00:00"} {"global_step": 8571, "acc_step": 0, "speed/wps": 12950.928093176157, "speed/FLOPS": 203412122143465.22, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8942044377326965, "optim/lr": 0.0029689210609804358, "optim/total_tokens": 2246836224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 272116, "loss/out": 4.979527473449707, "created_at": "2025-01-15T01:04:52.060433+00:00"} {"global_step": 8572, "acc_step": 0, "speed/wps": 12957.947337460238, "speed/FLOPS": 203522369020399.12, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1539990901947021, "optim/lr": 0.002968903707789035, "optim/total_tokens": 2247098368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448368, "loss/out": 4.9361252784729, "created_at": "2025-01-15T01:04:57.146522+00:00"} {"global_step": 8573, "acc_step": 0, "speed/wps": 12962.667311439563, "speed/FLOPS": 203596502697668.34, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.141045093536377, "optim/lr": 0.0029688863498050773, "optim/total_tokens": 2247360512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425201, "loss/out": 4.984711170196533, "created_at": "2025-01-15T01:05:02.207648+00:00"} {"global_step": 8574, "acc_step": 0, "speed/wps": 12954.944934277724, "speed/FLOPS": 203475212152683.44, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.172813057899475, "optim/lr": 0.0029688689870286207, "optim/total_tokens": 2247622656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 502417, "loss/out": 4.845998287200928, "created_at": "2025-01-15T01:05:07.274915+00:00"} {"global_step": 8575, "acc_step": 0, "speed/wps": 12948.055159304857, "speed/FLOPS": 203366998769191.6, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.0546505451202393, "optim/lr": 0.0029688516194597203, "optim/total_tokens": 2247884800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318851, "loss/out": 4.841174602508545, "created_at": "2025-01-15T01:05:12.339859+00:00"} {"global_step": 8576, "acc_step": 0, "speed/wps": 12954.97228570181, "speed/FLOPS": 203475641744383.47, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2485382556915283, "optim/lr": 0.0029688342470984325, "optim/total_tokens": 2248146944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375475, "loss/out": 4.977828025817871, "created_at": "2025-01-15T01:05:17.405536+00:00"} {"global_step": 8577, "acc_step": 0, "speed/wps": 12954.775779209018, "speed/FLOPS": 203472555339885.97, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4118117094039917, "optim/lr": 0.002968816869944816, "optim/total_tokens": 2248409088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428403, "loss/out": 5.007554054260254, "created_at": "2025-01-15T01:05:22.468718+00:00"} {"global_step": 8578, "acc_step": 0, "speed/wps": 12950.102750586777, "speed/FLOPS": 203399159003959.44, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3523975610733032, "optim/lr": 0.0029687994879989254, "optim/total_tokens": 2248671232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 4.951615810394287, "created_at": "2025-01-15T01:05:27.538804+00:00"} {"global_step": 8579, "acc_step": 0, "speed/wps": 12950.652886253969, "speed/FLOPS": 203407799640578.3, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4384210109710693, "optim/lr": 0.0029687821012608184, "optim/total_tokens": 2248933376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 4.953673362731934, "created_at": "2025-01-15T01:05:32.607504+00:00"} {"global_step": 8580, "acc_step": 0, "speed/wps": 12955.52168311727, "speed/FLOPS": 203484270785744.44, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.424019455909729, "optim/lr": 0.0029687647097305517, "optim/total_tokens": 2249195520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456851, "loss/out": 4.90565299987793, "created_at": "2025-01-15T01:05:37.673380+00:00"} {"global_step": 8581, "acc_step": 0, "speed/wps": 12949.791074020895, "speed/FLOPS": 203394263695203.78, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9939783811569214, "optim/lr": 0.0029687473134081817, "optim/total_tokens": 2249457664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441971, "loss/out": 5.025237560272217, "created_at": "2025-01-15T01:05:42.745769+00:00"} {"global_step": 8582, "acc_step": 0, "speed/wps": 12950.964268225387, "speed/FLOPS": 203412690322323.1, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.239884614944458, "optim/lr": 0.0029687299122937655, "optim/total_tokens": 2249719808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 342518, "loss/out": 4.961688041687012, "created_at": "2025-01-15T01:05:47.811590+00:00"} {"global_step": 8583, "acc_step": 0, "speed/wps": 12959.623363393892, "speed/FLOPS": 203548693310788.12, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8999916315078735, "optim/lr": 0.0029687125063873596, "optim/total_tokens": 2249981952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 4.86159610748291, "created_at": "2025-01-15T01:05:52.876481+00:00"} {"global_step": 8584, "acc_step": 0, "speed/wps": 12952.97933044329, "speed/FLOPS": 203444339643440.9, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.5936098098754883, "optim/lr": 0.0029686950956890213, "optim/total_tokens": 2250244096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 4.894453048706055, "created_at": "2025-01-15T01:05:57.945287+00:00"} {"global_step": 8585, "acc_step": 0, "speed/wps": 12954.45405452414, "speed/FLOPS": 203467502211613.9, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.033735752105713, "optim/lr": 0.0029686776801988064, "optim/total_tokens": 2250506240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 4.957857131958008, "created_at": "2025-01-15T01:06:03.007737+00:00"} {"global_step": 8586, "acc_step": 0, "speed/wps": 12955.179523953795, "speed/FLOPS": 203478896705906.12, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.879447877407074, "optim/lr": 0.0029686602599167736, "optim/total_tokens": 2250768384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 4.762533187866211, "created_at": "2025-01-15T01:06:08.070099+00:00"} {"global_step": 8587, "acc_step": 0, "speed/wps": 12957.117719303185, "speed/FLOPS": 203509338727227.8, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.238703966140747, "optim/lr": 0.002968642834842977, "optim/total_tokens": 2251030528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325819, "loss/out": 5.01396369934082, "created_at": "2025-01-15T01:06:13.131902+00:00"} {"global_step": 8588, "acc_step": 0, "speed/wps": 12947.784029217508, "speed/FLOPS": 203362740298598.0, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1184806823730469, "optim/lr": 0.0029686254049774764, "optim/total_tokens": 2251292672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366652, "loss/out": 5.014610767364502, "created_at": "2025-01-15T01:06:18.198186+00:00"} {"global_step": 8589, "acc_step": 0, "speed/wps": 12944.686222888855, "speed/FLOPS": 203314084993374.53, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.3020851612091064, "optim/lr": 0.002968607970320326, "optim/total_tokens": 2251554816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466392, "loss/out": 4.968874931335449, "created_at": "2025-01-15T01:06:23.264480+00:00"} {"global_step": 8590, "acc_step": 0, "speed/wps": 12948.586926875985, "speed/FLOPS": 203375350909621.4, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2328336238861084, "optim/lr": 0.0029685905308715847, "optim/total_tokens": 2251816960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452325, "loss/out": 5.022782802581787, "created_at": "2025-01-15T01:06:28.336315+00:00"} {"global_step": 8591, "acc_step": 0, "speed/wps": 12949.41773821772, "speed/FLOPS": 203388399943397.75, "speed/curr_iter_time": 1.2771, "speed/data_load_time": 0.001, "optim/grad_norm": 1.4574576616287231, "optim/lr": 0.002968573086631308, "optim/total_tokens": 2252079104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 4.922252178192139, "created_at": "2025-01-15T01:06:33.416486+00:00"} {"global_step": 8592, "acc_step": 0, "speed/wps": 12957.33932010043, "speed/FLOPS": 203512819272259.9, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2842333316802979, "optim/lr": 0.0029685556375995538, "optim/total_tokens": 2252341248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447839, "loss/out": 4.95196533203125, "created_at": "2025-01-15T01:06:38.479744+00:00"} {"global_step": 8593, "acc_step": 0, "speed/wps": 12955.07717046838, "speed/FLOPS": 203477289103769.22, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.587899088859558, "optim/lr": 0.0029685381837763784, "optim/total_tokens": 2252603392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316449, "loss/out": 4.979541778564453, "created_at": "2025-01-15T01:06:43.545318+00:00"} {"global_step": 8594, "acc_step": 0, "speed/wps": 12950.539225947387, "speed/FLOPS": 203406014449277.8, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0691496133804321, "optim/lr": 0.002968520725161839, "optim/total_tokens": 2252865536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423841, "loss/out": 4.988583564758301, "created_at": "2025-01-15T01:06:48.625099+00:00"} {"global_step": 8595, "acc_step": 0, "speed/wps": 12956.532538386391, "speed/FLOPS": 203500147656806.7, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.046898126602173, "optim/lr": 0.0029685032617559927, "optim/total_tokens": 2253127680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453131, "loss/out": 4.944446086883545, "created_at": "2025-01-15T01:06:53.690348+00:00"} {"global_step": 8596, "acc_step": 0, "speed/wps": 12956.229099451419, "speed/FLOPS": 203495381731364.16, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.205376386642456, "optim/lr": 0.002968485793558897, "optim/total_tokens": 2253389824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452847, "loss/out": 4.980544090270996, "created_at": "2025-01-15T01:06:58.756714+00:00"} {"global_step": 8597, "acc_step": 0, "speed/wps": 12952.401730331201, "speed/FLOPS": 203435267639974.84, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2709708213806152, "optim/lr": 0.0029684683205706067, "optim/total_tokens": 2253651968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452093, "loss/out": 4.928732395172119, "created_at": "2025-01-15T01:07:03.824355+00:00"} {"global_step": 8598, "acc_step": 0, "speed/wps": 12955.458558349595, "speed/FLOPS": 203483279324525.6, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3452972173690796, "optim/lr": 0.0029684508427911815, "optim/total_tokens": 2253914112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305105, "loss/out": 4.945113658905029, "created_at": "2025-01-15T01:07:08.887083+00:00"} {"global_step": 8599, "acc_step": 0, "speed/wps": 12951.249945472047, "speed/FLOPS": 203417177276043.56, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6230477094650269, "optim/lr": 0.002968433360220677, "optim/total_tokens": 2254176256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453382, "loss/out": 4.996060371398926, "created_at": "2025-01-15T01:07:13.950815+00:00"} {"global_step": 8600, "acc_step": 0, "speed/wps": 12954.793892959715, "speed/FLOPS": 203472839841231.56, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3319286108016968, "optim/lr": 0.0029684158728591507, "optim/total_tokens": 2254438400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 4.9287943840026855, "created_at": "2025-01-15T01:07:19.013235+00:00"} {"global_step": 8601, "acc_step": 0, "speed/wps": 12957.24394214618, "speed/FLOPS": 203511321230425.4, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2421742677688599, "optim/lr": 0.002968398380706659, "optim/total_tokens": 2254700544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448077, "loss/out": 4.776882171630859, "created_at": "2025-01-15T01:07:24.075195+00:00"} {"global_step": 8602, "acc_step": 0, "speed/wps": 12956.111951526265, "speed/FLOPS": 203493541762221.5, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1857399940490723, "optim/lr": 0.00296838088376326, "optim/total_tokens": 2254962688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 4.841368675231934, "created_at": "2025-01-15T01:07:29.141220+00:00"} {"global_step": 8603, "acc_step": 0, "speed/wps": 12958.266694336924, "speed/FLOPS": 203527384958989.8, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0637619495391846, "optim/lr": 0.00296836338202901, "optim/total_tokens": 2255224832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 4.9208879470825195, "created_at": "2025-01-15T01:07:34.208479+00:00"} {"global_step": 8604, "acc_step": 0, "speed/wps": 12961.345299867065, "speed/FLOPS": 203575738689288.0, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.071670413017273, "optim/lr": 0.002968345875503966, "optim/total_tokens": 2255486976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306979, "loss/out": 5.004879951477051, "created_at": "2025-01-15T01:07:39.273336+00:00"} {"global_step": 8605, "acc_step": 0, "speed/wps": 12945.232211969273, "speed/FLOPS": 203322660502150.12, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0011, "optim/grad_norm": 1.6276533603668213, "optim/lr": 0.0029683283641881854, "optim/total_tokens": 2255749120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 347078, "loss/out": 4.9942626953125, "created_at": "2025-01-15T01:07:44.342048+00:00"} {"global_step": 8606, "acc_step": 0, "speed/wps": 12950.773346238495, "speed/FLOPS": 203409691630164.66, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1951520442962646, "optim/lr": 0.0029683108480817256, "optim/total_tokens": 2256011264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431605, "loss/out": 4.844259262084961, "created_at": "2025-01-15T01:07:49.407297+00:00"} {"global_step": 8607, "acc_step": 0, "speed/wps": 12950.5548157485, "speed/FLOPS": 203406259308526.9, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.650342345237732, "optim/lr": 0.0029682933271846434, "optim/total_tokens": 2256273408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 4.779524803161621, "created_at": "2025-01-15T01:07:54.472123+00:00"} {"global_step": 8608, "acc_step": 0, "speed/wps": 12953.01736717427, "speed/FLOPS": 203444937062568.9, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1793230772018433, "optim/lr": 0.002968275801496996, "optim/total_tokens": 2256535552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445404, "loss/out": 4.966610908508301, "created_at": "2025-01-15T01:07:59.537652+00:00"} {"global_step": 8609, "acc_step": 0, "speed/wps": 12958.728522109079, "speed/FLOPS": 203534638598769.38, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5208792686462402, "optim/lr": 0.0029682582710188416, "optim/total_tokens": 2256797696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 4.8749494552612305, "created_at": "2025-01-15T01:08:04.604605+00:00"} {"global_step": 8610, "acc_step": 0, "speed/wps": 12958.948330864214, "speed/FLOPS": 203538090997317.66, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1071968078613281, "optim/lr": 0.002968240735750235, "optim/total_tokens": 2257059840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328460, "loss/out": 4.927866458892822, "created_at": "2025-01-15T01:08:09.665258+00:00"} {"global_step": 8611, "acc_step": 0, "speed/wps": 12948.309801402878, "speed/FLOPS": 203370998273255.94, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.001, "optim/grad_norm": 1.1859594583511353, "optim/lr": 0.0029682231956912357, "optim/total_tokens": 2257321984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350002, "loss/out": 4.873734474182129, "created_at": "2025-01-15T01:08:14.736044+00:00"} {"global_step": 8612, "acc_step": 0, "speed/wps": 12954.198841610449, "speed/FLOPS": 203463493742105.1, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.347361445426941, "optim/lr": 0.0029682056508419, "optim/total_tokens": 2257584128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453945, "loss/out": 4.8506879806518555, "created_at": "2025-01-15T01:08:19.798759+00:00"} {"global_step": 8613, "acc_step": 0, "speed/wps": 12952.114180914507, "speed/FLOPS": 203430751281251.34, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.218454360961914, "optim/lr": 0.0029681881012022853, "optim/total_tokens": 2257846272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447555, "loss/out": 4.836045742034912, "created_at": "2025-01-15T01:08:24.864583+00:00"} {"global_step": 8614, "acc_step": 0, "speed/wps": 12948.51930927273, "speed/FLOPS": 203374288882245.94, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.814380645751953, "optim/lr": 0.0029681705467724485, "optim/total_tokens": 2258108416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 4.772692680358887, "created_at": "2025-01-15T01:08:29.931920+00:00"} {"global_step": 8615, "acc_step": 0, "speed/wps": 12953.342380157452, "speed/FLOPS": 203450041838084.4, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1894445419311523, "optim/lr": 0.0029681529875524477, "optim/total_tokens": 2258370560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435044, "loss/out": 4.908351421356201, "created_at": "2025-01-15T01:08:34.998177+00:00"} {"global_step": 8616, "acc_step": 0, "speed/wps": 12955.688725216005, "speed/FLOPS": 203486894411444.97, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.711334705352783, "optim/lr": 0.0029681354235423394, "optim/total_tokens": 2258632704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351876, "loss/out": 4.766841888427734, "created_at": "2025-01-15T01:08:40.082599+00:00"} {"global_step": 8617, "acc_step": 0, "speed/wps": 12953.765505642761, "speed/FLOPS": 203456687605266.56, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2492598295211792, "optim/lr": 0.0029681178547421813, "optim/total_tokens": 2258894848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 5.002995491027832, "created_at": "2025-01-15T01:08:45.147862+00:00"} {"global_step": 8618, "acc_step": 0, "speed/wps": 12956.455472009562, "speed/FLOPS": 203498937223458.3, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.890960454940796, "optim/lr": 0.0029681002811520305, "optim/total_tokens": 2259156992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456851, "loss/out": 4.842422962188721, "created_at": "2025-01-15T01:08:50.222155+00:00"} {"global_step": 8619, "acc_step": 0, "speed/wps": 12952.935984149517, "speed/FLOPS": 203443658830333.16, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5120924711227417, "optim/lr": 0.002968082702771944, "optim/total_tokens": 2259419136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 5.133450031280518, "created_at": "2025-01-15T01:08:55.285424+00:00"} {"global_step": 8620, "acc_step": 0, "speed/wps": 12951.511967001388, "speed/FLOPS": 203421292684217.2, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1670172214508057, "optim/lr": 0.00296806511960198, "optim/total_tokens": 2259681280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446218, "loss/out": 4.869524955749512, "created_at": "2025-01-15T01:09:00.350068+00:00"} {"global_step": 8621, "acc_step": 0, "speed/wps": 12961.174769637397, "speed/FLOPS": 203573060277697.97, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1287353038787842, "optim/lr": 0.0029680475316421952, "optim/total_tokens": 2259943424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297534, "loss/out": 4.935268402099609, "created_at": "2025-01-15T01:09:05.416012+00:00"} {"global_step": 8622, "acc_step": 0, "speed/wps": 12953.802881688882, "speed/FLOPS": 203457274647430.94, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5601818561553955, "optim/lr": 0.002968029938892648, "optim/total_tokens": 2260205568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 404906, "loss/out": 4.901727676391602, "created_at": "2025-01-15T01:09:10.482332+00:00"} {"global_step": 8623, "acc_step": 0, "speed/wps": 12947.79550991845, "speed/FLOPS": 203362920618782.44, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.208954334259033, "optim/lr": 0.0029680123413533942, "optim/total_tokens": 2260467712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 4.945042610168457, "created_at": "2025-01-15T01:09:15.553532+00:00"} {"global_step": 8624, "acc_step": 0, "speed/wps": 12941.237334188298, "speed/FLOPS": 203259915457062.56, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0013, "optim/grad_norm": 2.1398744583129883, "optim/lr": 0.0029679947390244925, "optim/total_tokens": 2260729856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464007, "loss/out": 4.96718692779541, "created_at": "2025-01-15T01:09:20.622349+00:00"} {"global_step": 8625, "acc_step": 0, "speed/wps": 12952.195813469254, "speed/FLOPS": 203432033432698.53, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0996044874191284, "optim/lr": 0.0029679771319059997, "optim/total_tokens": 2260992000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436096, "loss/out": 4.784233093261719, "created_at": "2025-01-15T01:09:25.691043+00:00"} {"global_step": 8626, "acc_step": 0, "speed/wps": 12956.021703577691, "speed/FLOPS": 203492124294175.47, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3293449878692627, "optim/lr": 0.002967959519997973, "optim/total_tokens": 2261254144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349355, "loss/out": 4.998736381530762, "created_at": "2025-01-15T01:09:30.753110+00:00"} {"global_step": 8627, "acc_step": 0, "speed/wps": 12956.24868196107, "speed/FLOPS": 203495689301587.4, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1130650043487549, "optim/lr": 0.0029679419033004713, "optim/total_tokens": 2261516288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433464, "loss/out": 4.975811004638672, "created_at": "2025-01-15T01:09:35.819193+00:00"} {"global_step": 8628, "acc_step": 0, "speed/wps": 12953.699836971953, "speed/FLOPS": 203455656188553.3, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5149545669555664, "optim/lr": 0.0029679242818135506, "optim/total_tokens": 2261778432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 4.88680362701416, "created_at": "2025-01-15T01:09:40.883375+00:00"} {"global_step": 8629, "acc_step": 0, "speed/wps": 12950.070908279853, "speed/FLOPS": 203398658876773.28, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3321460485458374, "optim/lr": 0.0029679066555372688, "optim/total_tokens": 2262040576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464535, "loss/out": 4.901033401489258, "created_at": "2025-01-15T01:09:45.957571+00:00"} {"global_step": 8630, "acc_step": 0, "speed/wps": 12955.174288655648, "speed/FLOPS": 203478814478354.66, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.086158037185669, "optim/lr": 0.002967889024471683, "optim/total_tokens": 2262302720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434230, "loss/out": 4.964982032775879, "created_at": "2025-01-15T01:09:51.019859+00:00"} {"global_step": 8631, "acc_step": 0, "speed/wps": 12949.96735811363, "speed/FLOPS": 203397032479119.9, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3129481077194214, "optim/lr": 0.0029678713886168518, "optim/total_tokens": 2262564864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289017, "loss/out": 4.948559761047363, "created_at": "2025-01-15T01:09:56.101462+00:00"} {"global_step": 8632, "acc_step": 0, "speed/wps": 12942.92726814831, "speed/FLOPS": 203286458192117.2, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.240026593208313, "optim/lr": 0.002967853747972832, "optim/total_tokens": 2262827008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 414801, "loss/out": 4.806821346282959, "created_at": "2025-01-15T01:10:01.170930+00:00"} {"global_step": 8633, "acc_step": 0, "speed/wps": 12951.950897436733, "speed/FLOPS": 203428186689857.9, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4015467166900635, "optim/lr": 0.002967836102539681, "optim/total_tokens": 2263089152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455808, "loss/out": 4.976924896240234, "created_at": "2025-01-15T01:10:06.237336+00:00"} {"global_step": 8634, "acc_step": 0, "speed/wps": 12953.56603604542, "speed/FLOPS": 203453554661140.38, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.194871425628662, "optim/lr": 0.002967818452317457, "optim/total_tokens": 2263351296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449696, "loss/out": 4.879273414611816, "created_at": "2025-01-15T01:10:11.303682+00:00"} {"global_step": 8635, "acc_step": 0, "speed/wps": 12950.582863661308, "speed/FLOPS": 203406699839539.44, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.2715048789978027, "optim/lr": 0.002967800797306217, "optim/total_tokens": 2263613440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 4.955532550811768, "created_at": "2025-01-15T01:10:16.369715+00:00"} {"global_step": 8636, "acc_step": 0, "speed/wps": 12951.954692628879, "speed/FLOPS": 203428246298565.7, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6013401746749878, "optim/lr": 0.0029677831375060195, "optim/total_tokens": 2263875584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424963, "loss/out": 4.939630508422852, "created_at": "2025-01-15T01:10:21.436911+00:00"} {"global_step": 8637, "acc_step": 0, "speed/wps": 12953.317245969416, "speed/FLOPS": 203449647071123.3, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1416188478469849, "optim/lr": 0.0029677654729169207, "optim/total_tokens": 2264137728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348659, "loss/out": 4.946040153503418, "created_at": "2025-01-15T01:10:26.505772+00:00"} {"global_step": 8638, "acc_step": 0, "speed/wps": 12953.386809323632, "speed/FLOPS": 203450739659191.34, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1699144840240479, "optim/lr": 0.002967747803538979, "optim/total_tokens": 2264399872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 4.960777282714844, "created_at": "2025-01-15T01:10:31.568824+00:00"} {"global_step": 8639, "acc_step": 0, "speed/wps": 12951.876348108268, "speed/FLOPS": 203427015790212.84, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.378128170967102, "optim/lr": 0.0029677301293722525, "optim/total_tokens": 2264662016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457380, "loss/out": 4.891231060028076, "created_at": "2025-01-15T01:10:36.633948+00:00"} {"global_step": 8640, "acc_step": 0, "speed/wps": 12957.4556432132, "speed/FLOPS": 203514646286590.7, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5939210653305054, "optim/lr": 0.0029677124504167984, "optim/total_tokens": 2264924160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458430, "loss/out": 4.962185382843018, "created_at": "2025-01-15T01:10:41.701795+00:00"} {"global_step": 8641, "acc_step": 0, "speed/wps": 12955.300783706254, "speed/FLOPS": 203480801256946.84, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.064028263092041, "optim/lr": 0.0029676947666726745, "optim/total_tokens": 2265186304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445691, "loss/out": 5.031210422515869, "created_at": "2025-01-15T01:10:46.769798+00:00"} {"global_step": 8642, "acc_step": 0, "speed/wps": 12954.319598893677, "speed/FLOPS": 203465390401176.1, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1159430742263794, "optim/lr": 0.0029676770781399383, "optim/total_tokens": 2265448448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 343860, "loss/out": 4.936817169189453, "created_at": "2025-01-15T01:10:51.836865+00:00"} {"global_step": 8643, "acc_step": 0, "speed/wps": 12949.25200873727, "speed/FLOPS": 203385796934171.12, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4807652235031128, "optim/lr": 0.0029676593848186467, "optim/total_tokens": 2265710592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437195, "loss/out": 4.725861549377441, "created_at": "2025-01-15T01:10:56.901598+00:00"} {"global_step": 8644, "acc_step": 0, "speed/wps": 12948.982415483302, "speed/FLOPS": 203381562601657.6, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1681151390075684, "optim/lr": 0.0029676416867088595, "optim/total_tokens": 2265972736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 4.875452518463135, "created_at": "2025-01-15T01:11:01.966334+00:00"} {"global_step": 8645, "acc_step": 0, "speed/wps": 12949.40869004306, "speed/FLOPS": 203388257829381.22, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9975676536560059, "optim/lr": 0.0029676239838106327, "optim/total_tokens": 2266234880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 4.918639183044434, "created_at": "2025-01-15T01:11:07.033756+00:00"} {"global_step": 8646, "acc_step": 0, "speed/wps": 12950.284960898362, "speed/FLOPS": 203402020867286.06, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.4183619022369385, "optim/lr": 0.002967606276124025, "optim/total_tokens": 2266497024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425492, "loss/out": 5.026066303253174, "created_at": "2025-01-15T01:11:12.099845+00:00"} {"global_step": 8647, "acc_step": 0, "speed/wps": 12949.971764622065, "speed/FLOPS": 203397101689390.06, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3140711784362793, "optim/lr": 0.0029675885636490937, "optim/total_tokens": 2266759168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 4.98085880279541, "created_at": "2025-01-15T01:11:17.168731+00:00"} {"global_step": 8648, "acc_step": 0, "speed/wps": 12951.920199849343, "speed/FLOPS": 203427704542065.03, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2146559953689575, "optim/lr": 0.0029675708463858964, "optim/total_tokens": 2267021312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308260, "loss/out": 4.845878601074219, "created_at": "2025-01-15T01:11:22.235029+00:00"} {"global_step": 8649, "acc_step": 0, "speed/wps": 12947.795793422216, "speed/FLOPS": 203362925071598.66, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.001, "optim/grad_norm": 1.4945483207702637, "optim/lr": 0.0029675531243344914, "optim/total_tokens": 2267283456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449411, "loss/out": 4.941889762878418, "created_at": "2025-01-15T01:11:27.302990+00:00"} {"global_step": 8650, "acc_step": 0, "speed/wps": 12951.365570503056, "speed/FLOPS": 203418993326041.0, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.748353123664856, "optim/lr": 0.0029675353974949366, "optim/total_tokens": 2267545600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 4.81148624420166, "created_at": "2025-01-15T01:11:32.372037+00:00"} {"global_step": 8651, "acc_step": 0, "speed/wps": 12953.902901945778, "speed/FLOPS": 203458845603007.72, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.3035879135131836, "optim/lr": 0.002967517665867289, "optim/total_tokens": 2267807744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422813, "loss/out": 4.94075870513916, "created_at": "2025-01-15T01:11:37.438459+00:00"} {"global_step": 8652, "acc_step": 0, "speed/wps": 12949.172898640762, "speed/FLOPS": 203384554401396.78, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4877996444702148, "optim/lr": 0.0029674999294516074, "optim/total_tokens": 2268069888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 4.768135070800781, "created_at": "2025-01-15T01:11:42.504145+00:00"} {"global_step": 8653, "acc_step": 0, "speed/wps": 12945.794668553934, "speed/FLOPS": 203331494655708.56, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1097939014434814, "optim/lr": 0.00296748218824795, "optim/total_tokens": 2268332032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432941, "loss/out": 4.839452266693115, "created_at": "2025-01-15T01:11:47.572576+00:00"} {"global_step": 8654, "acc_step": 0, "speed/wps": 12959.761779036624, "speed/FLOPS": 203550867318660.28, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.806466579437256, "optim/lr": 0.0029674644422563725, "optim/total_tokens": 2268594176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330329, "loss/out": 4.873321533203125, "created_at": "2025-01-15T01:11:52.639009+00:00"} {"global_step": 8655, "acc_step": 0, "speed/wps": 12952.916771677115, "speed/FLOPS": 203443357072053.4, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8944776058197021, "optim/lr": 0.0029674466914769346, "optim/total_tokens": 2268856320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 4.940075874328613, "created_at": "2025-01-15T01:11:57.705981+00:00"} {"global_step": 8656, "acc_step": 0, "speed/wps": 12948.680611983349, "speed/FLOPS": 203376822362969.38, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.13905930519104, "optim/lr": 0.002967428935909695, "optim/total_tokens": 2269118464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450225, "loss/out": 4.942814826965332, "created_at": "2025-01-15T01:12:02.776012+00:00"} {"global_step": 8657, "acc_step": 0, "speed/wps": 12953.527171056963, "speed/FLOPS": 203452944233090.56, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.4622366428375244, "optim/lr": 0.0029674111755547093, "optim/total_tokens": 2269380608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 4.92655086517334, "created_at": "2025-01-15T01:12:07.844421+00:00"} {"global_step": 8658, "acc_step": 0, "speed/wps": 12950.989962953836, "speed/FLOPS": 203413093893341.66, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.4533920288085938, "optim/lr": 0.002967393410412037, "optim/total_tokens": 2269642752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 4.911921501159668, "created_at": "2025-01-15T01:12:12.918329+00:00"} {"global_step": 8659, "acc_step": 0, "speed/wps": 12953.970704558978, "speed/FLOPS": 203459910536218.6, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4803587198257446, "optim/lr": 0.002967375640481736, "optim/total_tokens": 2269904896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315630, "loss/out": 4.957001686096191, "created_at": "2025-01-15T01:12:17.984862+00:00"} {"global_step": 8660, "acc_step": 0, "speed/wps": 12952.105846546505, "speed/FLOPS": 203430620378549.25, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.843446969985962, "optim/lr": 0.002967357865763864, "optim/total_tokens": 2270167040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 412372, "loss/out": 4.943428039550781, "created_at": "2025-01-15T01:12:23.053018+00:00"} {"global_step": 8661, "acc_step": 0, "speed/wps": 12948.973417956158, "speed/FLOPS": 203381421283129.9, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.9657719135284424, "optim/lr": 0.002967340086258479, "optim/total_tokens": 2270429184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 4.832509994506836, "created_at": "2025-01-15T01:12:28.122935+00:00"} {"global_step": 8662, "acc_step": 0, "speed/wps": 12954.576354813322, "speed/FLOPS": 203469423105713.72, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 8.683327674865723, "optim/lr": 0.002967322301965639, "optim/total_tokens": 2270691328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431014, "loss/out": 4.937248229980469, "created_at": "2025-01-15T01:12:33.191591+00:00"} {"global_step": 8663, "acc_step": 0, "speed/wps": 12948.778785156299, "speed/FLOPS": 203378364307554.62, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.373274803161621, "optim/lr": 0.0029673045128854017, "optim/total_tokens": 2270953472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 4.955121994018555, "created_at": "2025-01-15T01:12:38.261932+00:00"} {"global_step": 8664, "acc_step": 0, "speed/wps": 12951.491862464925, "speed/FLOPS": 203420976914845.5, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0011, "optim/grad_norm": 3.9283297061920166, "optim/lr": 0.0029672867190178253, "optim/total_tokens": 2271215616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 282328, "loss/out": 4.826094627380371, "created_at": "2025-01-15T01:12:43.329827+00:00"} {"global_step": 8665, "acc_step": 0, "speed/wps": 12948.493038217106, "speed/FLOPS": 203373876259217.16, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.2169787883758545, "optim/lr": 0.002967268920362969, "optim/total_tokens": 2271477760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433464, "loss/out": 4.92418909072876, "created_at": "2025-01-15T01:12:48.399051+00:00"} {"global_step": 8666, "acc_step": 0, "speed/wps": 12953.075773411314, "speed/FLOPS": 203445854412780.25, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0855966806411743, "optim/lr": 0.002967251116920889, "optim/total_tokens": 2271739904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461605, "loss/out": 4.996156692504883, "created_at": "2025-01-15T01:12:53.467305+00:00"} {"global_step": 8667, "acc_step": 0, "speed/wps": 12949.301792905098, "speed/FLOPS": 203386578862937.47, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1445093154907227, "optim/lr": 0.0029672333086916444, "optim/total_tokens": 2272002048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 4.919720649719238, "created_at": "2025-01-15T01:12:58.534388+00:00"} {"global_step": 8668, "acc_step": 0, "speed/wps": 12947.154618258406, "speed/FLOPS": 203352854534584.03, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0021, "optim/grad_norm": 1.2288367748260498, "optim/lr": 0.0029672154956752934, "optim/total_tokens": 2272264192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322156, "loss/out": 4.806867599487305, "created_at": "2025-01-15T01:13:03.605195+00:00"} {"global_step": 8669, "acc_step": 0, "speed/wps": 12948.96419169131, "speed/FLOPS": 203381276371962.25, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.4253082275390625, "optim/lr": 0.0029671976778718936, "optim/total_tokens": 2272526336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 403274, "loss/out": 4.890093803405762, "created_at": "2025-01-15T01:13:08.669864+00:00"} {"global_step": 8670, "acc_step": 0, "speed/wps": 12948.046756303009, "speed/FLOPS": 203366866788500.66, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4725993871688843, "optim/lr": 0.002967179855281504, "optim/total_tokens": 2272788480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 4.798730373382568, "created_at": "2025-01-15T01:13:13.738280+00:00"} {"global_step": 8671, "acc_step": 0, "speed/wps": 12952.940366451403, "speed/FLOPS": 203443727660406.2, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.446451425552368, "optim/lr": 0.0029671620279041818, "optim/total_tokens": 2273050624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 4.830733776092529, "created_at": "2025-01-15T01:13:18.801481+00:00"} {"global_step": 8672, "acc_step": 0, "speed/wps": 12953.536744808469, "speed/FLOPS": 203453094602013.66, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2020494937896729, "optim/lr": 0.0029671441957399855, "optim/total_tokens": 2273312768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462938, "loss/out": 4.905653953552246, "created_at": "2025-01-15T01:13:23.900778+00:00"} {"global_step": 8673, "acc_step": 0, "speed/wps": 12949.018124542441, "speed/FLOPS": 203382123461500.78, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.0415470600128174, "optim/lr": 0.0029671263587889734, "optim/total_tokens": 2273574912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319663, "loss/out": 4.909872531890869, "created_at": "2025-01-15T01:13:28.976452+00:00"} {"global_step": 8674, "acc_step": 0, "speed/wps": 12951.740072285122, "speed/FLOPS": 203424875391147.72, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4287830591201782, "optim/lr": 0.0029671085170512035, "optim/total_tokens": 2273837056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 4.922622203826904, "created_at": "2025-01-15T01:13:34.040077+00:00"} {"global_step": 8675, "acc_step": 0, "speed/wps": 12955.626705584451, "speed/FLOPS": 203485920307907.53, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.318475365638733, "optim/lr": 0.002967090670526734, "optim/total_tokens": 2274099200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 4.94618034362793, "created_at": "2025-01-15T01:13:39.106261+00:00"} {"global_step": 8676, "acc_step": 0, "speed/wps": 12950.188406955082, "speed/FLOPS": 203400504354927.84, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.356513977050781, "optim/lr": 0.0029670728192156236, "optim/total_tokens": 2274361344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 4.9446258544921875, "created_at": "2025-01-15T01:13:44.170696+00:00"} {"global_step": 8677, "acc_step": 0, "speed/wps": 12951.84877613543, "speed/FLOPS": 203426582734491.56, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.6600526571273804, "optim/lr": 0.0029670549631179302, "optim/total_tokens": 2274623488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447555, "loss/out": 4.924351692199707, "created_at": "2025-01-15T01:13:49.238538+00:00"} {"global_step": 8678, "acc_step": 0, "speed/wps": 12954.219519737382, "speed/FLOPS": 203463818520503.3, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5386438369750977, "optim/lr": 0.0029670371022337124, "optim/total_tokens": 2274885632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287143, "loss/out": 4.9238481521606445, "created_at": "2025-01-15T01:13:54.306260+00:00"} {"global_step": 8679, "acc_step": 0, "speed/wps": 12952.263085647055, "speed/FLOPS": 203433090034692.56, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6472187042236328, "optim/lr": 0.002967019236563028, "optim/total_tokens": 2275147776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436388, "loss/out": 4.911606311798096, "created_at": "2025-01-15T01:13:59.374553+00:00"} {"global_step": 8680, "acc_step": 0, "speed/wps": 12953.298468713008, "speed/FLOPS": 203449352148508.78, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2301846742630005, "optim/lr": 0.002967001366105935, "optim/total_tokens": 2275409920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 4.894920349121094, "created_at": "2025-01-15T01:14:04.442768+00:00"} {"global_step": 8681, "acc_step": 0, "speed/wps": 12953.438549549817, "speed/FLOPS": 203451552310542.12, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.1592278480529785, "optim/lr": 0.0029669834908624924, "optim/total_tokens": 2275672064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461100, "loss/out": 4.864175319671631, "created_at": "2025-01-15T01:14:09.513597+00:00"} {"global_step": 8682, "acc_step": 0, "speed/wps": 12952.941830398397, "speed/FLOPS": 203443750653705.4, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2090237140655518, "optim/lr": 0.0029669656108327583, "optim/total_tokens": 2275934208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 4.8594865798950195, "created_at": "2025-01-15T01:14:14.581639+00:00"} {"global_step": 8683, "acc_step": 0, "speed/wps": 12948.940275424966, "speed/FLOPS": 203380900734134.94, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2030410766601562, "optim/lr": 0.0029669477260167907, "optim/total_tokens": 2276196352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353219, "loss/out": 4.998786926269531, "created_at": "2025-01-15T01:14:19.651618+00:00"} {"global_step": 8684, "acc_step": 0, "speed/wps": 12949.791562560302, "speed/FLOPS": 203394271368386.47, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.680027961730957, "optim/lr": 0.0029669298364146486, "optim/total_tokens": 2276458496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431842, "loss/out": 4.956079959869385, "created_at": "2025-01-15T01:14:24.715981+00:00"} {"global_step": 8685, "acc_step": 0, "speed/wps": 12953.570371083439, "speed/FLOPS": 203453622748869.4, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.169886827468872, "optim/lr": 0.0029669119420263896, "optim/total_tokens": 2276720640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451523, "loss/out": 4.866030216217041, "created_at": "2025-01-15T01:14:29.786870+00:00"} {"global_step": 8686, "acc_step": 0, "speed/wps": 12951.811387526312, "speed/FLOPS": 203425995495008.03, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7500444650650024, "optim/lr": 0.002966894042852073, "optim/total_tokens": 2276982784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431333, "loss/out": 4.877033233642578, "created_at": "2025-01-15T01:14:34.856747+00:00"} {"global_step": 8687, "acc_step": 0, "speed/wps": 12951.027833069724, "speed/FLOPS": 203413688695550.66, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4628968238830566, "optim/lr": 0.0029668761388917564, "optim/total_tokens": 2277244928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462938, "loss/out": 4.973529815673828, "created_at": "2025-01-15T01:14:39.920486+00:00"} {"global_step": 8688, "acc_step": 0, "speed/wps": 12952.405449496957, "speed/FLOPS": 203435326054583.7, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6164575815200806, "optim/lr": 0.0029668582301454987, "optim/total_tokens": 2277507072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322066, "loss/out": 4.805663108825684, "created_at": "2025-01-15T01:14:44.984084+00:00"} {"global_step": 8689, "acc_step": 0, "speed/wps": 12952.031166701361, "speed/FLOPS": 203429447428960.4, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.3570914268493652, "optim/lr": 0.002966840316613358, "optim/total_tokens": 2277769216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367993, "loss/out": 4.818297863006592, "created_at": "2025-01-15T01:14:50.057226+00:00"} {"global_step": 8690, "acc_step": 0, "speed/wps": 12950.058987414575, "speed/FLOPS": 203398471643203.1, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.211417317390442, "optim/lr": 0.0029668223982953932, "optim/total_tokens": 2278031360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 4.929888725280762, "created_at": "2025-01-15T01:14:55.121384+00:00"} {"global_step": 8691, "acc_step": 0, "speed/wps": 12949.216122053027, "speed/FLOPS": 203385233284481.62, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.104625940322876, "optim/lr": 0.002966804475191662, "optim/total_tokens": 2278293504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441687, "loss/out": 4.908921718597412, "created_at": "2025-01-15T01:15:00.189788+00:00"} {"global_step": 8692, "acc_step": 0, "speed/wps": 12949.63764110733, "speed/FLOPS": 203391853820457.34, "speed/curr_iter_time": 1.2771, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3202028274536133, "optim/lr": 0.002966786547302224, "optim/total_tokens": 2278555648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480954, "loss/out": 4.890539169311523, "created_at": "2025-01-15T01:15:05.254738+00:00"} {"global_step": 8693, "acc_step": 0, "speed/wps": 12947.558447937627, "speed/FLOPS": 203359197234616.3, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3654344081878662, "optim/lr": 0.002966768614627136, "optim/total_tokens": 2278817792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327499, "loss/out": 4.955885887145996, "created_at": "2025-01-15T01:15:10.322684+00:00"} {"global_step": 8694, "acc_step": 0, "speed/wps": 12941.742755209278, "speed/FLOPS": 203267853788718.3, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.207415223121643, "optim/lr": 0.002966750677166458, "optim/total_tokens": 2279079936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401411, "loss/out": 4.75150203704834, "created_at": "2025-01-15T01:15:15.396105+00:00"} {"global_step": 8695, "acc_step": 0, "speed/wps": 12954.397920509165, "speed/FLOPS": 203466620549772.53, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0484230518341064, "optim/lr": 0.002966732734920248, "optim/total_tokens": 2279342080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 4.973730564117432, "created_at": "2025-01-15T01:15:20.459844+00:00"} {"global_step": 8696, "acc_step": 0, "speed/wps": 12951.62764438687, "speed/FLOPS": 203423109556513.56, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6568467617034912, "optim/lr": 0.0029667147878885652, "optim/total_tokens": 2279604224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435338, "loss/out": 4.8947625160217285, "created_at": "2025-01-15T01:15:25.529195+00:00"} {"global_step": 8697, "acc_step": 0, "speed/wps": 12951.10301381946, "speed/FLOPS": 203414869512534.44, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1860777139663696, "optim/lr": 0.002966696836071467, "optim/total_tokens": 2279866368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 4.954947471618652, "created_at": "2025-01-15T01:15:30.595105+00:00"} {"global_step": 8698, "acc_step": 0, "speed/wps": 12957.094207125832, "speed/FLOPS": 203508969436173.34, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.252590298652649, "optim/lr": 0.0029666788794690123, "optim/total_tokens": 2280128512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 4.9156084060668945, "created_at": "2025-01-15T01:15:35.690631+00:00"} {"global_step": 8699, "acc_step": 0, "speed/wps": 12952.81266100045, "speed/FLOPS": 203441721870813.28, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.965692043304443, "optim/lr": 0.00296666091808126, "optim/total_tokens": 2280390656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453660, "loss/out": 4.994726657867432, "created_at": "2025-01-15T01:15:40.753847+00:00"} {"global_step": 8700, "acc_step": 0, "speed/wps": 12953.529968069326, "speed/FLOPS": 203452988164013.22, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.510656476020813, "optim/lr": 0.002966642951908269, "optim/total_tokens": 2280652800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453375, "loss/out": 4.976284503936768, "created_at": "2025-01-15T01:15:45.817581+00:00"} {"global_step": 8701, "acc_step": 0, "speed/wps": 12958.67638165606, "speed/FLOPS": 203533819661306.22, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6809712648391724, "optim/lr": 0.002966624980950098, "optim/total_tokens": 2280914944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 4.892838478088379, "created_at": "2025-01-15T01:15:50.881138+00:00"} {"global_step": 8702, "acc_step": 0, "speed/wps": 12950.399694012794, "speed/FLOPS": 203403822908508.06, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5472208261489868, "optim/lr": 0.002966607005206804, "optim/total_tokens": 2281177088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434515, "loss/out": 4.9428253173828125, "created_at": "2025-01-15T01:15:55.948148+00:00"} {"global_step": 8703, "acc_step": 0, "speed/wps": 12953.24637194723, "speed/FLOPS": 203448533897213.78, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.2252943515777588, "optim/lr": 0.0029665890246784476, "optim/total_tokens": 2281439232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 4.992051601409912, "created_at": "2025-01-15T01:16:01.016374+00:00"} {"global_step": 8704, "acc_step": 0, "speed/wps": 12957.94944440619, "speed/FLOPS": 203522402112880.53, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4920685291290283, "optim/lr": 0.0029665710393650857, "optim/total_tokens": 2281701376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318319, "loss/out": 4.953135013580322, "created_at": "2025-01-15T01:16:06.077558+00:00"} {"global_step": 8705, "acc_step": 0, "speed/wps": 12956.44487053421, "speed/FLOPS": 203498770712719.94, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4378365278244019, "optim/lr": 0.0029665530492667793, "optim/total_tokens": 2281963520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 347316, "loss/out": 4.990303039550781, "created_at": "2025-01-15T01:16:11.139281+00:00"} {"global_step": 8706, "acc_step": 0, "speed/wps": 12955.919494219228, "speed/FLOPS": 203490518955750.44, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.4789204597473145, "optim/lr": 0.002966535054383585, "optim/total_tokens": 2282225664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428093, "loss/out": 4.964930057525635, "created_at": "2025-01-15T01:16:16.204565+00:00"} {"global_step": 8707, "acc_step": 0, "speed/wps": 12932.425288624065, "speed/FLOPS": 203121510172456.75, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.256988525390625, "optim/lr": 0.002966517054715562, "optim/total_tokens": 2282487808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440108, "loss/out": 4.851011753082275, "created_at": "2025-01-15T01:16:21.298555+00:00"} {"global_step": 8708, "acc_step": 0, "speed/wps": 12951.677089250967, "speed/FLOPS": 203423886156048.56, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6998769044876099, "optim/lr": 0.00296649905026277, "optim/total_tokens": 2282749952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326344, "loss/out": 4.941445350646973, "created_at": "2025-01-15T01:16:26.362588+00:00"} {"global_step": 8709, "acc_step": 0, "speed/wps": 12951.469834314928, "speed/FLOPS": 203420630932479.97, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6963424682617188, "optim/lr": 0.0029664810410252665, "optim/total_tokens": 2283012096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352926, "loss/out": 4.984848499298096, "created_at": "2025-01-15T01:16:31.429325+00:00"} {"global_step": 8710, "acc_step": 0, "speed/wps": 12950.962268426587, "speed/FLOPS": 203412658912734.94, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6489405632019043, "optim/lr": 0.002966463027003111, "optim/total_tokens": 2283274240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438240, "loss/out": 4.809866905212402, "created_at": "2025-01-15T01:16:36.493760+00:00"} {"global_step": 8711, "acc_step": 0, "speed/wps": 12952.839157016046, "speed/FLOPS": 203442138027147.56, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.089888095855713, "optim/lr": 0.0029664450081963623, "optim/total_tokens": 2283536384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441157, "loss/out": 4.933882713317871, "created_at": "2025-01-15T01:16:41.560294+00:00"} {"global_step": 8712, "acc_step": 0, "speed/wps": 12950.03827013735, "speed/FLOPS": 203398146249895.75, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3218486309051514, "optim/lr": 0.0029664269846050788, "optim/total_tokens": 2283798528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 4.80430793762207, "created_at": "2025-01-15T01:16:46.624618+00:00"} {"global_step": 8713, "acc_step": 0, "speed/wps": 12955.077404204747, "speed/FLOPS": 203477292774920.06, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.106238603591919, "optim/lr": 0.0029664089562293195, "optim/total_tokens": 2284060672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444359, "loss/out": 4.9627838134765625, "created_at": "2025-01-15T01:16:51.692656+00:00"} {"global_step": 8714, "acc_step": 0, "speed/wps": 12959.036569566026, "speed/FLOPS": 203539476907382.28, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1407685279846191, "optim/lr": 0.0029663909230691426, "optim/total_tokens": 2284322816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 4.9077606201171875, "created_at": "2025-01-15T01:16:56.758473+00:00"} {"global_step": 8715, "acc_step": 0, "speed/wps": 12956.055421474426, "speed/FLOPS": 203492653880076.94, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0014, "optim/grad_norm": 1.529648780822754, "optim/lr": 0.0029663728851246083, "optim/total_tokens": 2284584960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326344, "loss/out": 4.956981658935547, "created_at": "2025-01-15T01:17:01.820331+00:00"} {"global_step": 8716, "acc_step": 0, "speed/wps": 12954.754097628802, "speed/FLOPS": 203472214800875.06, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.632983684539795, "optim/lr": 0.0029663548423957743, "optim/total_tokens": 2284847104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 346666, "loss/out": 4.895481109619141, "created_at": "2025-01-15T01:17:06.901595+00:00"} {"global_step": 8717, "acc_step": 0, "speed/wps": 12955.080275823673, "speed/FLOPS": 203477337877641.34, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4028310775756836, "optim/lr": 0.0029663367948827, "optim/total_tokens": 2285109248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 4.949929237365723, "created_at": "2025-01-15T01:17:11.966774+00:00"} {"global_step": 8718, "acc_step": 0, "speed/wps": 12949.949074352957, "speed/FLOPS": 203396745307534.0, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.560464382171631, "optim/lr": 0.0029663187425854436, "optim/total_tokens": 2285371392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 4.981287956237793, "created_at": "2025-01-15T01:17:17.040449+00:00"} {"global_step": 8719, "acc_step": 0, "speed/wps": 12950.183895435614, "speed/FLOPS": 203400433495315.12, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.7108492851257324, "optim/lr": 0.0029663006855040652, "optim/total_tokens": 2285633536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321253, "loss/out": 4.871922492980957, "created_at": "2025-01-15T01:17:22.109660+00:00"} {"global_step": 8720, "acc_step": 0, "speed/wps": 12958.52129916095, "speed/FLOPS": 203531383877614.84, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.524327754974365, "optim/lr": 0.0029662826236386227, "optim/total_tokens": 2285895680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355086, "loss/out": 5.013343811035156, "created_at": "2025-01-15T01:17:27.175864+00:00"} {"global_step": 8721, "acc_step": 0, "speed/wps": 12951.224278315738, "speed/FLOPS": 203416774138083.38, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2445955276489258, "optim/lr": 0.002966264556989175, "optim/total_tokens": 2286157824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457095, "loss/out": 4.909355640411377, "created_at": "2025-01-15T01:17:32.243193+00:00"} {"global_step": 8722, "acc_step": 0, "speed/wps": 12955.806943658426, "speed/FLOPS": 203488751194532.38, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2806017398834229, "optim/lr": 0.0029662464855557823, "optim/total_tokens": 2286419968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452081, "loss/out": 4.88167142868042, "created_at": "2025-01-15T01:17:37.310131+00:00"} {"global_step": 8723, "acc_step": 0, "speed/wps": 12953.813090151381, "speed/FLOPS": 203457434985362.4, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.120833158493042, "optim/lr": 0.0029662284093385023, "optim/total_tokens": 2286682112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 5.113676071166992, "created_at": "2025-01-15T01:17:42.378815+00:00"} {"global_step": 8724, "acc_step": 0, "speed/wps": 12956.394804800313, "speed/FLOPS": 203497984361571.56, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4130547046661377, "optim/lr": 0.002966210328337394, "optim/total_tokens": 2286944256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437201, "loss/out": 4.928450584411621, "created_at": "2025-01-15T01:17:47.444755+00:00"} {"global_step": 8725, "acc_step": 0, "speed/wps": 12955.237024263102, "speed/FLOPS": 203479799827277.5, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.8130818605422974, "optim/lr": 0.002966192242552517, "optim/total_tokens": 2287206400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436666, "loss/out": 4.7744646072387695, "created_at": "2025-01-15T01:17:52.508814+00:00"} {"global_step": 8726, "acc_step": 0, "speed/wps": 12955.204410162396, "speed/FLOPS": 203479287578009.44, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3030240535736084, "optim/lr": 0.00296617415198393, "optim/total_tokens": 2287468544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 282524, "loss/out": 4.801053047180176, "created_at": "2025-01-15T01:17:57.572407+00:00"} {"global_step": 8727, "acc_step": 0, "speed/wps": 12948.427575479964, "speed/FLOPS": 203372848076975.12, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9286085367202759, "optim/lr": 0.002966156056631692, "optim/total_tokens": 2287730688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439871, "loss/out": 4.783661842346191, "created_at": "2025-01-15T01:18:02.638402+00:00"} {"global_step": 8728, "acc_step": 0, "speed/wps": 12952.580159715095, "speed/FLOPS": 203438070118636.56, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.618529319763184, "optim/lr": 0.0029661379564958623, "optim/total_tokens": 2287992832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 4.878353118896484, "created_at": "2025-01-15T01:18:07.706295+00:00"} {"global_step": 8729, "acc_step": 0, "speed/wps": 12948.890648918026, "speed/FLOPS": 203380121281649.47, "speed/curr_iter_time": 1.2778, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6376255750656128, "optim/lr": 0.0029661198515765, "optim/total_tokens": 2288254976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448368, "loss/out": 4.970701217651367, "created_at": "2025-01-15T01:18:12.773212+00:00"} {"global_step": 8730, "acc_step": 0, "speed/wps": 12954.737963489324, "speed/FLOPS": 203471961392043.94, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3900816440582275, "optim/lr": 0.0029661017418736637, "optim/total_tokens": 2288517120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450742, "loss/out": 4.995361804962158, "created_at": "2025-01-15T01:18:17.844962+00:00"} {"global_step": 8731, "acc_step": 0, "speed/wps": 12949.335525875855, "speed/FLOPS": 203387108685597.16, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0713993310928345, "optim/lr": 0.0029660836273874124, "optim/total_tokens": 2288779264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294309, "loss/out": 4.858912944793701, "created_at": "2025-01-15T01:18:22.912710+00:00"} {"global_step": 8732, "acc_step": 0, "speed/wps": 12957.488750612129, "speed/FLOPS": 203515166283785.2, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6278352737426758, "optim/lr": 0.002966065508117806, "optim/total_tokens": 2289041408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417976, "loss/out": 4.88502311706543, "created_at": "2025-01-15T01:18:27.978563+00:00"} {"global_step": 8733, "acc_step": 0, "speed/wps": 12955.159148054112, "speed/FLOPS": 203478576674402.16, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5017390251159668, "optim/lr": 0.002966047384064903, "optim/total_tokens": 2289303552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 4.827236175537109, "created_at": "2025-01-15T01:18:33.045237+00:00"} {"global_step": 8734, "acc_step": 0, "speed/wps": 12954.151225651067, "speed/FLOPS": 203462745868031.8, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.532378911972046, "optim/lr": 0.002966029255228763, "optim/total_tokens": 2289565696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439583, "loss/out": 4.873866081237793, "created_at": "2025-01-15T01:18:38.111567+00:00"} {"global_step": 8735, "acc_step": 0, "speed/wps": 12943.613515333733, "speed/FLOPS": 203297236647166.7, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.219101071357727, "optim/lr": 0.002966011121609444, "optim/total_tokens": 2289827840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444641, "loss/out": 4.929943561553955, "created_at": "2025-01-15T01:18:43.197122+00:00"} {"global_step": 8736, "acc_step": 0, "speed/wps": 12952.714018953955, "speed/FLOPS": 203440172561924.72, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.161119818687439, "optim/lr": 0.002965992983207006, "optim/total_tokens": 2290089984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 414278, "loss/out": 4.8722944259643555, "created_at": "2025-01-15T01:18:48.264222+00:00"} {"global_step": 8737, "acc_step": 0, "speed/wps": 12955.215725142605, "speed/FLOPS": 203479465295322.03, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5201640129089355, "optim/lr": 0.0029659748400215086, "optim/total_tokens": 2290352128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 4.95939826965332, "created_at": "2025-01-15T01:18:53.331489+00:00"} {"global_step": 8738, "acc_step": 0, "speed/wps": 12958.325660969113, "speed/FLOPS": 203528311110976.97, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3896846771240234, "optim/lr": 0.002965956692053011, "optim/total_tokens": 2290614272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479178, "loss/out": 5.019548416137695, "created_at": "2025-01-15T01:18:58.393655+00:00"} {"global_step": 8739, "acc_step": 0, "speed/wps": 12951.96615688801, "speed/FLOPS": 203428426360508.88, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2839505672454834, "optim/lr": 0.002965938539301571, "optim/total_tokens": 2290876416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458430, "loss/out": 4.84013032913208, "created_at": "2025-01-15T01:19:03.457103+00:00"} {"global_step": 8740, "acc_step": 0, "speed/wps": 12955.315848313958, "speed/FLOPS": 203481037867311.78, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2434005737304688, "optim/lr": 0.0029659203817672497, "optim/total_tokens": 2291138560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444067, "loss/out": 4.961210250854492, "created_at": "2025-01-15T01:19:08.522622+00:00"} {"global_step": 8741, "acc_step": 0, "speed/wps": 12954.20325576855, "speed/FLOPS": 203463563072523.78, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1878259181976318, "optim/lr": 0.0029659022194501047, "optim/total_tokens": 2291400704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 4.7718400955200195, "created_at": "2025-01-15T01:19:13.590567+00:00"} {"global_step": 8742, "acc_step": 0, "speed/wps": 12945.467035871574, "speed/FLOPS": 203326348734217.6, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2342334985733032, "optim/lr": 0.0029658840523501965, "optim/total_tokens": 2291662848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287957, "loss/out": 4.897010803222656, "created_at": "2025-01-15T01:19:18.656651+00:00"} {"global_step": 8743, "acc_step": 0, "speed/wps": 12946.114860506634, "speed/FLOPS": 203336523710315.56, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.001, "optim/grad_norm": 1.5598652362823486, "optim/lr": 0.0029658658804675836, "optim/total_tokens": 2291924992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443306, "loss/out": 4.745867729187012, "created_at": "2025-01-15T01:19:23.754753+00:00"} {"global_step": 8744, "acc_step": 0, "speed/wps": 12950.586863025188, "speed/FLOPS": 203406762655044.88, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.3321104049682617, "optim/lr": 0.0029658477038023256, "optim/total_tokens": 2292187136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 4.93014669418335, "created_at": "2025-01-15T01:19:28.819067+00:00"} {"global_step": 8745, "acc_step": 0, "speed/wps": 12956.135462524368, "speed/FLOPS": 203493911034754.22, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7802537679672241, "optim/lr": 0.0029658295223544823, "optim/total_tokens": 2292449280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453144, "loss/out": 5.064313888549805, "created_at": "2025-01-15T01:19:33.884603+00:00"} {"global_step": 8746, "acc_step": 0, "speed/wps": 12945.132170772793, "speed/FLOPS": 203321089217688.38, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.8269211053848267, "optim/lr": 0.002965811336124112, "optim/total_tokens": 2292711424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358017, "loss/out": 4.867971420288086, "created_at": "2025-01-15T01:19:38.955409+00:00"} {"global_step": 8747, "acc_step": 0, "speed/wps": 12953.217616257103, "speed/FLOPS": 203448082249585.88, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0012, "optim/grad_norm": 2.3185439109802246, "optim/lr": 0.002965793145111274, "optim/total_tokens": 2292973568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 4.834953308105469, "created_at": "2025-01-15T01:19:44.022465+00:00"} {"global_step": 8748, "acc_step": 0, "speed/wps": 12953.015135463707, "speed/FLOPS": 203444902010487.84, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.001, "optim/grad_norm": 1.5566269159317017, "optim/lr": 0.0029657749493160294, "optim/total_tokens": 2293235712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 4.853155136108398, "created_at": "2025-01-15T01:19:49.086273+00:00"} {"global_step": 8749, "acc_step": 0, "speed/wps": 12955.058981988695, "speed/FLOPS": 203477003428702.22, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7420066595077515, "optim/lr": 0.0029657567487384358, "optim/total_tokens": 2293497856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 4.781520843505859, "created_at": "2025-01-15T01:19:54.154838+00:00"} {"global_step": 8750, "acc_step": 0, "speed/wps": 12954.659993113808, "speed/FLOPS": 203470736760153.94, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.350708246231079, "optim/lr": 0.002965738543378553, "optim/total_tokens": 2293760000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452325, "loss/out": 5.002655029296875, "created_at": "2025-01-15T01:19:59.218515+00:00"} {"global_step": 8751, "acc_step": 0, "speed/wps": 12950.26496026938, "speed/FLOPS": 203401706729924.2, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3454560041427612, "optim/lr": 0.0029657203332364404, "optim/total_tokens": 2294022144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435866, "loss/out": 4.949244976043701, "created_at": "2025-01-15T01:20:04.282733+00:00"} {"global_step": 8752, "acc_step": 0, "speed/wps": 12956.6716347224, "speed/FLOPS": 203502332355902.1, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 44.58814239501953, "optim/lr": 0.0029657021183121576, "optim/total_tokens": 2294284288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320967, "loss/out": 4.847818374633789, "created_at": "2025-01-15T01:20:09.348346+00:00"} {"global_step": 8753, "acc_step": 0, "speed/wps": 12951.93248649534, "speed/FLOPS": 203427897520724.03, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4058384895324707, "optim/lr": 0.0029656838986057642, "optim/total_tokens": 2294546432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 399826, "loss/out": 4.921226501464844, "created_at": "2025-01-15T01:20:14.436447+00:00"} {"global_step": 8754, "acc_step": 0, "speed/wps": 12948.324437570318, "speed/FLOPS": 203371228154377.78, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2853158712387085, "optim/lr": 0.0029656656741173194, "optim/total_tokens": 2294808576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437195, "loss/out": 4.865561008453369, "created_at": "2025-01-15T01:20:19.505615+00:00"} {"global_step": 8755, "acc_step": 0, "speed/wps": 12948.647752109375, "speed/FLOPS": 203376306253494.4, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7965986728668213, "optim/lr": 0.0029656474448468825, "optim/total_tokens": 2295070720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 4.9046759605407715, "created_at": "2025-01-15T01:20:24.575104+00:00"} {"global_step": 8756, "acc_step": 0, "speed/wps": 12955.450324625894, "speed/FLOPS": 203483150002580.53, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.9114439487457275, "optim/lr": 0.002965629210794513, "optim/total_tokens": 2295332864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293489, "loss/out": 5.109408378601074, "created_at": "2025-01-15T01:20:29.641378+00:00"} {"global_step": 8757, "acc_step": 0, "speed/wps": 12956.215425842307, "speed/FLOPS": 203495166968543.56, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.443627953529358, "optim/lr": 0.002965610971960271, "optim/total_tokens": 2295595008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441971, "loss/out": 4.934120178222656, "created_at": "2025-01-15T01:20:34.703622+00:00"} {"global_step": 8758, "acc_step": 0, "speed/wps": 12953.39945403497, "speed/FLOPS": 203450938261758.56, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.8182194232940674, "optim/lr": 0.002965592728344215, "optim/total_tokens": 2295857152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 4.827404975891113, "created_at": "2025-01-15T01:20:39.766948+00:00"} {"global_step": 8759, "acc_step": 0, "speed/wps": 12957.280441178447, "speed/FLOPS": 203511894497882.06, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.6477551460266113, "optim/lr": 0.0029655744799464054, "optim/total_tokens": 2296119296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448361, "loss/out": 4.8039703369140625, "created_at": "2025-01-15T01:20:44.833022+00:00"} {"global_step": 8760, "acc_step": 0, "speed/wps": 12953.265015526324, "speed/FLOPS": 203448826720242.4, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.253487467765808, "optim/lr": 0.0029655562267669015, "optim/total_tokens": 2296381440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 4.884555339813232, "created_at": "2025-01-15T01:20:49.896068+00:00"} {"global_step": 8761, "acc_step": 0, "speed/wps": 12955.715337952557, "speed/FLOPS": 203487312401042.25, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.9710898399353027, "optim/lr": 0.0029655379688057625, "optim/total_tokens": 2296643584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450225, "loss/out": 4.759505271911621, "created_at": "2025-01-15T01:20:54.963624+00:00"} {"global_step": 8762, "acc_step": 0, "speed/wps": 12954.486996340796, "speed/FLOPS": 203468019608111.3, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3993021249771118, "optim/lr": 0.002965519706063048, "optim/total_tokens": 2296905728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345322, "loss/out": 4.864111423492432, "created_at": "2025-01-15T01:21:00.026484+00:00"} {"global_step": 8763, "acc_step": 0, "speed/wps": 12955.865937998948, "speed/FLOPS": 203489677781716.97, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.468397617340088, "optim/lr": 0.0029655014385388187, "optim/total_tokens": 2297167872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435320, "loss/out": 4.8677144050598145, "created_at": "2025-01-15T01:21:05.092753+00:00"} {"global_step": 8764, "acc_step": 0, "speed/wps": 12952.211562219652, "speed/FLOPS": 203432280788464.6, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.6313843727111816, "optim/lr": 0.002965483166233133, "optim/total_tokens": 2297430016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441442, "loss/out": 4.922188758850098, "created_at": "2025-01-15T01:21:10.160918+00:00"} {"global_step": 8765, "acc_step": 0, "speed/wps": 12952.352310926923, "speed/FLOPS": 203434491440321.28, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8238353729248047, "optim/lr": 0.0029654648891460503, "optim/total_tokens": 2297692160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478284, "loss/out": 4.8096394538879395, "created_at": "2025-01-15T01:21:15.229460+00:00"} {"global_step": 8766, "acc_step": 0, "speed/wps": 12956.891781664099, "speed/FLOPS": 203505790066135.1, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7862460613250732, "optim/lr": 0.002965446607277631, "optim/total_tokens": 2297954304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438531, "loss/out": 4.992312908172607, "created_at": "2025-01-15T01:21:20.291149+00:00"} {"global_step": 8767, "acc_step": 0, "speed/wps": 12953.363093701164, "speed/FLOPS": 203450367172751.8, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2333999872207642, "optim/lr": 0.0029654283206279348, "optim/total_tokens": 2298216448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 4.801470756530762, "created_at": "2025-01-15T01:21:25.358981+00:00"} {"global_step": 8768, "acc_step": 0, "speed/wps": 12959.326361992127, "speed/FLOPS": 203544028495650.72, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.649585247039795, "optim/lr": 0.002965410029197021, "optim/total_tokens": 2298478592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309195, "loss/out": 4.888237953186035, "created_at": "2025-01-15T01:21:30.425231+00:00"} {"global_step": 8769, "acc_step": 0, "speed/wps": 12950.465393578601, "speed/FLOPS": 203404854810469.72, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1942459344863892, "optim/lr": 0.002965391732984949, "optim/total_tokens": 2298740736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 4.844213485717773, "created_at": "2025-01-15T01:21:35.493417+00:00"} {"global_step": 8770, "acc_step": 0, "speed/wps": 12951.7715485369, "speed/FLOPS": 203425369768934.8, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.495382308959961, "optim/lr": 0.0029653734319917795, "optim/total_tokens": 2299002880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 4.760956764221191, "created_at": "2025-01-15T01:21:40.585677+00:00"} {"global_step": 8771, "acc_step": 0, "speed/wps": 12960.743494836099, "speed/FLOPS": 203566286514309.9, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.06820011138916, "optim/lr": 0.0029653551262175705, "optim/total_tokens": 2299265024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468540, "loss/out": 5.0120439529418945, "created_at": "2025-01-15T01:21:45.650997+00:00"} {"global_step": 8772, "acc_step": 0, "speed/wps": 12955.130088120319, "speed/FLOPS": 203478120248209.4, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2717758417129517, "optim/lr": 0.002965336815662384, "optim/total_tokens": 2299527168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290078, "loss/out": 4.948031425476074, "created_at": "2025-01-15T01:21:50.718380+00:00"} {"global_step": 8773, "acc_step": 0, "speed/wps": 12949.354148667064, "speed/FLOPS": 203387401182123.5, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3057180643081665, "optim/lr": 0.002965318500326278, "optim/total_tokens": 2299789312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427589, "loss/out": 4.99739933013916, "created_at": "2025-01-15T01:21:55.785285+00:00"} {"global_step": 8774, "acc_step": 0, "speed/wps": 12958.952602683677, "speed/FLOPS": 203538158092112.4, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.2673227787017822, "optim/lr": 0.002965300180209313, "optim/total_tokens": 2300051456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 4.939043998718262, "created_at": "2025-01-15T01:22:00.846214+00:00"} {"global_step": 8775, "acc_step": 0, "speed/wps": 12953.896756750519, "speed/FLOPS": 203458749084271.8, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8157061338424683, "optim/lr": 0.002965281855311548, "optim/total_tokens": 2300313600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431320, "loss/out": 5.009897232055664, "created_at": "2025-01-15T01:22:05.911283+00:00"} {"global_step": 8776, "acc_step": 0, "speed/wps": 12960.18443845959, "speed/FLOPS": 203557505765692.84, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8048348426818848, "optim/lr": 0.0029652635256330437, "optim/total_tokens": 2300575744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 4.850336074829102, "created_at": "2025-01-15T01:22:10.975724+00:00"} {"global_step": 8777, "acc_step": 0, "speed/wps": 12948.014419023395, "speed/FLOPS": 203366358887088.22, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1406478881835938, "optim/lr": 0.0029652451911738598, "optim/total_tokens": 2300837888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433179, "loss/out": 4.741905689239502, "created_at": "2025-01-15T01:22:16.043644+00:00"} {"global_step": 8778, "acc_step": 0, "speed/wps": 12955.121055811906, "speed/FLOPS": 203477978383394.1, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.0800228118896484, "optim/lr": 0.0029652268519340555, "optim/total_tokens": 2301100032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291719, "loss/out": 4.821257591247559, "created_at": "2025-01-15T01:22:21.112224+00:00"} {"global_step": 8779, "acc_step": 0, "speed/wps": 12955.866978013973, "speed/FLOPS": 203489694116582.1, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.530981421470642, "optim/lr": 0.0029652085079136916, "optim/total_tokens": 2301362176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355086, "loss/out": 4.907792091369629, "created_at": "2025-01-15T01:22:26.182593+00:00"} {"global_step": 8780, "acc_step": 0, "speed/wps": 12960.799965544322, "speed/FLOPS": 203567173464381.44, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1999653577804565, "optim/lr": 0.0029651901591128266, "optim/total_tokens": 2301624320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 4.962987899780273, "created_at": "2025-01-15T01:22:31.249128+00:00"} {"global_step": 8781, "acc_step": 0, "speed/wps": 12956.360837041162, "speed/FLOPS": 203497450851237.5, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.722632884979248, "optim/lr": 0.002965171805531521, "optim/total_tokens": 2301886464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424156, "loss/out": 4.740471839904785, "created_at": "2025-01-15T01:22:36.317578+00:00"} {"global_step": 8782, "acc_step": 0, "speed/wps": 12950.69704909338, "speed/FLOPS": 203408493278657.38, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.5215500593185425, "optim/lr": 0.0029651534471698353, "optim/total_tokens": 2302148608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 4.901291370391846, "created_at": "2025-01-15T01:22:41.387628+00:00"} {"global_step": 8783, "acc_step": 0, "speed/wps": 12959.336147181573, "speed/FLOPS": 203544182185497.25, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4315937757492065, "optim/lr": 0.0029651350840278297, "optim/total_tokens": 2302410752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297001, "loss/out": 4.8998918533325195, "created_at": "2025-01-15T01:22:46.453416+00:00"} {"global_step": 8784, "acc_step": 0, "speed/wps": 12949.887302696683, "speed/FLOPS": 203395775098789.1, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.038668632507324, "optim/lr": 0.002965116716105562, "optim/total_tokens": 2302672896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441450, "loss/out": 4.893148422241211, "created_at": "2025-01-15T01:22:51.521912+00:00"} {"global_step": 8785, "acc_step": 0, "speed/wps": 12959.072780381193, "speed/FLOPS": 203540045647993.38, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.219207763671875, "optim/lr": 0.002965098343403094, "optim/total_tokens": 2302935040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490472, "loss/out": 5.00160026550293, "created_at": "2025-01-15T01:22:56.583782+00:00"} {"global_step": 8786, "acc_step": 0, "speed/wps": 12962.265786033431, "speed/FLOPS": 203590196189411.3, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.38249933719635, "optim/lr": 0.0029650799659204845, "optim/total_tokens": 2303197184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433179, "loss/out": 5.001646995544434, "created_at": "2025-01-15T01:23:01.650158+00:00"} {"global_step": 8787, "acc_step": 0, "speed/wps": 12954.961624642934, "speed/FLOPS": 203475474297804.0, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.143273830413818, "optim/lr": 0.002965061583657795, "optim/total_tokens": 2303459328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 4.785561561584473, "created_at": "2025-01-15T01:23:06.717933+00:00"} {"global_step": 8788, "acc_step": 0, "speed/wps": 12958.135507376996, "speed/FLOPS": 203525324487514.38, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5287836790084839, "optim/lr": 0.002965043196615084, "optim/total_tokens": 2303721472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316287, "loss/out": 4.774356842041016, "created_at": "2025-01-15T01:23:11.779159+00:00"} {"global_step": 8789, "acc_step": 0, "speed/wps": 12953.060570953381, "speed/FLOPS": 203445615637288.06, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2707635164260864, "optim/lr": 0.002965024804792412, "optim/total_tokens": 2303983616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365501, "loss/out": 4.836922645568848, "created_at": "2025-01-15T01:23:16.842410+00:00"} {"global_step": 8790, "acc_step": 0, "speed/wps": 12954.03582273001, "speed/FLOPS": 203460933306576.38, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2860064506530762, "optim/lr": 0.002965006408189839, "optim/total_tokens": 2304245760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430506, "loss/out": 4.864487648010254, "created_at": "2025-01-15T01:23:21.907260+00:00"} {"global_step": 8791, "acc_step": 0, "speed/wps": 12951.898053431441, "speed/FLOPS": 203427356702139.53, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8047183752059937, "optim/lr": 0.0029649880068074257, "optim/total_tokens": 2304507904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446742, "loss/out": 4.885497093200684, "created_at": "2025-01-15T01:23:26.974077+00:00"} {"global_step": 8792, "acc_step": 0, "speed/wps": 12953.916506288968, "speed/FLOPS": 203459059277911.84, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1995292901992798, "optim/lr": 0.0029649696006452306, "optim/total_tokens": 2304770048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 4.848226070404053, "created_at": "2025-01-15T01:23:32.044497+00:00"} {"global_step": 8793, "acc_step": 0, "speed/wps": 12954.173224767443, "speed/FLOPS": 203463091394384.47, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3004529476165771, "optim/lr": 0.002964951189703315, "optim/total_tokens": 2305032192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438240, "loss/out": 4.929201126098633, "created_at": "2025-01-15T01:23:37.107097+00:00"} {"global_step": 8794, "acc_step": 0, "speed/wps": 12955.016456539846, "speed/FLOPS": 203476335508091.94, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.4564008712768555, "optim/lr": 0.0029649327739817385, "optim/total_tokens": 2305294336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 4.720500946044922, "created_at": "2025-01-15T01:23:42.169594+00:00"} {"global_step": 8795, "acc_step": 0, "speed/wps": 12945.603315468465, "speed/FLOPS": 203328489192554.56, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2527332305908203, "optim/lr": 0.0029649143534805616, "optim/total_tokens": 2305556480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334467, "loss/out": 4.951690673828125, "created_at": "2025-01-15T01:23:47.239884+00:00"} {"global_step": 8796, "acc_step": 0, "speed/wps": 12952.733955622332, "speed/FLOPS": 203440485694697.44, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1867958307266235, "optim/lr": 0.0029648959281998438, "optim/total_tokens": 2305818624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415004, "loss/out": 4.840958595275879, "created_at": "2025-01-15T01:23:52.304716+00:00"} {"global_step": 8797, "acc_step": 0, "speed/wps": 12960.049885332995, "speed/FLOPS": 203555392423944.3, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.632086157798767, "optim/lr": 0.0029648774981396457, "optim/total_tokens": 2306080768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 4.968916893005371, "created_at": "2025-01-15T01:23:57.367992+00:00"} {"global_step": 8798, "acc_step": 0, "speed/wps": 12949.38062815224, "speed/FLOPS": 203387817078824.8, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3436620235443115, "optim/lr": 0.002964859063300027, "optim/total_tokens": 2306342912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447269, "loss/out": 4.961091995239258, "created_at": "2025-01-15T01:24:02.434176+00:00"} {"global_step": 8799, "acc_step": 0, "speed/wps": 12957.602008018022, "speed/FLOPS": 203516945146976.75, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.252413272857666, "optim/lr": 0.0029648406236810485, "optim/total_tokens": 2306605056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498484, "loss/out": 4.727721214294434, "created_at": "2025-01-15T01:24:07.497211+00:00"} {"global_step": 8800, "acc_step": 0, "speed/wps": 12957.025870796864, "speed/FLOPS": 203507896120222.94, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1865310668945312, "optim/lr": 0.00296482217928277, "optim/total_tokens": 2306867200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449420, "loss/out": 4.943603515625, "created_at": "2025-01-15T01:24:12.562975+00:00"} {"global_step": 8801, "acc_step": 0, "speed/wps": 12943.17774712769, "speed/FLOPS": 203290392308682.5, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1442171335220337, "optim/lr": 0.002964803730105251, "optim/total_tokens": 2307129344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368187, "loss/out": 4.8893842697143555, "created_at": "2025-01-15T01:24:17.630905+00:00"} {"global_step": 8802, "acc_step": 0, "speed/wps": 12929.185365070185, "speed/FLOPS": 203070622720921.56, "speed/curr_iter_time": 1.2778, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.545330286026001, "optim/lr": 0.0029647852761485525, "optim/total_tokens": 2307391488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 4.736742973327637, "created_at": "2025-01-15T01:24:22.704952+00:00"} {"global_step": 8803, "acc_step": 0, "speed/wps": 12944.61691249363, "speed/FLOPS": 203312996378375.06, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4283766746520996, "optim/lr": 0.0029647668174127345, "optim/total_tokens": 2307653632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440921, "loss/out": 4.850669860839844, "created_at": "2025-01-15T01:24:27.777106+00:00"} {"global_step": 8804, "acc_step": 0, "speed/wps": 12952.325671475879, "speed/FLOPS": 203434073031136.2, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1231751441955566, "optim/lr": 0.0029647483538978573, "optim/total_tokens": 2307915776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443545, "loss/out": 4.8418378829956055, "created_at": "2025-01-15T01:24:32.846399+00:00"} {"global_step": 8805, "acc_step": 0, "speed/wps": 12938.927072568753, "speed/FLOPS": 203223629623695.47, "speed/curr_iter_time": 1.2787, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.148681402206421, "optim/lr": 0.0029647298856039814, "optim/total_tokens": 2308177920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449696, "loss/out": 4.891312599182129, "created_at": "2025-01-15T01:24:37.917387+00:00"} {"global_step": 8806, "acc_step": 0, "speed/wps": 12945.167125123027, "speed/FLOPS": 203321638223791.12, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2430310249328613, "optim/lr": 0.002964711412531166, "optim/total_tokens": 2308440064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306165, "loss/out": 5.003572463989258, "created_at": "2025-01-15T01:24:42.989016+00:00"} {"global_step": 8807, "acc_step": 0, "speed/wps": 12950.146037404824, "speed/FLOPS": 203399838882919.2, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.292665958404541, "optim/lr": 0.002964692934679473, "optim/total_tokens": 2308702208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431320, "loss/out": 4.889947891235352, "created_at": "2025-01-15T01:24:48.055051+00:00"} {"global_step": 8808, "acc_step": 0, "speed/wps": 12944.354098371046, "speed/FLOPS": 203308868521435.44, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1527220010757446, "optim/lr": 0.0029646744520489612, "optim/total_tokens": 2308964352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460578, "loss/out": 4.90565299987793, "created_at": "2025-01-15T01:24:53.122496+00:00"} {"global_step": 8809, "acc_step": 0, "speed/wps": 12949.34674243735, "speed/FLOPS": 203387284857108.62, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4469356536865234, "optim/lr": 0.002964655964639692, "optim/total_tokens": 2309226496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467205, "loss/out": 4.873907089233398, "created_at": "2025-01-15T01:24:58.187871+00:00"} {"global_step": 8810, "acc_step": 0, "speed/wps": 12961.422359965627, "speed/FLOPS": 203576949024027.6, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3503749370574951, "optim/lr": 0.002964637472451725, "optim/total_tokens": 2309488640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 4.78820276260376, "created_at": "2025-01-15T01:25:03.253107+00:00"} {"global_step": 8811, "acc_step": 0, "speed/wps": 12947.78519420159, "speed/FLOPS": 203362758596273.88, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1565560102462769, "optim/lr": 0.0029646189754851207, "optim/total_tokens": 2309750784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301350, "loss/out": 4.9415411949157715, "created_at": "2025-01-15T01:25:08.320690+00:00"} {"global_step": 8812, "acc_step": 0, "speed/wps": 12953.745174728534, "speed/FLOPS": 203456368280321.0, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2283447980880737, "optim/lr": 0.0029646004737399396, "optim/total_tokens": 2310012928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437432, "loss/out": 4.912930488586426, "created_at": "2025-01-15T01:25:13.388210+00:00"} {"global_step": 8813, "acc_step": 0, "speed/wps": 12947.567274355166, "speed/FLOPS": 203359335865632.53, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.9022905826568604, "optim/lr": 0.002964581967216242, "optim/total_tokens": 2310275072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 4.920226097106934, "created_at": "2025-01-15T01:25:18.457275+00:00"} {"global_step": 8814, "acc_step": 0, "speed/wps": 12947.225428015883, "speed/FLOPS": 203353966699128.0, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.2601399421691895, "optim/lr": 0.0029645634559140885, "optim/total_tokens": 2310537216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467490, "loss/out": 4.777275085449219, "created_at": "2025-01-15T01:25:23.526281+00:00"} {"global_step": 8815, "acc_step": 0, "speed/wps": 12951.212336257817, "speed/FLOPS": 203416586571653.6, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3318417072296143, "optim/lr": 0.0029645449398335386, "optim/total_tokens": 2310799360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448361, "loss/out": 4.833476543426514, "created_at": "2025-01-15T01:25:28.597775+00:00"} {"global_step": 8816, "acc_step": 0, "speed/wps": 12951.394301284845, "speed/FLOPS": 203419444582449.22, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.397231101989746, "optim/lr": 0.002964526418974654, "optim/total_tokens": 2311061504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350820, "loss/out": 4.775506496429443, "created_at": "2025-01-15T01:25:33.666212+00:00"} {"global_step": 8817, "acc_step": 0, "speed/wps": 12946.309220302497, "speed/FLOPS": 203339576397987.94, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.879019856452942, "optim/lr": 0.002964507893337494, "optim/total_tokens": 2311323648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428403, "loss/out": 4.776094436645508, "created_at": "2025-01-15T01:25:38.733385+00:00"} {"global_step": 8818, "acc_step": 0, "speed/wps": 12949.212495269046, "speed/FLOPS": 203385176320855.47, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.158699870109558, "optim/lr": 0.00296448936292212, "optim/total_tokens": 2311585792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450462, "loss/out": 4.872408390045166, "created_at": "2025-01-15T01:25:43.801729+00:00"} {"global_step": 8819, "acc_step": 0, "speed/wps": 12954.560330820803, "speed/FLOPS": 203469171426891.94, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7773849964141846, "optim/lr": 0.0029644708277285917, "optim/total_tokens": 2311847936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 4.948408603668213, "created_at": "2025-01-15T01:25:48.874116+00:00"} {"global_step": 8820, "acc_step": 0, "speed/wps": 12948.513374092036, "speed/FLOPS": 203374195662077.34, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4463006258010864, "optim/lr": 0.00296445228775697, "optim/total_tokens": 2312110080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459765, "loss/out": 4.807138919830322, "created_at": "2025-01-15T01:25:53.940052+00:00"} {"global_step": 8821, "acc_step": 0, "speed/wps": 12942.880951489333, "speed/FLOPS": 203285730725342.2, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.337232232093811, "optim/lr": 0.0029644337430073152, "optim/total_tokens": 2312372224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291378, "loss/out": 4.756613254547119, "created_at": "2025-01-15T01:25:59.008154+00:00"} {"global_step": 8822, "acc_step": 0, "speed/wps": 12945.69068424296, "speed/FLOPS": 203329861439215.28, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0336270332336426, "optim/lr": 0.002964415193479688, "optim/total_tokens": 2312634368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 4.930771827697754, "created_at": "2025-01-15T01:26:04.079745+00:00"} {"global_step": 8823, "acc_step": 0, "speed/wps": 12946.137034787085, "speed/FLOPS": 203336871987860.72, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3262089490890503, "optim/lr": 0.0029643966391741487, "optim/total_tokens": 2312896512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 4.822385311126709, "created_at": "2025-01-15T01:26:09.147940+00:00"} {"global_step": 8824, "acc_step": 0, "speed/wps": 12957.313714694264, "speed/FLOPS": 203512417104170.6, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.1672723293304443, "optim/lr": 0.0029643780800907575, "optim/total_tokens": 2313158656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429961, "loss/out": 4.807925701141357, "created_at": "2025-01-15T01:26:14.209361+00:00"} {"global_step": 8825, "acc_step": 0, "speed/wps": 12952.064339360335, "speed/FLOPS": 203429968451153.6, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.258617877960205, "optim/lr": 0.002964359516229576, "optim/total_tokens": 2313420800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 4.706972122192383, "created_at": "2025-01-15T01:26:19.273264+00:00"} {"global_step": 8826, "acc_step": 0, "speed/wps": 12957.489378117793, "speed/FLOPS": 203515176139623.94, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.4589593410491943, "optim/lr": 0.0029643409475906637, "optim/total_tokens": 2313682944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427351, "loss/out": 4.760746479034424, "created_at": "2025-01-15T01:26:24.334627+00:00"} {"global_step": 8827, "acc_step": 0, "speed/wps": 12952.747910373031, "speed/FLOPS": 203440704873233.0, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7324928045272827, "optim/lr": 0.002964322374174082, "optim/total_tokens": 2313945088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330338, "loss/out": 4.8113603591918945, "created_at": "2025-01-15T01:26:29.403409+00:00"} {"global_step": 8828, "acc_step": 0, "speed/wps": 12954.636017644541, "speed/FLOPS": 203470360192463.34, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.260505199432373, "optim/lr": 0.0029643037959798915, "optim/total_tokens": 2314207232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441679, "loss/out": 4.808259963989258, "created_at": "2025-01-15T01:26:34.471354+00:00"} {"global_step": 8829, "acc_step": 0, "speed/wps": 12945.50311849904, "speed/FLOPS": 203326915461464.7, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4010721445083618, "optim/lr": 0.002964285213008152, "optim/total_tokens": 2314469376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452325, "loss/out": 4.790913105010986, "created_at": "2025-01-15T01:26:39.542360+00:00"} {"global_step": 8830, "acc_step": 0, "speed/wps": 12953.390230999637, "speed/FLOPS": 203450793401314.9, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2076411247253418, "optim/lr": 0.002964266625258924, "optim/total_tokens": 2314731520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 4.856283187866211, "created_at": "2025-01-15T01:26:44.608976+00:00"} {"global_step": 8831, "acc_step": 0, "speed/wps": 12951.809785570835, "speed/FLOPS": 203425970334095.94, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.5668283700942993, "optim/lr": 0.002964248032732269, "optim/total_tokens": 2314993664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464535, "loss/out": 4.882730484008789, "created_at": "2025-01-15T01:26:49.681091+00:00"} {"global_step": 8832, "acc_step": 0, "speed/wps": 12950.187058027113, "speed/FLOPS": 203400483168160.47, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.341947317123413, "optim/lr": 0.0029642294354282482, "optim/total_tokens": 2315255808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302630, "loss/out": 4.682504653930664, "created_at": "2025-01-15T01:26:54.749483+00:00"} {"global_step": 8833, "acc_step": 0, "speed/wps": 12956.208710708135, "speed/FLOPS": 203495061498133.84, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2724111080169678, "optim/lr": 0.002964210833346921, "optim/total_tokens": 2315517952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431299, "loss/out": 4.800788879394531, "created_at": "2025-01-15T01:26:59.813831+00:00"} {"global_step": 8834, "acc_step": 0, "speed/wps": 12958.555194727147, "speed/FLOPS": 203531916254059.34, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.1738767623901367, "optim/lr": 0.002964192226488348, "optim/total_tokens": 2315780096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468018, "loss/out": 4.923598289489746, "created_at": "2025-01-15T01:27:04.875196+00:00"} {"global_step": 8835, "acc_step": 0, "speed/wps": 12955.39639781071, "speed/FLOPS": 203482303007844.88, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2636277675628662, "optim/lr": 0.0029641736148525912, "optim/total_tokens": 2316042240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 4.836580753326416, "created_at": "2025-01-15T01:27:09.946815+00:00"} {"global_step": 8836, "acc_step": 0, "speed/wps": 12955.47517371151, "speed/FLOPS": 203483540291616.34, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.8538413047790527, "optim/lr": 0.0029641549984397103, "optim/total_tokens": 2316304384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464007, "loss/out": 4.934812545776367, "created_at": "2025-01-15T01:27:15.008937+00:00"} {"global_step": 8837, "acc_step": 0, "speed/wps": 12953.729398402753, "speed/FLOPS": 203456120491445.7, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6077041625976562, "optim/lr": 0.0029641363772497666, "optim/total_tokens": 2316566528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451275, "loss/out": 4.7963786125183105, "created_at": "2025-01-15T01:27:20.077770+00:00"} {"global_step": 8838, "acc_step": 0, "speed/wps": 12956.181995754829, "speed/FLOPS": 203494641903081.97, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4162286520004272, "optim/lr": 0.00296411775128282, "optim/total_tokens": 2316828672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 284534, "loss/out": 4.804093360900879, "created_at": "2025-01-15T01:27:25.144142+00:00"} {"global_step": 8839, "acc_step": 0, "speed/wps": 12955.85437857711, "speed/FLOPS": 203489496225112.72, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.220540165901184, "optim/lr": 0.0029640991205389317, "optim/total_tokens": 2317090816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432146, "loss/out": 4.780088424682617, "created_at": "2025-01-15T01:27:30.211885+00:00"} {"global_step": 8840, "acc_step": 0, "speed/wps": 12956.285335215636, "speed/FLOPS": 203496264991317.25, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8378139734268188, "optim/lr": 0.0029640804850181637, "optim/total_tokens": 2317352960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 4.8045759201049805, "created_at": "2025-01-15T01:27:35.280010+00:00"} {"global_step": 8841, "acc_step": 0, "speed/wps": 12958.243397522736, "speed/FLOPS": 203527019050509.8, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.964005470275879, "optim/lr": 0.0029640618447205748, "optim/total_tokens": 2317615104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472260, "loss/out": 4.774560928344727, "created_at": "2025-01-15T01:27:40.341125+00:00"} {"global_step": 8842, "acc_step": 0, "speed/wps": 12962.810768521727, "speed/FLOPS": 203598755888274.44, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.577987551689148, "optim/lr": 0.002964043199646227, "optim/total_tokens": 2317877248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 4.763339996337891, "created_at": "2025-01-15T01:27:45.400724+00:00"} {"global_step": 8843, "acc_step": 0, "speed/wps": 12951.50541406938, "speed/FLOPS": 203421189761415.3, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.118121862411499, "optim/lr": 0.002964024549795181, "optim/total_tokens": 2318139392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296535, "loss/out": 4.83926248550415, "created_at": "2025-01-15T01:27:50.470337+00:00"} {"global_step": 8844, "acc_step": 0, "speed/wps": 12956.603434060162, "speed/FLOPS": 203501261170784.12, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4827921390533447, "optim/lr": 0.002964005895167497, "optim/total_tokens": 2318401536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293833, "loss/out": 4.842411041259766, "created_at": "2025-01-15T01:27:55.532048+00:00"} {"global_step": 8845, "acc_step": 0, "speed/wps": 12959.909200807104, "speed/FLOPS": 203553182780144.4, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 8.348485946655273, "optim/lr": 0.0029639872357632375, "optim/total_tokens": 2318663680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351876, "loss/out": 4.73712158203125, "created_at": "2025-01-15T01:28:00.596645+00:00"} {"global_step": 8846, "acc_step": 0, "speed/wps": 12957.166029878352, "speed/FLOPS": 203510097511196.47, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.517317771911621, "optim/lr": 0.0029639685715824615, "optim/total_tokens": 2318925824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431320, "loss/out": 4.988640785217285, "created_at": "2025-01-15T01:28:05.662353+00:00"} {"global_step": 8847, "acc_step": 0, "speed/wps": 12955.596825410094, "speed/FLOPS": 203485450998709.56, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.185594081878662, "optim/lr": 0.0029639499026252306, "optim/total_tokens": 2319187968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 4.767536163330078, "created_at": "2025-01-15T01:28:10.726679+00:00"} {"global_step": 8848, "acc_step": 0, "speed/wps": 12948.113968710662, "speed/FLOPS": 203367922451722.38, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.389650821685791, "optim/lr": 0.002963931228891606, "optim/total_tokens": 2319450112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438240, "loss/out": 4.867527484893799, "created_at": "2025-01-15T01:28:15.795644+00:00"} {"global_step": 8849, "acc_step": 0, "speed/wps": 12957.052794408484, "speed/FLOPS": 203508318992540.38, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.390514373779297, "optim/lr": 0.0029639125503816486, "optim/total_tokens": 2319712256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461605, "loss/out": 4.9399094581604, "created_at": "2025-01-15T01:28:20.862070+00:00"} {"global_step": 8850, "acc_step": 0, "speed/wps": 12955.74965773873, "speed/FLOPS": 203487851440444.44, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2517973184585571, "optim/lr": 0.002963893867095419, "optim/total_tokens": 2319974400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306620, "loss/out": 4.8233489990234375, "created_at": "2025-01-15T01:28:25.924208+00:00"} {"global_step": 8851, "acc_step": 0, "speed/wps": 12955.916910858514, "speed/FLOPS": 203490478380520.5, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3196064233779907, "optim/lr": 0.0029638751790329785, "optim/total_tokens": 2320236544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402407, "loss/out": 4.887897491455078, "created_at": "2025-01-15T01:28:30.986129+00:00"} {"global_step": 8852, "acc_step": 0, "speed/wps": 12949.770603061339, "speed/FLOPS": 203393942170653.9, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.145349383354187, "optim/lr": 0.0029638564861943872, "optim/total_tokens": 2320498688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457902, "loss/out": 4.759152412414551, "created_at": "2025-01-15T01:28:36.053188+00:00"} {"global_step": 8853, "acc_step": 0, "speed/wps": 12952.372787665543, "speed/FLOPS": 203434813055639.28, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2756444215774536, "optim/lr": 0.0029638377885797076, "optim/total_tokens": 2320760832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437195, "loss/out": 4.796193599700928, "created_at": "2025-01-15T01:28:41.123450+00:00"} {"global_step": 8854, "acc_step": 0, "speed/wps": 12952.223932858955, "speed/FLOPS": 203432475086353.88, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2911559343338013, "optim/lr": 0.0029638190861889994, "optim/total_tokens": 2321022976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 4.876469135284424, "created_at": "2025-01-15T01:28:46.190768+00:00"} {"global_step": 8855, "acc_step": 0, "speed/wps": 12951.01482115133, "speed/FLOPS": 203413484325492.03, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3167747259140015, "optim/lr": 0.0029638003790223243, "optim/total_tokens": 2321285120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298347, "loss/out": 4.776650428771973, "created_at": "2025-01-15T01:28:51.273489+00:00"} {"global_step": 8856, "acc_step": 0, "speed/wps": 12961.582852654423, "speed/FLOPS": 203579469782245.2, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2417434453964233, "optim/lr": 0.002963781667079743, "optim/total_tokens": 2321547264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425492, "loss/out": 4.795248985290527, "created_at": "2025-01-15T01:28:56.334367+00:00"} {"global_step": 8857, "acc_step": 0, "speed/wps": 12951.477922339222, "speed/FLOPS": 203420757966015.6, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1353944540023804, "optim/lr": 0.0029637629503613165, "optim/total_tokens": 2321809408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 4.951462268829346, "created_at": "2025-01-15T01:29:01.402900+00:00"} {"global_step": 8858, "acc_step": 0, "speed/wps": 12955.429141678567, "speed/FLOPS": 203482817295284.38, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.48486065864563, "optim/lr": 0.0029637442288671066, "optim/total_tokens": 2322071552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462938, "loss/out": 4.870073318481445, "created_at": "2025-01-15T01:29:06.465096+00:00"} {"global_step": 8859, "acc_step": 0, "speed/wps": 12959.102225533836, "speed/FLOPS": 203540508124577.5, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.5722346305847168, "optim/lr": 0.0029637255025971735, "optim/total_tokens": 2322333696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417230, "loss/out": 4.742650032043457, "created_at": "2025-01-15T01:29:11.531337+00:00"} {"global_step": 8860, "acc_step": 0, "speed/wps": 12958.627436974402, "speed/FLOPS": 203533050917823.75, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.365010142326355, "optim/lr": 0.0029637067715515787, "optim/total_tokens": 2322595840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 4.7749786376953125, "created_at": "2025-01-15T01:29:16.594761+00:00"} {"global_step": 8861, "acc_step": 0, "speed/wps": 12955.533983752906, "speed/FLOPS": 203484463984129.97, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.6199610233306885, "optim/lr": 0.002963688035730383, "optim/total_tokens": 2322857984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318319, "loss/out": 4.912236213684082, "created_at": "2025-01-15T01:29:21.660828+00:00"} {"global_step": 8862, "acc_step": 0, "speed/wps": 12955.2507339692, "speed/FLOPS": 203480015157050.88, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.9629642963409424, "optim/lr": 0.002963669295133648, "optim/total_tokens": 2323120128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 400171, "loss/out": 4.747196197509766, "created_at": "2025-01-15T01:29:26.722949+00:00"} {"global_step": 8863, "acc_step": 0, "speed/wps": 12957.722735702064, "speed/FLOPS": 203518841341151.84, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.4801740646362305, "optim/lr": 0.0029636505497614347, "optim/total_tokens": 2323382272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444641, "loss/out": 4.707051753997803, "created_at": "2025-01-15T01:29:31.784237+00:00"} {"global_step": 8864, "acc_step": 0, "speed/wps": 12952.621002642398, "speed/FLOPS": 203438711612933.9, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3943889141082764, "optim/lr": 0.002963631799613804, "optim/total_tokens": 2323644416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462435, "loss/out": 4.924296855926514, "created_at": "2025-01-15T01:29:36.850303+00:00"} {"global_step": 8865, "acc_step": 0, "speed/wps": 12962.09124659553, "speed/FLOPS": 203587454807696.75, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.820532321929932, "optim/lr": 0.0029636130446908174, "optim/total_tokens": 2323906560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450225, "loss/out": 4.884312629699707, "created_at": "2025-01-15T01:29:41.915331+00:00"} {"global_step": 8866, "acc_step": 0, "speed/wps": 12957.27808870488, "speed/FLOPS": 203511857549052.06, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7143981456756592, "optim/lr": 0.0029635942849925356, "optim/total_tokens": 2324168704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438773, "loss/out": 4.897997856140137, "created_at": "2025-01-15T01:29:46.982118+00:00"} {"global_step": 8867, "acc_step": 0, "speed/wps": 12953.873053568477, "speed/FLOPS": 203458376793226.25, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3596887588500977, "optim/lr": 0.0029635755205190207, "optim/total_tokens": 2324430848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310734, "loss/out": 4.877138137817383, "created_at": "2025-01-15T01:29:52.045111+00:00"} {"global_step": 8868, "acc_step": 0, "speed/wps": 12960.741642406594, "speed/FLOPS": 203566257419359.03, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2809559106826782, "optim/lr": 0.0029635567512703324, "optim/total_tokens": 2324692992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 398201, "loss/out": 4.846024513244629, "created_at": "2025-01-15T01:29:57.107105+00:00"} {"global_step": 8869, "acc_step": 0, "speed/wps": 12949.004344625271, "speed/FLOPS": 203381907028966.03, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7130614519119263, "optim/lr": 0.002963537977246534, "optim/total_tokens": 2324955136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 4.827117919921875, "created_at": "2025-01-15T01:30:02.176203+00:00"} {"global_step": 8870, "acc_step": 0, "speed/wps": 12951.699025136948, "speed/FLOPS": 203424230689280.97, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0011, "optim/grad_norm": 1.2337002754211426, "optim/lr": 0.0029635191984476853, "optim/total_tokens": 2325217280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440108, "loss/out": 4.966804504394531, "created_at": "2025-01-15T01:30:07.246626+00:00"} {"global_step": 8871, "acc_step": 0, "speed/wps": 12955.168884019722, "speed/FLOPS": 203478729591120.7, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3606890439987183, "optim/lr": 0.0029635004148738475, "optim/total_tokens": 2325479424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443016, "loss/out": 4.832582950592041, "created_at": "2025-01-15T01:30:12.308842+00:00"} {"global_step": 8872, "acc_step": 0, "speed/wps": 12950.853972735655, "speed/FLOPS": 203410957980094.97, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7296712398529053, "optim/lr": 0.0029634816265250827, "optim/total_tokens": 2325741568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 4.923295974731445, "created_at": "2025-01-15T01:30:17.375288+00:00"} {"global_step": 8873, "acc_step": 0, "speed/wps": 12952.567244811182, "speed/FLOPS": 203437867272323.66, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3766571283340454, "optim/lr": 0.0029634628334014525, "optim/total_tokens": 2326003712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312250, "loss/out": 4.931787967681885, "created_at": "2025-01-15T01:30:22.441227+00:00"} {"global_step": 8874, "acc_step": 0, "speed/wps": 12953.87302733764, "speed/FLOPS": 203458376381234.9, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.458539366722107, "optim/lr": 0.0029634440355030163, "optim/total_tokens": 2326265856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401934, "loss/out": 4.73817253112793, "created_at": "2025-01-15T01:30:27.504961+00:00"} {"global_step": 8875, "acc_step": 0, "speed/wps": 12959.77668694504, "speed/FLOPS": 203551101467847.56, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2688331604003906, "optim/lr": 0.002963425232829838, "optim/total_tokens": 2326528000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457902, "loss/out": 4.833975315093994, "created_at": "2025-01-15T01:30:32.571384+00:00"} {"global_step": 8876, "acc_step": 0, "speed/wps": 12959.239379467344, "speed/FLOPS": 203542662315572.5, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8383108377456665, "optim/lr": 0.002963406425381976, "optim/total_tokens": 2326790144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430567, "loss/out": 4.797172546386719, "created_at": "2025-01-15T01:30:37.639387+00:00"} {"global_step": 8877, "acc_step": 0, "speed/wps": 12950.817357355678, "speed/FLOPS": 203410382885237.62, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.6359786987304688, "optim/lr": 0.0029633876131594946, "optim/total_tokens": 2327052288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457380, "loss/out": 4.812472343444824, "created_at": "2025-01-15T01:30:42.703530+00:00"} {"global_step": 8878, "acc_step": 0, "speed/wps": 12952.120732078576, "speed/FLOPS": 203430854176285.3, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4650073051452637, "optim/lr": 0.0029633687961624533, "optim/total_tokens": 2327314432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314047, "loss/out": 4.803735733032227, "created_at": "2025-01-15T01:30:47.767738+00:00"} {"global_step": 8879, "acc_step": 0, "speed/wps": 12955.871672395808, "speed/FLOPS": 203489767848299.56, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 6.390439510345459, "optim/lr": 0.002963349974390914, "optim/total_tokens": 2327576576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 378157, "loss/out": 4.836705684661865, "created_at": "2025-01-15T01:30:52.833051+00:00"} {"global_step": 8880, "acc_step": 0, "speed/wps": 12958.905467249519, "speed/FLOPS": 203537417765348.06, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9434980154037476, "optim/lr": 0.0029633311478449386, "optim/total_tokens": 2327838720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458430, "loss/out": 4.819642543792725, "created_at": "2025-01-15T01:30:57.894014+00:00"} {"global_step": 8881, "acc_step": 0, "speed/wps": 12959.253727708798, "speed/FLOPS": 203542887674421.06, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.950963973999023, "optim/lr": 0.0029633123165245874, "optim/total_tokens": 2328100864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457124, "loss/out": 4.795916557312012, "created_at": "2025-01-15T01:31:02.959717+00:00"} {"global_step": 8882, "acc_step": 0, "speed/wps": 12949.660880774483, "speed/FLOPS": 203392218831364.62, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.429826021194458, "optim/lr": 0.0029632934804299225, "optim/total_tokens": 2328363008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 4.8015851974487305, "created_at": "2025-01-15T01:31:08.031524+00:00"} {"global_step": 8883, "acc_step": 0, "speed/wps": 12950.184777243187, "speed/FLOPS": 203400447345314.78, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1621931791305542, "optim/lr": 0.002963274639561006, "optim/total_tokens": 2328625152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308786, "loss/out": 4.740894317626953, "created_at": "2025-01-15T01:31:13.100834+00:00"} {"global_step": 8884, "acc_step": 0, "speed/wps": 12959.218475291525, "speed/FLOPS": 203542333986765.9, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1491869688034058, "optim/lr": 0.0029632557939178986, "optim/total_tokens": 2328887296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309845, "loss/out": 4.910843849182129, "created_at": "2025-01-15T01:31:18.166856+00:00"} {"global_step": 8885, "acc_step": 0, "speed/wps": 12956.247410491658, "speed/FLOPS": 203495669331413.12, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0938897132873535, "optim/lr": 0.002963236943500662, "optim/total_tokens": 2329149440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411794, "loss/out": 4.957237720489502, "created_at": "2025-01-15T01:31:23.237369+00:00"} {"global_step": 8886, "acc_step": 0, "speed/wps": 12956.968046844118, "speed/FLOPS": 203506987915586.66, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.001, "optim/grad_norm": 1.15572190284729, "optim/lr": 0.0029632180883093573, "optim/total_tokens": 2329411584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 4.6595964431762695, "created_at": "2025-01-15T01:31:28.300022+00:00"} {"global_step": 8887, "acc_step": 0, "speed/wps": 12943.696231208, "speed/FLOPS": 203298535813635.56, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.433351755142212, "optim/lr": 0.0029631992283440465, "optim/total_tokens": 2329673728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453660, "loss/out": 4.7562103271484375, "created_at": "2025-01-15T01:31:33.367812+00:00"} {"global_step": 8888, "acc_step": 0, "speed/wps": 12956.463976580977, "speed/FLOPS": 203499070799438.88, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4405391216278076, "optim/lr": 0.0029631803636047914, "optim/total_tokens": 2329935872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444067, "loss/out": 4.801692962646484, "created_at": "2025-01-15T01:31:38.434610+00:00"} {"global_step": 8889, "acc_step": 0, "speed/wps": 12959.41720014454, "speed/FLOPS": 203545455233659.1, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.543455719947815, "optim/lr": 0.0029631614940916528, "optim/total_tokens": 2330198016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296240, "loss/out": 4.805081367492676, "created_at": "2025-01-15T01:31:43.503521+00:00"} {"global_step": 8890, "acc_step": 0, "speed/wps": 12936.151791582228, "speed/FLOPS": 203180040022166.38, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.196766376495361, "optim/lr": 0.002963142619804692, "optim/total_tokens": 2330460160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447839, "loss/out": 4.865462303161621, "created_at": "2025-01-15T01:31:48.578000+00:00"} {"global_step": 8891, "acc_step": 0, "speed/wps": 12951.717385354235, "speed/FLOPS": 203424519061723.03, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6037958860397339, "optim/lr": 0.002963123740743972, "optim/total_tokens": 2330722304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439302, "loss/out": 4.838021278381348, "created_at": "2025-01-15T01:31:53.647475+00:00"} {"global_step": 8892, "acc_step": 0, "speed/wps": 12952.43865029821, "speed/FLOPS": 203435847518790.06, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.344119071960449, "optim/lr": 0.0029631048569095536, "optim/total_tokens": 2330984448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 4.809158802032471, "created_at": "2025-01-15T01:31:58.718233+00:00"} {"global_step": 8893, "acc_step": 0, "speed/wps": 12958.41089877986, "speed/FLOPS": 203529649887923.8, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2562788724899292, "optim/lr": 0.002963085968301498, "optim/total_tokens": 2331246592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324468, "loss/out": 4.810088157653809, "created_at": "2025-01-15T01:32:03.781003+00:00"} {"global_step": 8894, "acc_step": 0, "speed/wps": 12956.39602143099, "speed/FLOPS": 203498003470428.16, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.138521432876587, "optim/lr": 0.0029630670749198674, "optim/total_tokens": 2331508736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367900, "loss/out": 4.740393161773682, "created_at": "2025-01-15T01:32:08.843661+00:00"} {"global_step": 8895, "acc_step": 0, "speed/wps": 12957.060982453275, "speed/FLOPS": 203508447597035.34, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.3892738819122314, "optim/lr": 0.0029630481767647233, "optim/total_tokens": 2331770880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420669, "loss/out": 5.13151741027832, "created_at": "2025-01-15T01:32:13.907910+00:00"} {"global_step": 8896, "acc_step": 0, "speed/wps": 12949.15155750012, "speed/FLOPS": 203384219209457.22, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2191417217254639, "optim/lr": 0.0029630292738361272, "optim/total_tokens": 2332033024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 4.838268280029297, "created_at": "2025-01-15T01:32:18.975300+00:00"} {"global_step": 8897, "acc_step": 0, "speed/wps": 12960.258154927038, "speed/FLOPS": 203558663584111.2, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1223299503326416, "optim/lr": 0.0029630103661341415, "optim/total_tokens": 2332295168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453660, "loss/out": 4.777998924255371, "created_at": "2025-01-15T01:32:24.047587+00:00"} {"global_step": 8898, "acc_step": 0, "speed/wps": 12955.566418234377, "speed/FLOPS": 203484973412231.12, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1819111108779907, "optim/lr": 0.0029629914536588265, "optim/total_tokens": 2332557312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 4.918426513671875, "created_at": "2025-01-15T01:32:29.110882+00:00"} {"global_step": 8899, "acc_step": 0, "speed/wps": 12959.495535678538, "speed/FLOPS": 203546685600864.8, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1623541116714478, "optim/lr": 0.002962972536410245, "optim/total_tokens": 2332819456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459507, "loss/out": 4.784660339355469, "created_at": "2025-01-15T01:32:34.202598+00:00"} {"global_step": 8900, "acc_step": 0, "speed/wps": 12960.569953613565, "speed/FLOPS": 203563560810939.38, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2251379489898682, "optim/lr": 0.0029629536143884585, "optim/total_tokens": 2333081600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 4.8974995613098145, "created_at": "2025-01-15T01:32:39.263614+00:00"} {"global_step": 8901, "acc_step": 0, "speed/wps": 12950.091070369039, "speed/FLOPS": 203398975550089.56, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0822129249572754, "optim/lr": 0.0029629346875935285, "optim/total_tokens": 2333343744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 4.847409248352051, "created_at": "2025-01-15T01:32:44.332436+00:00"} {"global_step": 8902, "acc_step": 0, "speed/wps": 12959.470193831568, "speed/FLOPS": 203546287572334.8, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8873976469039917, "optim/lr": 0.002962915756025517, "optim/total_tokens": 2333605888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 4.911185264587402, "created_at": "2025-01-15T01:32:49.398809+00:00"} {"global_step": 8903, "acc_step": 0, "speed/wps": 12957.342872703446, "speed/FLOPS": 203512875070772.1, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.962324619293213, "optim/lr": 0.0029628968196844857, "optim/total_tokens": 2333868032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 4.810273170471191, "created_at": "2025-01-15T01:32:54.465499+00:00"} {"global_step": 8904, "acc_step": 0, "speed/wps": 12952.537931865652, "speed/FLOPS": 203437406872233.9, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.461155652999878, "optim/lr": 0.002962877878570496, "optim/total_tokens": 2334130176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309674, "loss/out": 4.887583255767822, "created_at": "2025-01-15T01:32:59.531539+00:00"} {"global_step": 8905, "acc_step": 0, "speed/wps": 12954.11814245933, "speed/FLOPS": 203462226251044.53, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.3392959833145142, "optim/lr": 0.00296285893268361, "optim/total_tokens": 2334392320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432127, "loss/out": 4.979987621307373, "created_at": "2025-01-15T01:33:04.597771+00:00"} {"global_step": 8906, "acc_step": 0, "speed/wps": 12949.288345994923, "speed/FLOPS": 203386367660735.2, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0742676258087158, "optim/lr": 0.0029628399820238903, "optim/total_tokens": 2334654464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458173, "loss/out": 4.769525527954102, "created_at": "2025-01-15T01:33:09.663639+00:00"} {"global_step": 8907, "acc_step": 0, "speed/wps": 12958.963823979282, "speed/FLOPS": 203538334337979.56, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.115917682647705, "optim/lr": 0.0029628210265913975, "optim/total_tokens": 2334916608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465633, "loss/out": 4.726196765899658, "created_at": "2025-01-15T01:33:14.729280+00:00"} {"global_step": 8908, "acc_step": 0, "speed/wps": 12948.524653560926, "speed/FLOPS": 203374372821635.88, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.0050275325775146, "optim/lr": 0.0029628020663861943, "optim/total_tokens": 2335178752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438536, "loss/out": 4.879726886749268, "created_at": "2025-01-15T01:33:19.794219+00:00"} {"global_step": 8909, "acc_step": 0, "speed/wps": 12955.90382709454, "speed/FLOPS": 203490272882028.34, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3735100030899048, "optim/lr": 0.002962783101408342, "optim/total_tokens": 2335440896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306381, "loss/out": 4.866039752960205, "created_at": "2025-01-15T01:33:24.861821+00:00"} {"global_step": 8910, "acc_step": 0, "speed/wps": 12959.145383997418, "speed/FLOPS": 203541185987554.0, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.238419532775879, "optim/lr": 0.0029627641316579026, "optim/total_tokens": 2335703040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402512, "loss/out": 4.900906562805176, "created_at": "2025-01-15T01:33:29.923780+00:00"} {"global_step": 8911, "acc_step": 0, "speed/wps": 12953.938416376246, "speed/FLOPS": 203459403405940.2, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1385931968688965, "optim/lr": 0.002962745157134938, "optim/total_tokens": 2335965184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441962, "loss/out": 4.814258575439453, "created_at": "2025-01-15T01:33:34.990794+00:00"} {"global_step": 8912, "acc_step": 0, "speed/wps": 12957.479376209661, "speed/FLOPS": 203515019045912.7, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.4448951482772827, "optim/lr": 0.00296272617783951, "optim/total_tokens": 2336227328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494003, "loss/out": 4.597635746002197, "created_at": "2025-01-15T01:33:40.052085+00:00"} {"global_step": 8913, "acc_step": 0, "speed/wps": 12965.909673906439, "speed/FLOPS": 203647428455683.56, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4726803302764893, "optim/lr": 0.002962707193771681, "optim/total_tokens": 2336489472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 4.858110427856445, "created_at": "2025-01-15T01:33:45.113751+00:00"} {"global_step": 8914, "acc_step": 0, "speed/wps": 12953.341311934591, "speed/FLOPS": 203450025060176.5, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.412455677986145, "optim/lr": 0.002962688204931513, "optim/total_tokens": 2336751616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 4.915039539337158, "created_at": "2025-01-15T01:33:50.177002+00:00"} {"global_step": 8915, "acc_step": 0, "speed/wps": 12963.55985723988, "speed/FLOPS": 203610521355948.78, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2816250324249268, "optim/lr": 0.0029626692113190675, "optim/total_tokens": 2337013760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442778, "loss/out": 4.789034843444824, "created_at": "2025-01-15T01:33:55.238355+00:00"} {"global_step": 8916, "acc_step": 0, "speed/wps": 12955.918757138366, "speed/FLOPS": 203490507378882.66, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0809268951416016, "optim/lr": 0.0029626502129344065, "optim/total_tokens": 2337275904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334608, "loss/out": 4.794279098510742, "created_at": "2025-01-15T01:34:00.306105+00:00"} {"global_step": 8917, "acc_step": 0, "speed/wps": 12961.164740553175, "speed/FLOPS": 203572902757148.84, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.3833929300308228, "optim/lr": 0.0029626312097775914, "optim/total_tokens": 2337538048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443016, "loss/out": 4.818229675292969, "created_at": "2025-01-15T01:34:05.366995+00:00"} {"global_step": 8918, "acc_step": 0, "speed/wps": 12963.239007167991, "speed/FLOPS": 203605481964675.56, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.224028468132019, "optim/lr": 0.0029626122018486852, "optim/total_tokens": 2337800192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 4.885523796081543, "created_at": "2025-01-15T01:34:10.430373+00:00"} {"global_step": 8919, "acc_step": 0, "speed/wps": 12958.34335750634, "speed/FLOPS": 203528589059411.75, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.021362781524658, "optim/lr": 0.00296259318914775, "optim/total_tokens": 2338062336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435329, "loss/out": 4.902215957641602, "created_at": "2025-01-15T01:34:15.496568+00:00"} {"global_step": 8920, "acc_step": 0, "speed/wps": 12959.407109340267, "speed/FLOPS": 203545296743711.75, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4599567651748657, "optim/lr": 0.002962574171674847, "optim/total_tokens": 2338324480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 4.896261215209961, "created_at": "2025-01-15T01:34:20.563795+00:00"} {"global_step": 8921, "acc_step": 0, "speed/wps": 12964.118896764774, "speed/FLOPS": 203619301839888.06, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.304448127746582, "optim/lr": 0.0029625551494300386, "optim/total_tokens": 2338586624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317507, "loss/out": 4.882552146911621, "created_at": "2025-01-15T01:34:25.639740+00:00"} {"global_step": 8922, "acc_step": 0, "speed/wps": 12960.946912004943, "speed/FLOPS": 203569481460471.1, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2980951070785522, "optim/lr": 0.002962536122413387, "optim/total_tokens": 2338848768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441971, "loss/out": 4.808399677276611, "created_at": "2025-01-15T01:34:30.700318+00:00"} {"global_step": 8923, "acc_step": 0, "speed/wps": 12956.035388704997, "speed/FLOPS": 203492339237905.16, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.7297921180725098, "optim/lr": 0.0029625170906249546, "optim/total_tokens": 2339110912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471976, "loss/out": 4.762959003448486, "created_at": "2025-01-15T01:34:35.762186+00:00"} {"global_step": 8924, "acc_step": 0, "speed/wps": 12961.986687546432, "speed/FLOPS": 203585812564151.5, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.083003282546997, "optim/lr": 0.002962498054064803, "optim/total_tokens": 2339373056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474123, "loss/out": 4.782871246337891, "created_at": "2025-01-15T01:34:40.821671+00:00"} {"global_step": 8925, "acc_step": 0, "speed/wps": 12957.539507188649, "speed/FLOPS": 203515963485566.28, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.31026029586792, "optim/lr": 0.002962479012732994, "optim/total_tokens": 2339635200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 4.709827423095703, "created_at": "2025-01-15T01:34:45.889400+00:00"} {"global_step": 8926, "acc_step": 0, "speed/wps": 12959.821231865173, "speed/FLOPS": 203551801107029.38, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.259628176689148, "optim/lr": 0.0029624599666295904, "optim/total_tokens": 2339897344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455808, "loss/out": 4.79074764251709, "created_at": "2025-01-15T01:34:50.951044+00:00"} {"global_step": 8927, "acc_step": 0, "speed/wps": 12956.449834904284, "speed/FLOPS": 203498848684973.72, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.9021334648132324, "optim/lr": 0.0029624409157546537, "optim/total_tokens": 2340159488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459765, "loss/out": 4.806121826171875, "created_at": "2025-01-15T01:34:56.013203+00:00"} {"global_step": 8928, "acc_step": 0, "speed/wps": 12962.077715806974, "speed/FLOPS": 203587242288069.25, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.297913908958435, "optim/lr": 0.0029624218601082466, "optim/total_tokens": 2340421632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323417, "loss/out": 4.736728668212891, "created_at": "2025-01-15T01:35:01.079153+00:00"} {"global_step": 8929, "acc_step": 0, "speed/wps": 12959.019019099407, "speed/FLOPS": 203539201253187.16, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.358110427856445, "optim/lr": 0.0029624027996904313, "optim/total_tokens": 2340683776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440339, "loss/out": 4.699049949645996, "created_at": "2025-01-15T01:35:06.142104+00:00"} {"global_step": 8930, "acc_step": 0, "speed/wps": 12959.86244269303, "speed/FLOPS": 203552448379710.47, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.3853132724761963, "optim/lr": 0.0029623837345012696, "optim/total_tokens": 2340945920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 4.718236923217773, "created_at": "2025-01-15T01:35:11.203591+00:00"} {"global_step": 8931, "acc_step": 0, "speed/wps": 12950.791074590796, "speed/FLOPS": 203409970078298.84, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.63382887840271, "optim/lr": 0.002962364664540824, "optim/total_tokens": 2341208064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437432, "loss/out": 4.810347080230713, "created_at": "2025-01-15T01:35:16.272269+00:00"} {"global_step": 8932, "acc_step": 0, "speed/wps": 12956.014451927042, "speed/FLOPS": 203492010397037.22, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.001, "optim/grad_norm": 1.1533262729644775, "optim/lr": 0.0029623455898091567, "optim/total_tokens": 2341470208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 4.9341840744018555, "created_at": "2025-01-15T01:35:21.334295+00:00"} {"global_step": 8933, "acc_step": 0, "speed/wps": 12958.853917213588, "speed/FLOPS": 203536608101196.62, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9407052993774414, "optim/lr": 0.0029623265103063292, "optim/total_tokens": 2341732352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436381, "loss/out": 4.817883491516113, "created_at": "2025-01-15T01:35:26.400808+00:00"} {"global_step": 8934, "acc_step": 0, "speed/wps": 12949.450897491033, "speed/FLOPS": 203388920755350.75, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.763082504272461, "optim/lr": 0.002962307426032405, "optim/total_tokens": 2341994496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 408639, "loss/out": 4.91354513168335, "created_at": "2025-01-15T01:35:31.469607+00:00"} {"global_step": 8935, "acc_step": 0, "speed/wps": 12955.48725958939, "speed/FLOPS": 203483730116936.22, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0756458044052124, "optim/lr": 0.002962288336987446, "optim/total_tokens": 2342256640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471217, "loss/out": 4.883459568023682, "created_at": "2025-01-15T01:35:36.540539+00:00"} {"global_step": 8936, "acc_step": 0, "speed/wps": 12952.158768535108, "speed/FLOPS": 203431451591102.7, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1550308465957642, "optim/lr": 0.002962269243171514, "optim/total_tokens": 2342518784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450994, "loss/out": 4.874782085418701, "created_at": "2025-01-15T01:35:41.609452+00:00"} {"global_step": 8937, "acc_step": 0, "speed/wps": 12959.316505218672, "speed/FLOPS": 203543873681479.0, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.508406400680542, "optim/lr": 0.0029622501445846715, "optim/total_tokens": 2342780928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480757, "loss/out": 4.780002593994141, "created_at": "2025-01-15T01:35:46.670452+00:00"} {"global_step": 8938, "acc_step": 0, "speed/wps": 12955.747603984944, "speed/FLOPS": 203487819183419.1, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0255420207977295, "optim/lr": 0.0029622310412269807, "optim/total_tokens": 2343043072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475904, "loss/out": 4.809823989868164, "created_at": "2025-01-15T01:35:51.736368+00:00"} {"global_step": 8939, "acc_step": 0, "speed/wps": 12960.57945184304, "speed/FLOPS": 203563709993685.22, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5447328090667725, "optim/lr": 0.002962211933098504, "optim/total_tokens": 2343305216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322156, "loss/out": 4.7135910987854, "created_at": "2025-01-15T01:35:56.801373+00:00"} {"global_step": 8940, "acc_step": 0, "speed/wps": 12959.464902558077, "speed/FLOPS": 203546204465613.66, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.0812464952468872, "optim/lr": 0.002962192820199304, "optim/total_tokens": 2343567360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375999, "loss/out": 4.822531700134277, "created_at": "2025-01-15T01:36:01.862907+00:00"} {"global_step": 8941, "acc_step": 0, "speed/wps": 12958.646819119578, "speed/FLOPS": 203533355341047.6, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8815019130706787, "optim/lr": 0.0029621737025294433, "optim/total_tokens": 2343829504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447791, "loss/out": 4.708227157592773, "created_at": "2025-01-15T01:36:06.930557+00:00"} {"global_step": 8942, "acc_step": 0, "speed/wps": 12960.678726958045, "speed/FLOPS": 203565269245784.7, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.087569236755371, "optim/lr": 0.002962154580088983, "optim/total_tokens": 2344091648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 4.6510009765625, "created_at": "2025-01-15T01:36:11.994148+00:00"} {"global_step": 8943, "acc_step": 0, "speed/wps": 12956.397765268692, "speed/FLOPS": 203498030859795.56, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.212046504020691, "optim/lr": 0.002962135452877987, "optim/total_tokens": 2344353792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463248, "loss/out": 4.718587875366211, "created_at": "2025-01-15T01:36:17.065226+00:00"} {"global_step": 8944, "acc_step": 0, "speed/wps": 12961.735133866383, "speed/FLOPS": 203581861567935.34, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1369736194610596, "optim/lr": 0.0029621163208965167, "optim/total_tokens": 2344615936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443016, "loss/out": 4.765673637390137, "created_at": "2025-01-15T01:36:22.127775+00:00"} {"global_step": 8945, "acc_step": 0, "speed/wps": 12959.002101105534, "speed/FLOPS": 203538935532845.62, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.848653793334961, "optim/lr": 0.002962097184144635, "optim/total_tokens": 2344878080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307440, "loss/out": 4.832009315490723, "created_at": "2025-01-15T01:36:27.192606+00:00"} {"global_step": 8946, "acc_step": 0, "speed/wps": 12945.842999523222, "speed/FLOPS": 203332253759994.94, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4174656867980957, "optim/lr": 0.0029620780426224042, "optim/total_tokens": 2345140224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455769, "loss/out": 4.88040828704834, "created_at": "2025-01-15T01:36:32.259132+00:00"} {"global_step": 8947, "acc_step": 0, "speed/wps": 12956.863537031086, "speed/FLOPS": 203505346445361.44, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2684568166732788, "optim/lr": 0.0029620588963298867, "optim/total_tokens": 2345402368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 4.695043087005615, "created_at": "2025-01-15T01:36:37.328289+00:00"} {"global_step": 8948, "acc_step": 0, "speed/wps": 12952.079136590433, "speed/FLOPS": 203430200861985.8, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.108432412147522, "optim/lr": 0.0029620397452671454, "optim/total_tokens": 2345664512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 4.816201686859131, "created_at": "2025-01-15T01:36:42.399720+00:00"} {"global_step": 8949, "acc_step": 0, "speed/wps": 12955.357319098819, "speed/FLOPS": 203481689222974.84, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.2435253858566284, "optim/lr": 0.002962020589434242, "optim/total_tokens": 2345926656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297588, "loss/out": 4.8732218742370605, "created_at": "2025-01-15T01:36:47.467558+00:00"} {"global_step": 8950, "acc_step": 0, "speed/wps": 12964.193792497677, "speed/FLOPS": 203620478180290.8, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.713073968887329, "optim/lr": 0.0029620014288312397, "optim/total_tokens": 2346188800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313309, "loss/out": 4.9110941886901855, "created_at": "2025-01-15T01:36:52.530056+00:00"} {"global_step": 8951, "acc_step": 0, "speed/wps": 12959.462336876726, "speed/FLOPS": 203546164168062.44, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.459546446800232, "optim/lr": 0.0029619822634582007, "optim/total_tokens": 2346450944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351583, "loss/out": 4.7409257888793945, "created_at": "2025-01-15T01:36:57.591608+00:00"} {"global_step": 8952, "acc_step": 0, "speed/wps": 12961.317236352854, "speed/FLOPS": 203575297913234.0, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.3311210870742798, "optim/lr": 0.002961963093315187, "optim/total_tokens": 2346713088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446742, "loss/out": 4.798015117645264, "created_at": "2025-01-15T01:37:02.652898+00:00"} {"global_step": 8953, "acc_step": 0, "speed/wps": 12963.597466554595, "speed/FLOPS": 203611112061917.0, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2500989437103271, "optim/lr": 0.002961943918402262, "optim/total_tokens": 2346975232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459528, "loss/out": 4.700214385986328, "created_at": "2025-01-15T01:37:07.719979+00:00"} {"global_step": 8954, "acc_step": 0, "speed/wps": 12959.028978040704, "speed/FLOPS": 203539357672045.25, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.091992735862732, "optim/lr": 0.0029619247387194883, "optim/total_tokens": 2347237376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453911, "loss/out": 4.822822570800781, "created_at": "2025-01-15T01:37:12.783556+00:00"} {"global_step": 8955, "acc_step": 0, "speed/wps": 12951.035384237139, "speed/FLOPS": 203413807297011.3, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1022175550460815, "optim/lr": 0.0029619055542669283, "optim/total_tokens": 2347499520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 4.784460067749023, "created_at": "2025-01-15T01:37:17.853182+00:00"} {"global_step": 8956, "acc_step": 0, "speed/wps": 12959.397715455238, "speed/FLOPS": 203545149199838.8, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2239478826522827, "optim/lr": 0.0029618863650446438, "optim/total_tokens": 2347761664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427855, "loss/out": 4.7392578125, "created_at": "2025-01-15T01:37:22.914564+00:00"} {"global_step": 8957, "acc_step": 0, "speed/wps": 12958.328238131722, "speed/FLOPS": 203528351588857.16, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.356801748275757, "optim/lr": 0.0029618671710526985, "optim/total_tokens": 2348023808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 344264, "loss/out": 4.859574317932129, "created_at": "2025-01-15T01:37:27.976050+00:00"} {"global_step": 8958, "acc_step": 0, "speed/wps": 12956.754286165424, "speed/FLOPS": 203503630510389.78, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8930081129074097, "optim/lr": 0.002961847972291154, "optim/total_tokens": 2348285952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 4.81669807434082, "created_at": "2025-01-15T01:37:33.040384+00:00"} {"global_step": 8959, "acc_step": 0, "speed/wps": 12953.147809131573, "speed/FLOPS": 203446985832754.38, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.065716028213501, "optim/lr": 0.002961828768760074, "optim/total_tokens": 2348548096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 4.826014995574951, "created_at": "2025-01-15T01:37:38.107755+00:00"} {"global_step": 8960, "acc_step": 0, "speed/wps": 12958.125702524907, "speed/FLOPS": 203525170488839.03, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1755625009536743, "optim/lr": 0.0029618095604595205, "optim/total_tokens": 2348810240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438769, "loss/out": 4.773137092590332, "created_at": "2025-01-15T01:37:43.173457+00:00"} {"global_step": 8961, "acc_step": 0, "speed/wps": 12953.914784580853, "speed/FLOPS": 203459032236120.03, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1092731952667236, "optim/lr": 0.0029617903473895565, "optim/total_tokens": 2349072384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 4.764283180236816, "created_at": "2025-01-15T01:37:48.242259+00:00"} {"global_step": 8962, "acc_step": 0, "speed/wps": 12957.931023565578, "speed/FLOPS": 203522112788265.84, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2927340269088745, "optim/lr": 0.002961771129550245, "optim/total_tokens": 2349334528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310665, "loss/out": 4.847727298736572, "created_at": "2025-01-15T01:37:53.304723+00:00"} {"global_step": 8963, "acc_step": 0, "speed/wps": 12953.02974412149, "speed/FLOPS": 203445131459532.72, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.092686414718628, "optim/lr": 0.002961751906941647, "optim/total_tokens": 2349596672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 344099, "loss/out": 4.842609882354736, "created_at": "2025-01-15T01:37:58.368071+00:00"} {"global_step": 8964, "acc_step": 0, "speed/wps": 12955.830768414758, "speed/FLOPS": 203489125395069.22, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.13429856300354, "optim/lr": 0.0029617326795638274, "optim/total_tokens": 2349858816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352687, "loss/out": 4.80180549621582, "created_at": "2025-01-15T01:38:03.434550+00:00"} {"global_step": 8965, "acc_step": 0, "speed/wps": 12956.401434247497, "speed/FLOPS": 203498088486149.4, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.4668080806732178, "optim/lr": 0.0029617134474168474, "optim/total_tokens": 2350120960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 4.658764839172363, "created_at": "2025-01-15T01:38:08.497249+00:00"} {"global_step": 8966, "acc_step": 0, "speed/wps": 12956.84374987695, "speed/FLOPS": 203505035660915.34, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.179840087890625, "optim/lr": 0.002961694210500771, "optim/total_tokens": 2350383104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 4.781733512878418, "created_at": "2025-01-15T01:38:13.560044+00:00"} {"global_step": 8967, "acc_step": 0, "speed/wps": 12956.748970866991, "speed/FLOPS": 203503547026323.9, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2500293254852295, "optim/lr": 0.0029616749688156593, "optim/total_tokens": 2350645248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 4.831271648406982, "created_at": "2025-01-15T01:38:18.622857+00:00"} {"global_step": 8968, "acc_step": 0, "speed/wps": 12956.745867097346, "speed/FLOPS": 203503498277356.62, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 23.732711791992188, "optim/lr": 0.002961655722361577, "optim/total_tokens": 2350907392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 4.769676208496094, "created_at": "2025-01-15T01:38:23.689795+00:00"} {"global_step": 8969, "acc_step": 0, "speed/wps": 12954.949625606385, "speed/FLOPS": 203475285836446.6, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2664241790771484, "optim/lr": 0.0029616364711385857, "optim/total_tokens": 2351169536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435044, "loss/out": 4.751437187194824, "created_at": "2025-01-15T01:38:28.753267+00:00"} {"global_step": 8970, "acc_step": 0, "speed/wps": 12961.517916428897, "speed/FLOPS": 203578449869591.56, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0850261449813843, "optim/lr": 0.0029616172151467482, "optim/total_tokens": 2351431680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302630, "loss/out": 4.8463358879089355, "created_at": "2025-01-15T01:38:33.813326+00:00"} {"global_step": 8971, "acc_step": 0, "speed/wps": 12947.969352470353, "speed/FLOPS": 203365651054944.78, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8645589351654053, "optim/lr": 0.002961597954386128, "optim/total_tokens": 2351693824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433167, "loss/out": 4.819953441619873, "created_at": "2025-01-15T01:38:38.878652+00:00"} {"global_step": 8972, "acc_step": 0, "speed/wps": 12955.864912296216, "speed/FLOPS": 203489661671646.12, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.184956669807434, "optim/lr": 0.002961578688856787, "optim/total_tokens": 2351955968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 4.789504051208496, "created_at": "2025-01-15T01:38:43.947134+00:00"} {"global_step": 8973, "acc_step": 0, "speed/wps": 12957.845768206893, "speed/FLOPS": 203520773735704.56, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1172447204589844, "optim/lr": 0.0029615594185587886, "optim/total_tokens": 2352218112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432656, "loss/out": 4.765713691711426, "created_at": "2025-01-15T01:38:49.008394+00:00"} {"global_step": 8974, "acc_step": 0, "speed/wps": 12964.036003258678, "speed/FLOPS": 203617999883467.2, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.6109001636505127, "optim/lr": 0.0029615401434921963, "optim/total_tokens": 2352480256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 4.8576507568359375, "created_at": "2025-01-15T01:38:54.072756+00:00"} {"global_step": 8975, "acc_step": 0, "speed/wps": 12953.37821341384, "speed/FLOPS": 203450604648616.06, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9771481156349182, "optim/lr": 0.002961520863657072, "optim/total_tokens": 2352742400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434531, "loss/out": 4.864192485809326, "created_at": "2025-01-15T01:38:59.135817+00:00"} {"global_step": 8976, "acc_step": 0, "speed/wps": 12962.745633736808, "speed/FLOPS": 203597732856972.7, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3853859901428223, "optim/lr": 0.002961501579053479, "optim/total_tokens": 2353004544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331148, "loss/out": 4.80744743347168, "created_at": "2025-01-15T01:39:04.195465+00:00"} {"global_step": 8977, "acc_step": 0, "speed/wps": 12962.479850387243, "speed/FLOPS": 203593558364243.2, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.160186767578125, "optim/lr": 0.0029614822896814803, "optim/total_tokens": 2353266688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 280775, "loss/out": 4.84742546081543, "created_at": "2025-01-15T01:39:09.262038+00:00"} {"global_step": 8978, "acc_step": 0, "speed/wps": 12956.798213886252, "speed/FLOPS": 203504320455608.4, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9968134164810181, "optim/lr": 0.002961462995541138, "optim/total_tokens": 2353528832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315023, "loss/out": 4.792109489440918, "created_at": "2025-01-15T01:39:14.326412+00:00"} {"global_step": 8979, "acc_step": 0, "speed/wps": 12959.01954651959, "speed/FLOPS": 203539209537045.9, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3629721403121948, "optim/lr": 0.0029614436966325162, "optim/total_tokens": 2353790976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452847, "loss/out": 4.808622360229492, "created_at": "2025-01-15T01:39:19.390558+00:00"} {"global_step": 8980, "acc_step": 0, "speed/wps": 12960.275995187696, "speed/FLOPS": 203558943789919.9, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.3460564613342285, "optim/lr": 0.002961424392955678, "optim/total_tokens": 2354053120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461100, "loss/out": 4.696805000305176, "created_at": "2025-01-15T01:39:24.458260+00:00"} {"global_step": 8981, "acc_step": 0, "speed/wps": 12960.918136438206, "speed/FLOPS": 203569029500653.7, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0518008470535278, "optim/lr": 0.0029614050845106856, "optim/total_tokens": 2354315264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440105, "loss/out": 4.725311279296875, "created_at": "2025-01-15T01:39:29.520692+00:00"} {"global_step": 8982, "acc_step": 0, "speed/wps": 12950.745040085825, "speed/FLOPS": 203409247043140.25, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0677059888839722, "optim/lr": 0.0029613857712976017, "optim/total_tokens": 2354577408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 4.773646831512451, "created_at": "2025-01-15T01:39:34.589215+00:00"} {"global_step": 8983, "acc_step": 0, "speed/wps": 12956.418980010369, "speed/FLOPS": 203498364066465.72, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2699183225631714, "optim/lr": 0.0029613664533164902, "optim/total_tokens": 2354839552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443022, "loss/out": 4.7879509925842285, "created_at": "2025-01-15T01:39:39.653273+00:00"} {"global_step": 8984, "acc_step": 0, "speed/wps": 12960.648415278763, "speed/FLOPS": 203564793159208.94, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0013, "optim/grad_norm": 3.5605838298797607, "optim/lr": 0.0029613471305674137, "optim/total_tokens": 2355101696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305807, "loss/out": 4.733551979064941, "created_at": "2025-01-15T01:39:44.716821+00:00"} {"global_step": 8985, "acc_step": 0, "speed/wps": 12953.816197285301, "speed/FLOPS": 203457483787170.25, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2938635349273682, "optim/lr": 0.0029613278030504353, "optim/total_tokens": 2355363840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441158, "loss/out": 4.834157466888428, "created_at": "2025-01-15T01:39:49.784842+00:00"} {"global_step": 8986, "acc_step": 0, "speed/wps": 12955.311047003108, "speed/FLOPS": 203480962456127.2, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9702240228652954, "optim/lr": 0.002961308470765618, "optim/total_tokens": 2355625984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 4.860547065734863, "created_at": "2025-01-15T01:39:54.852124+00:00"} {"global_step": 8987, "acc_step": 0, "speed/wps": 12959.862256520508, "speed/FLOPS": 203552445455615.2, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0814708471298218, "optim/lr": 0.002961289133713025, "optim/total_tokens": 2355888128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436096, "loss/out": 4.792910099029541, "created_at": "2025-01-15T01:39:59.917575+00:00"} {"global_step": 8988, "acc_step": 0, "speed/wps": 12957.773818615593, "speed/FLOPS": 203519643668504.1, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0802948474884033, "optim/lr": 0.0029612697918927192, "optim/total_tokens": 2356150272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 4.826247692108154, "created_at": "2025-01-15T01:40:04.980330+00:00"} {"global_step": 8989, "acc_step": 0, "speed/wps": 12955.217137131405, "speed/FLOPS": 203479487472546.4, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1629935503005981, "optim/lr": 0.002961250445304764, "optim/total_tokens": 2356412416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436666, "loss/out": 4.789008140563965, "created_at": "2025-01-15T01:40:10.042836+00:00"} {"global_step": 8990, "acc_step": 0, "speed/wps": 12961.40375735716, "speed/FLOPS": 203576656844498.97, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3091447353363037, "optim/lr": 0.0029612310939492216, "optim/total_tokens": 2356674560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333837, "loss/out": 4.8780670166015625, "created_at": "2025-01-15T01:40:15.102725+00:00"} {"global_step": 8991, "acc_step": 0, "speed/wps": 12965.77484181161, "speed/FLOPS": 203645310732355.56, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9334121942520142, "optim/lr": 0.0029612117378261566, "optim/total_tokens": 2356936704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324713, "loss/out": 4.83047342300415, "created_at": "2025-01-15T01:40:20.165585+00:00"} {"global_step": 8992, "acc_step": 0, "speed/wps": 12957.759676465486, "speed/FLOPS": 203519421546603.3, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2330522537231445, "optim/lr": 0.0029611923769356312, "optim/total_tokens": 2357198848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350820, "loss/out": 4.813716888427734, "created_at": "2025-01-15T01:40:25.229530+00:00"} {"global_step": 8993, "acc_step": 0, "speed/wps": 12963.678900018482, "speed/FLOPS": 203612391086368.66, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2151116132736206, "optim/lr": 0.0029611730112777093, "optim/total_tokens": 2357460992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447032, "loss/out": 4.790452480316162, "created_at": "2025-01-15T01:40:30.296564+00:00"} {"global_step": 8994, "acc_step": 0, "speed/wps": 12960.92589250795, "speed/FLOPS": 203569151320387.0, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5609605312347412, "optim/lr": 0.0029611536408524534, "optim/total_tokens": 2357723136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 4.771385669708252, "created_at": "2025-01-15T01:40:35.370803+00:00"} {"global_step": 8995, "acc_step": 0, "speed/wps": 12961.800793305838, "speed/FLOPS": 203582892839657.5, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0840857028961182, "optim/lr": 0.0029611342656599266, "optim/total_tokens": 2357985280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 4.697973251342773, "created_at": "2025-01-15T01:40:40.434169+00:00"} {"global_step": 8996, "acc_step": 0, "speed/wps": 12963.445392335887, "speed/FLOPS": 203608723527340.3, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9016704559326172, "optim/lr": 0.0029611148857001924, "optim/total_tokens": 2358247424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445976, "loss/out": 4.741213798522949, "created_at": "2025-01-15T01:40:45.493425+00:00"} {"global_step": 8997, "acc_step": 0, "speed/wps": 12962.729890401484, "speed/FLOPS": 203597485586257.8, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1146568059921265, "optim/lr": 0.002961095500973314, "optim/total_tokens": 2358509568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287094, "loss/out": 4.74867582321167, "created_at": "2025-01-15T01:40:50.557240+00:00"} {"global_step": 8998, "acc_step": 0, "speed/wps": 12953.537946595461, "speed/FLOPS": 203453113477729.8, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2988438606262207, "optim/lr": 0.0029610761114793552, "optim/total_tokens": 2358771712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424393, "loss/out": 4.8082380294799805, "created_at": "2025-01-15T01:40:55.625218+00:00"} {"global_step": 8999, "acc_step": 0, "speed/wps": 12959.55958287708, "speed/FLOPS": 203547691550128.0, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.460102081298828, "optim/lr": 0.002961056717218378, "optim/total_tokens": 2359033856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468799, "loss/out": 4.8616042137146, "created_at": "2025-01-15T01:41:00.690756+00:00"} {"global_step": 9000, "acc_step": 0, "speed/wps": 12962.338776429684, "speed/FLOPS": 203591342603883.78, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7212986946105957, "optim/lr": 0.002961037318190447, "optim/total_tokens": 2359296000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466155, "loss/out": 4.790530681610107, "created_at": "2025-01-15T01:41:05.755684+00:00"} {"global_step": 9001, "acc_step": 0, "speed/wps": 12292.97094894643, "speed/FLOPS": 193078001065475.34, "speed/curr_iter_time": 1.2709, "speed/data_load_time": 0.2803, "optim/grad_norm": 2.1364736557006836, "optim/lr": 0.002961017914395625, "optim/total_tokens": 2359558144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419380, "loss/out": 4.849999904632568, "created_at": "2025-01-15T01:41:11.094975+00:00"} {"global_step": 9002, "acc_step": 0, "speed/wps": 12964.027056450424, "speed/FLOPS": 203617859361549.38, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4775722026824951, "optim/lr": 0.002960998505833975, "optim/total_tokens": 2359820288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 4.783624172210693, "created_at": "2025-01-15T01:41:16.156227+00:00"} {"global_step": 9003, "acc_step": 0, "speed/wps": 12963.956585839864, "speed/FLOPS": 203616752523773.75, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.3318347930908203, "optim/lr": 0.0029609790925055604, "optim/total_tokens": 2360082432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462417, "loss/out": 4.787773132324219, "created_at": "2025-01-15T01:41:21.214940+00:00"} {"global_step": 9004, "acc_step": 0, "speed/wps": 12966.302331796287, "speed/FLOPS": 203653595687412.12, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 4.317779541015625, "optim/lr": 0.0029609596744104446, "optim/total_tokens": 2360344576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 4.7291998863220215, "created_at": "2025-01-15T01:41:26.276781+00:00"} {"global_step": 9005, "acc_step": 0, "speed/wps": 12964.472751965375, "speed/FLOPS": 203624859622061.5, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.191415309906006, "optim/lr": 0.002960940251548691, "optim/total_tokens": 2360606720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438769, "loss/out": 4.683332443237305, "created_at": "2025-01-15T01:41:31.335722+00:00"} {"global_step": 9006, "acc_step": 0, "speed/wps": 12966.504327458902, "speed/FLOPS": 203656768306864.62, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7366766929626465, "optim/lr": 0.002960920823920364, "optim/total_tokens": 2360868864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316000, "loss/out": 4.753750324249268, "created_at": "2025-01-15T01:41:36.396984+00:00"} {"global_step": 9007, "acc_step": 0, "speed/wps": 12958.515214014009, "speed/FLOPS": 203531288302020.25, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.41542911529541, "optim/lr": 0.0029609013915255253, "optim/total_tokens": 2361131008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427855, "loss/out": 4.785137176513672, "created_at": "2025-01-15T01:41:41.462410+00:00"} {"global_step": 9008, "acc_step": 0, "speed/wps": 12963.359918724513, "speed/FLOPS": 203607381046820.62, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6494163274765015, "optim/lr": 0.0029608819543642385, "optim/total_tokens": 2361393152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 4.77439022064209, "created_at": "2025-01-15T01:41:46.522466+00:00"} {"global_step": 9009, "acc_step": 0, "speed/wps": 12967.313608458026, "speed/FLOPS": 203669479177024.47, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 7.529836177825928, "optim/lr": 0.0029608625124365674, "optim/total_tokens": 2361655296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 4.781750202178955, "created_at": "2025-01-15T01:41:51.584280+00:00"} {"global_step": 9010, "acc_step": 0, "speed/wps": 12962.024710343054, "speed/FLOPS": 203586409764421.22, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.299580693244934, "optim/lr": 0.0029608430657425766, "optim/total_tokens": 2361917440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467490, "loss/out": 4.846890926361084, "created_at": "2025-01-15T01:41:56.650670+00:00"} {"global_step": 9011, "acc_step": 0, "speed/wps": 12962.373372478918, "speed/FLOPS": 203591885982376.97, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.417315721511841, "optim/lr": 0.002960823614282328, "optim/total_tokens": 2362179584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429739, "loss/out": 4.770901679992676, "created_at": "2025-01-15T01:42:01.711361+00:00"} {"global_step": 9012, "acc_step": 0, "speed/wps": 12962.080967777929, "speed/FLOPS": 203587293364741.75, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.583803653717041, "optim/lr": 0.0029608041580558857, "optim/total_tokens": 2362441728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294127, "loss/out": 4.769001483917236, "created_at": "2025-01-15T01:42:06.776883+00:00"} {"global_step": 9013, "acc_step": 0, "speed/wps": 12964.351265521273, "speed/FLOPS": 203622951510517.0, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.692258596420288, "optim/lr": 0.002960784697063313, "optim/total_tokens": 2362703872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456330, "loss/out": 4.74122428894043, "created_at": "2025-01-15T01:42:11.835476+00:00"} {"global_step": 9014, "acc_step": 0, "speed/wps": 12961.73814930114, "speed/FLOPS": 203581908929481.84, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.818656086921692, "optim/lr": 0.002960765231304673, "optim/total_tokens": 2362966016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 4.678435325622559, "created_at": "2025-01-15T01:42:16.896399+00:00"} {"global_step": 9015, "acc_step": 0, "speed/wps": 12960.873118473382, "speed/FLOPS": 203568322431654.97, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3730077743530273, "optim/lr": 0.00296074576078003, "optim/total_tokens": 2363228160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450462, "loss/out": 4.675837993621826, "created_at": "2025-01-15T01:42:21.961365+00:00"} {"global_step": 9016, "acc_step": 0, "speed/wps": 12962.527909643417, "speed/FLOPS": 203594313200901.72, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.422715187072754, "optim/lr": 0.002960726285489447, "optim/total_tokens": 2363490304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447263, "loss/out": 4.791214942932129, "created_at": "2025-01-15T01:42:27.020748+00:00"} {"global_step": 9017, "acc_step": 0, "speed/wps": 12972.46681414471, "speed/FLOPS": 203750417353580.5, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.151197671890259, "optim/lr": 0.002960706805432988, "optim/total_tokens": 2363752448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430791, "loss/out": 4.801340103149414, "created_at": "2025-01-15T01:42:32.081131+00:00"} {"global_step": 9018, "acc_step": 0, "speed/wps": 12965.06640660516, "speed/FLOPS": 203634183783946.28, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.121205449104309, "optim/lr": 0.0029606873206107164, "optim/total_tokens": 2364014592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 4.722888946533203, "created_at": "2025-01-15T01:42:37.143991+00:00"} {"global_step": 9019, "acc_step": 0, "speed/wps": 12950.757784512944, "speed/FLOPS": 203409447211880.8, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.087523102760315, "optim/lr": 0.0029606678310226953, "optim/total_tokens": 2364276736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437717, "loss/out": 4.791555404663086, "created_at": "2025-01-15T01:42:42.208513+00:00"} {"global_step": 9020, "acc_step": 0, "speed/wps": 12960.135406003797, "speed/FLOPS": 203556735643597.0, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8193509578704834, "optim/lr": 0.0029606483366689884, "optim/total_tokens": 2364538880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329272, "loss/out": 4.66861629486084, "created_at": "2025-01-15T01:42:47.269149+00:00"} {"global_step": 9021, "acc_step": 0, "speed/wps": 12968.02218651215, "speed/FLOPS": 203680608369052.78, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3874611854553223, "optim/lr": 0.00296062883754966, "optim/total_tokens": 2364801024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402750, "loss/out": 4.826984882354736, "created_at": "2025-01-15T01:42:52.334417+00:00"} {"global_step": 9022, "acc_step": 0, "speed/wps": 12970.688732860275, "speed/FLOPS": 203722490143668.75, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.2047945261001587, "optim/lr": 0.0029606093336647732, "optim/total_tokens": 2365063168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448361, "loss/out": 4.634970188140869, "created_at": "2025-01-15T01:42:57.391904+00:00"} {"global_step": 9023, "acc_step": 0, "speed/wps": 12968.38927297639, "speed/FLOPS": 203686373966403.34, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3580029010772705, "optim/lr": 0.0029605898250143915, "optim/total_tokens": 2365325312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475383, "loss/out": 4.799675941467285, "created_at": "2025-01-15T01:43:02.452142+00:00"} {"global_step": 9024, "acc_step": 0, "speed/wps": 12964.198000900209, "speed/FLOPS": 203620544279035.53, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.53794264793396, "optim/lr": 0.002960570311598579, "optim/total_tokens": 2365587456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428932, "loss/out": 4.669272422790527, "created_at": "2025-01-15T01:43:07.511260+00:00"} {"global_step": 9025, "acc_step": 0, "speed/wps": 12968.213801433409, "speed/FLOPS": 203683617944697.62, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4916257858276367, "optim/lr": 0.0029605507934173983, "optim/total_tokens": 2365849600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459765, "loss/out": 4.670139789581299, "created_at": "2025-01-15T01:43:12.571426+00:00"} {"global_step": 9026, "acc_step": 0, "speed/wps": 12971.869908627903, "speed/FLOPS": 203741042132204.94, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1969637870788574, "optim/lr": 0.002960531270470915, "optim/total_tokens": 2366111744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 4.752538204193115, "created_at": "2025-01-15T01:43:17.632600+00:00"} {"global_step": 9027, "acc_step": 0, "speed/wps": 12973.633237613658, "speed/FLOPS": 203768737636994.16, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.4976152181625366, "optim/lr": 0.002960511742759191, "optim/total_tokens": 2366373888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441157, "loss/out": 4.782196521759033, "created_at": "2025-01-15T01:43:22.689259+00:00"} {"global_step": 9028, "acc_step": 0, "speed/wps": 12957.410754278775, "speed/FLOPS": 203513941244191.7, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.168310284614563, "optim/lr": 0.002960492210282291, "optim/total_tokens": 2366636032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324035, "loss/out": 4.673799991607666, "created_at": "2025-01-15T01:43:27.759485+00:00"} {"global_step": 9029, "acc_step": 0, "speed/wps": 12963.904553771326, "speed/FLOPS": 203615935288637.8, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.173421025276184, "optim/lr": 0.0029604726730402785, "optim/total_tokens": 2366898176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442208, "loss/out": 4.759871006011963, "created_at": "2025-01-15T01:43:32.819221+00:00"} {"global_step": 9030, "acc_step": 0, "speed/wps": 12962.686068110837, "speed/FLOPS": 203596797296965.1, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7491950988769531, "optim/lr": 0.002960453131033217, "optim/total_tokens": 2367160320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437717, "loss/out": 4.808455467224121, "created_at": "2025-01-15T01:43:37.887850+00:00"} {"global_step": 9031, "acc_step": 0, "speed/wps": 12964.386414501976, "speed/FLOPS": 203623503573558.72, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0147089958190918, "optim/lr": 0.0029604335842611704, "optim/total_tokens": 2367422464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 4.701729774475098, "created_at": "2025-01-15T01:43:42.948297+00:00"} {"global_step": 9032, "acc_step": 0, "speed/wps": 12960.004734960441, "speed/FLOPS": 203554683275300.25, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4035176038742065, "optim/lr": 0.002960414032724202, "optim/total_tokens": 2367684608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450462, "loss/out": 4.712491989135742, "created_at": "2025-01-15T01:43:48.012564+00:00"} {"global_step": 9033, "acc_step": 0, "speed/wps": 12961.005604626342, "speed/FLOPS": 203570403308742.25, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2238681316375732, "optim/lr": 0.0029603944764223767, "optim/total_tokens": 2367946752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493719, "loss/out": 4.680596828460693, "created_at": "2025-01-15T01:43:53.076432+00:00"} {"global_step": 9034, "acc_step": 0, "speed/wps": 12958.210479551053, "speed/FLOPS": 203526502028530.5, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2350132465362549, "optim/lr": 0.0029603749153557578, "optim/total_tokens": 2368208896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427880, "loss/out": 4.825857639312744, "created_at": "2025-01-15T01:43:58.139096+00:00"} {"global_step": 9035, "acc_step": 0, "speed/wps": 12962.437130551665, "speed/FLOPS": 203592887390522.22, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.121402382850647, "optim/lr": 0.002960355349524409, "optim/total_tokens": 2368471040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297588, "loss/out": 4.739873886108398, "created_at": "2025-01-15T01:44:03.203180+00:00"} {"global_step": 9036, "acc_step": 0, "speed/wps": 12961.033136819697, "speed/FLOPS": 203570835739672.1, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6301485300064087, "optim/lr": 0.002960335778928394, "optim/total_tokens": 2368733184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445933, "loss/out": 4.742157936096191, "created_at": "2025-01-15T01:44:08.263493+00:00"} {"global_step": 9037, "acc_step": 0, "speed/wps": 12962.412316834241, "speed/FLOPS": 203592497656992.47, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1075003147125244, "optim/lr": 0.0029603162035677765, "optim/total_tokens": 2368995328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 4.750276565551758, "created_at": "2025-01-15T01:44:13.323186+00:00"} {"global_step": 9038, "acc_step": 0, "speed/wps": 12956.963082076922, "speed/FLOPS": 203506909937095.5, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2454180717468262, "optim/lr": 0.002960296623442621, "optim/total_tokens": 2369257472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457095, "loss/out": 4.763553619384766, "created_at": "2025-01-15T01:44:18.387509+00:00"} {"global_step": 9039, "acc_step": 0, "speed/wps": 12961.30672480157, "speed/FLOPS": 203575132814876.72, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.572794795036316, "optim/lr": 0.0029602770385529914, "optim/total_tokens": 2369519616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 4.728781700134277, "created_at": "2025-01-15T01:44:23.452269+00:00"} {"global_step": 9040, "acc_step": 0, "speed/wps": 12960.171019064013, "speed/FLOPS": 203557294995645.16, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.386572241783142, "optim/lr": 0.0029602574488989506, "optim/total_tokens": 2369781760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 4.748577117919922, "created_at": "2025-01-15T01:44:28.512840+00:00"} {"global_step": 9041, "acc_step": 0, "speed/wps": 12961.951334305457, "speed/FLOPS": 203585257292921.53, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2270159721374512, "optim/lr": 0.0029602378544805637, "optim/total_tokens": 2370043904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 4.806008338928223, "created_at": "2025-01-15T01:44:33.573276+00:00"} {"global_step": 9042, "acc_step": 0, "speed/wps": 12955.887136686857, "speed/FLOPS": 203490010736240.66, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3376238346099854, "optim/lr": 0.0029602182552978936, "optim/total_tokens": 2370306048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331967, "loss/out": 4.858203887939453, "created_at": "2025-01-15T01:44:38.637567+00:00"} {"global_step": 9043, "acc_step": 0, "speed/wps": 12952.914489931572, "speed/FLOPS": 203443321234104.2, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2521158456802368, "optim/lr": 0.0029601986513510055, "optim/total_tokens": 2370568192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 398779, "loss/out": 4.717955589294434, "created_at": "2025-01-15T01:44:43.701440+00:00"} {"global_step": 9044, "acc_step": 0, "speed/wps": 12960.959612063425, "speed/FLOPS": 203569680932341.25, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0842288732528687, "optim/lr": 0.0029601790426399617, "optim/total_tokens": 2370830336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401696, "loss/out": 4.751466751098633, "created_at": "2025-01-15T01:44:48.761991+00:00"} {"global_step": 9045, "acc_step": 0, "speed/wps": 12960.775711916418, "speed/FLOPS": 203566792527827.22, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3294461965560913, "optim/lr": 0.0029601594291648278, "optim/total_tokens": 2371092480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457665, "loss/out": 4.811460018157959, "created_at": "2025-01-15T01:44:53.822096+00:00"} {"global_step": 9046, "acc_step": 0, "speed/wps": 12960.23968455639, "speed/FLOPS": 203558373481559.22, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1404166221618652, "optim/lr": 0.002960139810925667, "optim/total_tokens": 2371354624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456312, "loss/out": 4.7356672286987305, "created_at": "2025-01-15T01:44:58.882652+00:00"} {"global_step": 9047, "acc_step": 0, "speed/wps": 12966.659472065607, "speed/FLOPS": 203659205066103.94, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.081660032272339, "optim/lr": 0.002960120187922543, "optim/total_tokens": 2371616768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437717, "loss/out": 4.627333641052246, "created_at": "2025-01-15T01:45:03.942767+00:00"} {"global_step": 9048, "acc_step": 0, "speed/wps": 12969.497794737048, "speed/FLOPS": 203703784823923.0, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.670762538909912, "optim/lr": 0.0029601005601555203, "optim/total_tokens": 2371878912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 4.684462547302246, "created_at": "2025-01-15T01:45:09.002107+00:00"} {"global_step": 9049, "acc_step": 0, "speed/wps": 12965.980239813709, "speed/FLOPS": 203648536790224.97, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7564815282821655, "optim/lr": 0.002960080927624663, "optim/total_tokens": 2372141056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450994, "loss/out": 4.639788627624512, "created_at": "2025-01-15T01:45:14.060458+00:00"} {"global_step": 9050, "acc_step": 0, "speed/wps": 12960.146653217218, "speed/FLOPS": 203556912296539.06, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.344962120056152, "optim/lr": 0.002960061290330035, "optim/total_tokens": 2372403200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297061, "loss/out": 4.756952285766602, "created_at": "2025-01-15T01:45:19.125731+00:00"} {"global_step": 9051, "acc_step": 0, "speed/wps": 12968.436140438744, "speed/FLOPS": 203687110084302.25, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5152018070220947, "optim/lr": 0.0029600416482717003, "optim/total_tokens": 2372665344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421416, "loss/out": 4.704710006713867, "created_at": "2025-01-15T01:45:24.188018+00:00"} {"global_step": 9052, "acc_step": 0, "speed/wps": 12962.941615554015, "speed/FLOPS": 203600811020719.16, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3597314357757568, "optim/lr": 0.0029600220014497237, "optim/total_tokens": 2372927488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460578, "loss/out": 4.678305625915527, "created_at": "2025-01-15T01:45:29.250726+00:00"} {"global_step": 9053, "acc_step": 0, "speed/wps": 12962.61016063861, "speed/FLOPS": 203595605065805.94, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.841670274734497, "optim/lr": 0.002960002349864168, "optim/total_tokens": 2373189632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459743, "loss/out": 4.772042751312256, "created_at": "2025-01-15T01:45:34.314911+00:00"} {"global_step": 9054, "acc_step": 0, "speed/wps": 12965.898280512238, "speed/FLOPS": 203647249506771.47, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1592906713485718, "optim/lr": 0.0029599826935150977, "optim/total_tokens": 2373451776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435567, "loss/out": 4.754293441772461, "created_at": "2025-01-15T01:45:39.373023+00:00"} {"global_step": 9055, "acc_step": 0, "speed/wps": 12961.335387500058, "speed/FLOPS": 203575583001943.12, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.049354076385498, "optim/lr": 0.002959963032402578, "optim/total_tokens": 2373713920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 4.729314804077148, "created_at": "2025-01-15T01:45:44.436552+00:00"} {"global_step": 9056, "acc_step": 0, "speed/wps": 12963.220115164911, "speed/FLOPS": 203605185239806.88, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1376389265060425, "optim/lr": 0.002959943366526672, "optim/total_tokens": 2373976064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452325, "loss/out": 4.77364444732666, "created_at": "2025-01-15T01:45:49.495767+00:00"} {"global_step": 9057, "acc_step": 0, "speed/wps": 12962.072362712966, "speed/FLOPS": 203587158210371.94, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1649086475372314, "optim/lr": 0.0029599236958874437, "optim/total_tokens": 2374238208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452847, "loss/out": 4.739175796508789, "created_at": "2025-01-15T01:45:54.556371+00:00"} {"global_step": 9058, "acc_step": 0, "speed/wps": 12970.207594073187, "speed/FLOPS": 203714933197863.53, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0849496126174927, "optim/lr": 0.0029599040204849577, "optim/total_tokens": 2374500352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324322, "loss/out": 4.664894104003906, "created_at": "2025-01-15T01:45:59.614126+00:00"} {"global_step": 9059, "acc_step": 0, "speed/wps": 12959.846565537766, "speed/FLOPS": 203552199007169.56, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5522581338882446, "optim/lr": 0.0029598843403192787, "optim/total_tokens": 2374762496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360941, "loss/out": 4.7239885330200195, "created_at": "2025-01-15T01:46:04.679496+00:00"} {"global_step": 9060, "acc_step": 0, "speed/wps": 12966.97645570068, "speed/FLOPS": 203664183729673.94, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1542574167251587, "optim/lr": 0.00295986465539047, "optim/total_tokens": 2375024640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437201, "loss/out": 4.678422451019287, "created_at": "2025-01-15T01:46:09.738416+00:00"} {"global_step": 9061, "acc_step": 0, "speed/wps": 12962.192440694029, "speed/FLOPS": 203589044200068.94, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3834927082061768, "optim/lr": 0.002959844965698597, "optim/total_tokens": 2375286784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 4.5816850662231445, "created_at": "2025-01-15T01:46:14.800294+00:00"} {"global_step": 9062, "acc_step": 0, "speed/wps": 12953.511271268791, "speed/FLOPS": 203452694505068.72, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.173640489578247, "optim/lr": 0.002959825271243722, "optim/total_tokens": 2375548928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441157, "loss/out": 4.693045616149902, "created_at": "2025-01-15T01:46:19.867770+00:00"} {"global_step": 9063, "acc_step": 0, "speed/wps": 12957.347148236322, "speed/FLOPS": 203512942223891.1, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.4632318019866943, "optim/lr": 0.0029598055720259117, "optim/total_tokens": 2375811072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 4.769888401031494, "created_at": "2025-01-15T01:46:24.935128+00:00"} {"global_step": 9064, "acc_step": 0, "speed/wps": 12966.632819127384, "speed/FLOPS": 203658786445084.2, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7644000053405762, "optim/lr": 0.0029597858680452285, "optim/total_tokens": 2376073216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329565, "loss/out": 4.753477096557617, "created_at": "2025-01-15T01:46:29.996168+00:00"} {"global_step": 9065, "acc_step": 0, "speed/wps": 12959.837758172864, "speed/FLOPS": 203552060675401.2, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3959659337997437, "optim/lr": 0.0029597661593017367, "optim/total_tokens": 2376335360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376530, "loss/out": 4.75604772567749, "created_at": "2025-01-15T01:46:35.059893+00:00"} {"global_step": 9066, "acc_step": 0, "speed/wps": 12959.491055859049, "speed/FLOPS": 203546615239143.84, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.30330228805542, "optim/lr": 0.002959746445795502, "optim/total_tokens": 2376597504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457095, "loss/out": 4.715864181518555, "created_at": "2025-01-15T01:46:40.125880+00:00"} {"global_step": 9067, "acc_step": 0, "speed/wps": 12965.90151528809, "speed/FLOPS": 203647300313371.28, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1239527463912964, "optim/lr": 0.0029597267275265878, "optim/total_tokens": 2376859648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 4.757854461669922, "created_at": "2025-01-15T01:46:45.184497+00:00"} {"global_step": 9068, "acc_step": 0, "speed/wps": 12958.126590186737, "speed/FLOPS": 203525184430787.84, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4458694458007812, "optim/lr": 0.0029597070044950586, "optim/total_tokens": 2377121792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436679, "loss/out": 4.65992546081543, "created_at": "2025-01-15T01:46:50.246157+00:00"} {"global_step": 9069, "acc_step": 0, "speed/wps": 12967.08349192224, "speed/FLOPS": 203665864880616.7, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.222621202468872, "optim/lr": 0.002959687276700978, "optim/total_tokens": 2377383936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 4.6895246505737305, "created_at": "2025-01-15T01:46:55.309780+00:00"} {"global_step": 9070, "acc_step": 0, "speed/wps": 12965.418722542396, "speed/FLOPS": 203639717389873.4, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1895663738250732, "optim/lr": 0.002959667544144412, "optim/total_tokens": 2377646080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 4.566150665283203, "created_at": "2025-01-15T01:47:00.369161+00:00"} {"global_step": 9071, "acc_step": 0, "speed/wps": 12956.241854674214, "speed/FLOPS": 203495582069665.7, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5074331760406494, "optim/lr": 0.0029596478068254235, "optim/total_tokens": 2377908224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 4.6863298416137695, "created_at": "2025-01-15T01:47:05.432162+00:00"} {"global_step": 9072, "acc_step": 0, "speed/wps": 12959.740340922974, "speed/FLOPS": 203550530603626.34, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.428189992904663, "optim/lr": 0.002959628064744078, "optim/total_tokens": 2378170368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429217, "loss/out": 4.685354709625244, "created_at": "2025-01-15T01:47:10.526201+00:00"} {"global_step": 9073, "acc_step": 0, "speed/wps": 12960.020113644207, "speed/FLOPS": 203554924818661.3, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2092928886413574, "optim/lr": 0.0029596083179004385, "optim/total_tokens": 2378432512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 4.813518524169922, "created_at": "2025-01-15T01:47:15.587497+00:00"} {"global_step": 9074, "acc_step": 0, "speed/wps": 12955.552373926539, "speed/FLOPS": 203484752827077.62, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6377623081207275, "optim/lr": 0.0029595885662945707, "optim/total_tokens": 2378694656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438240, "loss/out": 4.803403854370117, "created_at": "2025-01-15T01:47:20.652119+00:00"} {"global_step": 9075, "acc_step": 0, "speed/wps": 12960.64089109116, "speed/FLOPS": 203564674981503.3, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1922972202301025, "optim/lr": 0.0029595688099265387, "optim/total_tokens": 2378956800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 4.74766731262207, "created_at": "2025-01-15T01:47:25.712753+00:00"} {"global_step": 9076, "acc_step": 0, "speed/wps": 12956.882725427447, "speed/FLOPS": 203505647825493.9, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.1264045238494873, "optim/lr": 0.0029595490487964067, "optim/total_tokens": 2379218944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448083, "loss/out": 4.675362586975098, "created_at": "2025-01-15T01:47:30.779450+00:00"} {"global_step": 9077, "acc_step": 0, "speed/wps": 12959.7894849559, "speed/FLOPS": 203551302478194.4, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.676086664199829, "optim/lr": 0.0029595292829042393, "optim/total_tokens": 2379481088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354562, "loss/out": 4.687204360961914, "created_at": "2025-01-15T01:47:35.844034+00:00"} {"global_step": 9078, "acc_step": 0, "speed/wps": 12959.360729551961, "speed/FLOPS": 203544568285403.9, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3398001194000244, "optim/lr": 0.0029595095122501016, "optim/total_tokens": 2379743232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449942, "loss/out": 4.740000247955322, "created_at": "2025-01-15T01:47:40.909418+00:00"} {"global_step": 9079, "acc_step": 0, "speed/wps": 12958.881246962586, "speed/FLOPS": 203537037352459.66, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1609081029891968, "optim/lr": 0.002959489736834057, "optim/total_tokens": 2380005376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 4.656359672546387, "created_at": "2025-01-15T01:47:45.974809+00:00"} {"global_step": 9080, "acc_step": 0, "speed/wps": 12964.410206333218, "speed/FLOPS": 203623877256961.75, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4940778017044067, "optim/lr": 0.0029594699566561707, "optim/total_tokens": 2380267520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432656, "loss/out": 4.629949569702148, "created_at": "2025-01-15T01:47:51.038437+00:00"} {"global_step": 9081, "acc_step": 0, "speed/wps": 12961.620727445881, "speed/FLOPS": 203580064657890.47, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.8025777339935303, "optim/lr": 0.002959450171716507, "optim/total_tokens": 2380529664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461368, "loss/out": 4.741044044494629, "created_at": "2025-01-15T01:47:56.098916+00:00"} {"global_step": 9082, "acc_step": 0, "speed/wps": 12964.426085118288, "speed/FLOPS": 203624126655101.03, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4469534158706665, "optim/lr": 0.002959430382015131, "optim/total_tokens": 2380791808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444878, "loss/out": 4.72387170791626, "created_at": "2025-01-15T01:48:01.162353+00:00"} {"global_step": 9083, "acc_step": 0, "speed/wps": 12958.587637178673, "speed/FLOPS": 203532425807341.16, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0167877674102783, "optim/lr": 0.002959410587552107, "optim/total_tokens": 2381053952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316156, "loss/out": 4.727822780609131, "created_at": "2025-01-15T01:48:06.223568+00:00"} {"global_step": 9084, "acc_step": 0, "speed/wps": 12958.319528280732, "speed/FLOPS": 203528214788678.75, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4305592775344849, "optim/lr": 0.002959390788327499, "optim/total_tokens": 2381316096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287143, "loss/out": 4.688352108001709, "created_at": "2025-01-15T01:48:11.285303+00:00"} {"global_step": 9085, "acc_step": 0, "speed/wps": 12956.290922082026, "speed/FLOPS": 203496352740731.03, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.456169128417969, "optim/lr": 0.0029593709843413717, "optim/total_tokens": 2381578240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 400297, "loss/out": 4.768072128295898, "created_at": "2025-01-15T01:48:16.351514+00:00"} {"global_step": 9086, "acc_step": 0, "speed/wps": 12933.366034178156, "speed/FLOPS": 203136285874103.62, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.1403911113739014, "optim/lr": 0.0029593511755937898, "optim/total_tokens": 2381840384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 4.749284744262695, "created_at": "2025-01-15T01:48:21.427346+00:00"} {"global_step": 9087, "acc_step": 0, "speed/wps": 12952.19566089299, "speed/FLOPS": 203432031036278.66, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4262166023254395, "optim/lr": 0.002959331362084819, "optim/total_tokens": 2382102528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458430, "loss/out": 4.780129909515381, "created_at": "2025-01-15T01:48:26.494637+00:00"} {"global_step": 9088, "acc_step": 0, "speed/wps": 12955.473368106665, "speed/FLOPS": 203483511932111.12, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1141948699951172, "optim/lr": 0.002959311543814523, "optim/total_tokens": 2382364672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 4.697855472564697, "created_at": "2025-01-15T01:48:31.558040+00:00"} {"global_step": 9089, "acc_step": 0, "speed/wps": 12961.300894872915, "speed/FLOPS": 203575041247836.0, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.742766380310059, "optim/lr": 0.0029592917207829665, "optim/total_tokens": 2382626816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307153, "loss/out": 4.7567138671875, "created_at": "2025-01-15T01:48:36.623710+00:00"} {"global_step": 9090, "acc_step": 0, "speed/wps": 12956.135128560729, "speed/FLOPS": 203493905789396.34, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1592764854431152, "optim/lr": 0.0029592718929902138, "optim/total_tokens": 2382888960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450471, "loss/out": 4.83033561706543, "created_at": "2025-01-15T01:48:41.690538+00:00"} {"global_step": 9091, "acc_step": 0, "speed/wps": 12962.619716779065, "speed/FLOPS": 203595755158123.3, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3697893619537354, "optim/lr": 0.0029592520604363303, "optim/total_tokens": 2383151104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462150, "loss/out": 4.799497127532959, "created_at": "2025-01-15T01:48:46.749972+00:00"} {"global_step": 9092, "acc_step": 0, "speed/wps": 12961.560522770465, "speed/FLOPS": 203579119060733.16, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2698802947998047, "optim/lr": 0.0029592322231213803, "optim/total_tokens": 2383413248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464271, "loss/out": 4.680695056915283, "created_at": "2025-01-15T01:48:51.815131+00:00"} {"global_step": 9093, "acc_step": 0, "speed/wps": 12957.321454521236, "speed/FLOPS": 203512538668788.8, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.09321129322052, "optim/lr": 0.0029592123810454284, "optim/total_tokens": 2383675392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467205, "loss/out": 4.670760154724121, "created_at": "2025-01-15T01:48:56.882013+00:00"} {"global_step": 9094, "acc_step": 0, "speed/wps": 12962.8229325637, "speed/FLOPS": 203598946941268.75, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2982757091522217, "optim/lr": 0.0029591925342085405, "optim/total_tokens": 2383937536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452325, "loss/out": 4.65614128112793, "created_at": "2025-01-15T01:49:01.946464+00:00"} {"global_step": 9095, "acc_step": 0, "speed/wps": 12964.727185104406, "speed/FLOPS": 203628855844137.75, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3436577320098877, "optim/lr": 0.00295917268261078, "optim/total_tokens": 2384199680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313596, "loss/out": 4.709784984588623, "created_at": "2025-01-15T01:49:07.005303+00:00"} {"global_step": 9096, "acc_step": 0, "speed/wps": 12957.995802162734, "speed/FLOPS": 203523130225149.3, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1730283498764038, "optim/lr": 0.002959152826252212, "optim/total_tokens": 2384461824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358542, "loss/out": 4.7115302085876465, "created_at": "2025-01-15T01:49:12.071825+00:00"} {"global_step": 9097, "acc_step": 0, "speed/wps": 12956.602396312523, "speed/FLOPS": 203501244871531.44, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1888090372085571, "optim/lr": 0.0029591329651329013, "optim/total_tokens": 2384723968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 402277, "loss/out": 4.7265214920043945, "created_at": "2025-01-15T01:49:17.135942+00:00"} {"global_step": 9098, "acc_step": 0, "speed/wps": 12954.321010687145, "speed/FLOPS": 203465412575332.53, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.121730089187622, "optim/lr": 0.002959113099252913, "optim/total_tokens": 2384986112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467490, "loss/out": 4.542952060699463, "created_at": "2025-01-15T01:49:22.227173+00:00"} {"global_step": 9099, "acc_step": 0, "speed/wps": 12958.452733010367, "speed/FLOPS": 203530306952000.16, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5264594554901123, "optim/lr": 0.0029590932286123117, "optim/total_tokens": 2385248256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 4.874137878417969, "created_at": "2025-01-15T01:49:27.288577+00:00"} {"global_step": 9100, "acc_step": 0, "speed/wps": 12948.852274237915, "speed/FLOPS": 203379518554566.0, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.192220687866211, "optim/lr": 0.002959073353211162, "optim/total_tokens": 2385510400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335181, "loss/out": 4.786985397338867, "created_at": "2025-01-15T01:49:32.354859+00:00"} {"global_step": 9101, "acc_step": 0, "speed/wps": 12957.917771091654, "speed/FLOPS": 203521904639952.06, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.136949062347412, "optim/lr": 0.0029590534730495294, "optim/total_tokens": 2385772544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350002, "loss/out": 4.726114273071289, "created_at": "2025-01-15T01:49:37.415962+00:00"} {"global_step": 9102, "acc_step": 0, "speed/wps": 12964.169144020118, "speed/FLOPS": 203620091042080.2, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.260318398475647, "optim/lr": 0.0029590335881274784, "optim/total_tokens": 2386034688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436916, "loss/out": 4.72572135925293, "created_at": "2025-01-15T01:49:42.481083+00:00"} {"global_step": 9103, "acc_step": 0, "speed/wps": 12956.335524150987, "speed/FLOPS": 203497053277513.75, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1399264335632324, "optim/lr": 0.002959013698445074, "optim/total_tokens": 2386296832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492907, "loss/out": 4.683767318725586, "created_at": "2025-01-15T01:49:47.547709+00:00"} {"global_step": 9104, "acc_step": 0, "speed/wps": 12954.941344829265, "speed/FLOPS": 203475155775462.97, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.1433184146881104, "optim/lr": 0.0029589938040023803, "optim/total_tokens": 2386558976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431333, "loss/out": 4.731677055358887, "created_at": "2025-01-15T01:49:52.611514+00:00"} {"global_step": 9105, "acc_step": 0, "speed/wps": 12958.564787844114, "speed/FLOPS": 203532066927143.62, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.262059211730957, "optim/lr": 0.002958973904799463, "optim/total_tokens": 2386821120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 4.725171089172363, "created_at": "2025-01-15T01:49:57.673864+00:00"} {"global_step": 9106, "acc_step": 0, "speed/wps": 12961.699316434326, "speed/FLOPS": 203581299005946.56, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0529824495315552, "optim/lr": 0.002958954000836387, "optim/total_tokens": 2387083264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453660, "loss/out": 4.7030415534973145, "created_at": "2025-01-15T01:50:02.735062+00:00"} {"global_step": 9107, "acc_step": 0, "speed/wps": 12956.546527564837, "speed/FLOPS": 203500367376077.34, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5884549617767334, "optim/lr": 0.0029589340921132175, "optim/total_tokens": 2387345408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294471, "loss/out": 4.728616237640381, "created_at": "2025-01-15T01:50:07.797517+00:00"} {"global_step": 9108, "acc_step": 0, "speed/wps": 12959.672143752556, "speed/FLOPS": 203549459473352.2, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2959213256835938, "optim/lr": 0.0029589141786300183, "optim/total_tokens": 2387607552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 410502, "loss/out": 4.7455949783325195, "created_at": "2025-01-15T01:50:12.858535+00:00"} {"global_step": 9109, "acc_step": 0, "speed/wps": 12956.685982049805, "speed/FLOPS": 203502557700394.3, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.617095470428467, "optim/lr": 0.002958894260386856, "optim/total_tokens": 2387869696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429217, "loss/out": 4.650116920471191, "created_at": "2025-01-15T01:50:17.922854+00:00"} {"global_step": 9110, "acc_step": 0, "speed/wps": 12955.468211282918, "speed/FLOPS": 203483430937107.94, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.162768006324768, "optim/lr": 0.0029588743373837943, "optim/total_tokens": 2388131840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469590, "loss/out": 4.844945907592773, "created_at": "2025-01-15T01:50:22.986287+00:00"} {"global_step": 9111, "acc_step": 0, "speed/wps": 12955.624243989081, "speed/FLOPS": 203485881645169.66, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1192808151245117, "optim/lr": 0.002958854409620898, "optim/total_tokens": 2388393984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463750, "loss/out": 4.778114318847656, "created_at": "2025-01-15T01:50:28.051784+00:00"} {"global_step": 9112, "acc_step": 0, "speed/wps": 12959.387765493797, "speed/FLOPS": 203544992922021.72, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.381170630455017, "optim/lr": 0.0029588344770982333, "optim/total_tokens": 2388656128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 421768, "loss/out": 4.938032627105713, "created_at": "2025-01-15T01:50:33.118330+00:00"} {"global_step": 9113, "acc_step": 0, "speed/wps": 12953.007951554478, "speed/FLOPS": 203444789177321.66, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1222673654556274, "optim/lr": 0.002958814539815865, "optim/total_tokens": 2388918272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453375, "loss/out": 4.67676305770874, "created_at": "2025-01-15T01:50:38.186237+00:00"} {"global_step": 9114, "acc_step": 0, "speed/wps": 12955.780285224382, "speed/FLOPS": 203488332487193.2, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1708228588104248, "optim/lr": 0.002958794597773857, "optim/total_tokens": 2389180416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290365, "loss/out": 4.723339080810547, "created_at": "2025-01-15T01:50:43.250582+00:00"} {"global_step": 9115, "acc_step": 0, "speed/wps": 12960.04268407095, "speed/FLOPS": 203555279318228.3, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7084349393844604, "optim/lr": 0.002958774650972276, "optim/total_tokens": 2389442560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 390425, "loss/out": 4.7486395835876465, "created_at": "2025-01-15T01:50:48.316974+00:00"} {"global_step": 9116, "acc_step": 0, "speed/wps": 12959.113745407487, "speed/FLOPS": 203540689060023.16, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.8064537048339844, "optim/lr": 0.0029587546994111857, "optim/total_tokens": 2389704704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450990, "loss/out": 4.7881975173950195, "created_at": "2025-01-15T01:50:53.378120+00:00"} {"global_step": 9117, "acc_step": 0, "speed/wps": 12959.841684489884, "speed/FLOPS": 203552122343605.3, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2572101354599, "optim/lr": 0.0029587347430906526, "optim/total_tokens": 2389966848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 4.628570556640625, "created_at": "2025-01-15T01:50:58.444044+00:00"} {"global_step": 9118, "acc_step": 0, "speed/wps": 12966.204382508144, "speed/FLOPS": 203652057259245.4, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7871040105819702, "optim/lr": 0.00295871478201074, "optim/total_tokens": 2390228992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449135, "loss/out": 4.7053093910217285, "created_at": "2025-01-15T01:51:03.507058+00:00"} {"global_step": 9119, "acc_step": 0, "speed/wps": 12958.864403308682, "speed/FLOPS": 203536772799729.47, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1277287006378174, "optim/lr": 0.002958694816171515, "optim/total_tokens": 2390491136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 4.6790452003479, "created_at": "2025-01-15T01:51:08.571620+00:00"} {"global_step": 9120, "acc_step": 0, "speed/wps": 12954.240095385792, "speed/FLOPS": 203464141689335.5, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4091355800628662, "optim/lr": 0.0029586748455730405, "optim/total_tokens": 2390753280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 286904, "loss/out": 4.687497138977051, "created_at": "2025-01-15T01:51:13.640473+00:00"} {"global_step": 9121, "acc_step": 0, "speed/wps": 12953.761990769846, "speed/FLOPS": 203456632399357.47, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2897406816482544, "optim/lr": 0.002958654870215384, "optim/total_tokens": 2391015424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448890, "loss/out": 4.629645824432373, "created_at": "2025-01-15T01:51:18.703770+00:00"} {"global_step": 9122, "acc_step": 0, "speed/wps": 12958.486618672585, "speed/FLOPS": 203530839172889.06, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1043626070022583, "optim/lr": 0.002958634890098609, "optim/total_tokens": 2391277568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469353, "loss/out": 4.702968597412109, "created_at": "2025-01-15T01:51:23.769591+00:00"} {"global_step": 9123, "acc_step": 0, "speed/wps": 12960.637398744362, "speed/FLOPS": 203564620129397.88, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.001, "optim/grad_norm": 1.3358861207962036, "optim/lr": 0.0029586149052227818, "optim/total_tokens": 2391539712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 4.792356967926025, "created_at": "2025-01-15T01:51:28.839763+00:00"} {"global_step": 9124, "acc_step": 0, "speed/wps": 12961.545727742716, "speed/FLOPS": 203578886684491.84, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1701470613479614, "optim/lr": 0.0029585949155879663, "optim/total_tokens": 2391801856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429454, "loss/out": 4.751340389251709, "created_at": "2025-01-15T01:51:33.902530+00:00"} {"global_step": 9125, "acc_step": 0, "speed/wps": 12954.810852436993, "speed/FLOPS": 203473106213126.97, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.253801941871643, "optim/lr": 0.002958574921194229, "optim/total_tokens": 2392064000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 4.6929826736450195, "created_at": "2025-01-15T01:51:38.970113+00:00"} {"global_step": 9126, "acc_step": 0, "speed/wps": 12955.585140050107, "speed/FLOPS": 203485267464073.62, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1726131439208984, "optim/lr": 0.0029585549220416344, "optim/total_tokens": 2392326144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318319, "loss/out": 4.685461521148682, "created_at": "2025-01-15T01:51:44.032598+00:00"} {"global_step": 9127, "acc_step": 0, "speed/wps": 12962.849674910567, "speed/FLOPS": 203599366966574.12, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1862943172454834, "optim/lr": 0.0029585349181302477, "optim/total_tokens": 2392588288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325000, "loss/out": 4.660594463348389, "created_at": "2025-01-15T01:51:49.094902+00:00"} {"global_step": 9128, "acc_step": 0, "speed/wps": 12953.094342584942, "speed/FLOPS": 203446146067168.8, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.298520803451538, "optim/lr": 0.002958514909460135, "optim/total_tokens": 2392850432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435567, "loss/out": 4.696639060974121, "created_at": "2025-01-15T01:51:54.165804+00:00"} {"global_step": 9129, "acc_step": 0, "speed/wps": 12955.530098209283, "speed/FLOPS": 203484402956328.06, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1828244924545288, "optim/lr": 0.002958494896031361, "optim/total_tokens": 2393112576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469875, "loss/out": 4.703957557678223, "created_at": "2025-01-15T01:51:59.231965+00:00"} {"global_step": 9130, "acc_step": 0, "speed/wps": 12956.721115720842, "speed/FLOPS": 203503109522976.66, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2236546277999878, "optim/lr": 0.0029584748778439906, "optim/total_tokens": 2393374720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 4.768965244293213, "created_at": "2025-01-15T01:52:04.294422+00:00"} {"global_step": 9131, "acc_step": 0, "speed/wps": 12958.56096491371, "speed/FLOPS": 203532006882768.38, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1599465608596802, "optim/lr": 0.00295845485489809, "optim/total_tokens": 2393636864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454979, "loss/out": 4.615509033203125, "created_at": "2025-01-15T01:52:09.360758+00:00"} {"global_step": 9132, "acc_step": 0, "speed/wps": 12954.30413119012, "speed/FLOPS": 203465147459636.88, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2263062000274658, "optim/lr": 0.002958434827193724, "optim/total_tokens": 2393899008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328845, "loss/out": 4.709298133850098, "created_at": "2025-01-15T01:52:14.427840+00:00"} {"global_step": 9133, "acc_step": 0, "speed/wps": 12952.622068362296, "speed/FLOPS": 203438728351529.38, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7022509574890137, "optim/lr": 0.002958414794730958, "optim/total_tokens": 2394161152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419618, "loss/out": 4.712222099304199, "created_at": "2025-01-15T01:52:19.495995+00:00"} {"global_step": 9134, "acc_step": 0, "speed/wps": 12952.106080175668, "speed/FLOPS": 203430624048016.3, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.333634614944458, "optim/lr": 0.002958394757509857, "optim/total_tokens": 2394423296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444878, "loss/out": 4.688467979431152, "created_at": "2025-01-15T01:52:24.564117+00:00"} {"global_step": 9135, "acc_step": 0, "speed/wps": 12945.129660772895, "speed/FLOPS": 203321049794690.88, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3139827251434326, "optim/lr": 0.002958374715530487, "optim/total_tokens": 2394685440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 4.6646013259887695, "created_at": "2025-01-15T01:52:29.631323+00:00"} {"global_step": 9136, "acc_step": 0, "speed/wps": 12951.19301925107, "speed/FLOPS": 203416283171518.3, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7664552927017212, "optim/lr": 0.002958354668792913, "optim/total_tokens": 2394947584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443306, "loss/out": 4.698617458343506, "created_at": "2025-01-15T01:52:34.695004+00:00"} {"global_step": 9137, "acc_step": 0, "speed/wps": 12949.81442862888, "speed/FLOPS": 203394630511414.78, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.0405967235565186, "optim/lr": 0.0029583346172972007, "optim/total_tokens": 2395209728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439583, "loss/out": 4.670475006103516, "created_at": "2025-01-15T01:52:39.759595+00:00"} {"global_step": 9138, "acc_step": 0, "speed/wps": 12952.076893283358, "speed/FLOPS": 203430165627765.6, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1495559215545654, "optim/lr": 0.0029583145610434154, "optim/total_tokens": 2395471872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307966, "loss/out": 4.551928520202637, "created_at": "2025-01-15T01:52:44.823121+00:00"} {"global_step": 9139, "acc_step": 0, "speed/wps": 12952.980381916395, "speed/FLOPS": 203444356158270.9, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.401161551475525, "optim/lr": 0.002958294500031622, "optim/total_tokens": 2395734016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368187, "loss/out": 4.699895858764648, "created_at": "2025-01-15T01:52:49.886292+00:00"} {"global_step": 9140, "acc_step": 0, "speed/wps": 12952.720253626601, "speed/FLOPS": 203440270486025.97, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.214468002319336, "optim/lr": 0.002958274434261887, "optim/total_tokens": 2395996160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420425, "loss/out": 4.721835613250732, "created_at": "2025-01-15T01:52:54.950712+00:00"} {"global_step": 9141, "acc_step": 0, "speed/wps": 12953.625160205098, "speed/FLOPS": 203454483287313.22, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2711024284362793, "optim/lr": 0.0029582543637342753, "optim/total_tokens": 2396258304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 4.662780284881592, "created_at": "2025-01-15T01:53:00.018070+00:00"} {"global_step": 9142, "acc_step": 0, "speed/wps": 12959.115926715887, "speed/FLOPS": 203540723320469.0, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1576671600341797, "optim/lr": 0.0029582342884488517, "optim/total_tokens": 2396520448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450462, "loss/out": 4.809330940246582, "created_at": "2025-01-15T01:53:05.083604+00:00"} {"global_step": 9143, "acc_step": 0, "speed/wps": 12959.003606993732, "speed/FLOPS": 203538959184889.1, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9573434591293335, "optim/lr": 0.002958214208405683, "optim/total_tokens": 2396782592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444119, "loss/out": 4.644244194030762, "created_at": "2025-01-15T01:53:10.149511+00:00"} {"global_step": 9144, "acc_step": 0, "speed/wps": 12955.037175437901, "speed/FLOPS": 203476660926856.66, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.4372971057891846, "optim/lr": 0.002958194123604834, "optim/total_tokens": 2397044736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293593, "loss/out": 4.749388694763184, "created_at": "2025-01-15T01:53:15.216804+00:00"} {"global_step": 9145, "acc_step": 0, "speed/wps": 12947.256737014468, "speed/FLOPS": 203354458449973.84, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.263569951057434, "optim/lr": 0.0029581740340463712, "optim/total_tokens": 2397306880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434230, "loss/out": 4.732308387756348, "created_at": "2025-01-15T01:53:20.287715+00:00"} {"global_step": 9146, "acc_step": 0, "speed/wps": 12956.444438746314, "speed/FLOPS": 203498763930897.7, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.178571343421936, "optim/lr": 0.0029581539397303585, "optim/total_tokens": 2397569024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462150, "loss/out": 4.687570095062256, "created_at": "2025-01-15T01:53:25.354251+00:00"} {"global_step": 9147, "acc_step": 0, "speed/wps": 12954.691011344885, "speed/FLOPS": 203471223944096.5, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4317995309829712, "optim/lr": 0.002958133840656862, "optim/total_tokens": 2397831168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 4.640748977661133, "created_at": "2025-01-15T01:53:30.420940+00:00"} {"global_step": 9148, "acc_step": 0, "speed/wps": 12957.37065655985, "speed/FLOPS": 203513311454415.94, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.5546696186065674, "optim/lr": 0.002958113736825948, "optim/total_tokens": 2398093312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 4.734633445739746, "created_at": "2025-01-15T01:53:35.482404+00:00"} {"global_step": 9149, "acc_step": 0, "speed/wps": 12958.51878635252, "speed/FLOPS": 203531344410505.5, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.262704849243164, "optim/lr": 0.0029580936282376812, "optim/total_tokens": 2398355456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422290, "loss/out": 4.759479522705078, "created_at": "2025-01-15T01:53:40.549059+00:00"} {"global_step": 9150, "acc_step": 0, "speed/wps": 12961.838604907, "speed/FLOPS": 203583486722812.34, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.209912061691284, "optim/lr": 0.002958073514892128, "optim/total_tokens": 2398617600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447269, "loss/out": 4.571616172790527, "created_at": "2025-01-15T01:53:45.609813+00:00"} {"global_step": 9151, "acc_step": 0, "speed/wps": 12948.316181938677, "speed/FLOPS": 203371098488338.44, "speed/curr_iter_time": 1.2775, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1806541681289673, "optim/lr": 0.002958053396789353, "optim/total_tokens": 2398879744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 277225, "loss/out": 4.757349967956543, "created_at": "2025-01-15T01:53:50.678642+00:00"} {"global_step": 9152, "acc_step": 0, "speed/wps": 12958.429371097156, "speed/FLOPS": 203529940021050.56, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.190223455429077, "optim/lr": 0.002958033273929423, "optim/total_tokens": 2399141888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453144, "loss/out": 4.690943717956543, "created_at": "2025-01-15T01:53:55.740455+00:00"} {"global_step": 9153, "acc_step": 0, "speed/wps": 12957.552613344289, "speed/FLOPS": 203516169335750.38, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3256466388702393, "optim/lr": 0.0029580131463124034, "optim/total_tokens": 2399404032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474123, "loss/out": 4.722010612487793, "created_at": "2025-01-15T01:54:00.805149+00:00"} {"global_step": 9154, "acc_step": 0, "speed/wps": 12954.102894629823, "speed/FLOPS": 203461986762929.4, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7632486820220947, "optim/lr": 0.0029579930139383595, "optim/total_tokens": 2399666176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432127, "loss/out": 4.628366470336914, "created_at": "2025-01-15T01:54:05.869063+00:00"} {"global_step": 9155, "acc_step": 0, "speed/wps": 12951.80556375257, "speed/FLOPS": 203425904024638.72, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1442101001739502, "optim/lr": 0.002957972876807357, "optim/total_tokens": 2399928320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482522, "loss/out": 4.717929363250732, "created_at": "2025-01-15T01:54:10.938020+00:00"} {"global_step": 9156, "acc_step": 0, "speed/wps": 12951.403671652013, "speed/FLOPS": 203419591756941.84, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2315064668655396, "optim/lr": 0.0029579527349194615, "optim/total_tokens": 2400190464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316820, "loss/out": 4.737963676452637, "created_at": "2025-01-15T01:54:16.028344+00:00"} {"global_step": 9157, "acc_step": 0, "speed/wps": 12956.02069931901, "speed/FLOPS": 203492108520912.88, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.503381252288818, "optim/lr": 0.002957932588274739, "optim/total_tokens": 2400452608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492670, "loss/out": 4.644082546234131, "created_at": "2025-01-15T01:54:21.096399+00:00"} {"global_step": 9158, "acc_step": 0, "speed/wps": 12949.713451535217, "speed/FLOPS": 203393044527401.72, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.421571731567383, "optim/lr": 0.002957912436873255, "optim/total_tokens": 2400714752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457635, "loss/out": 4.712677001953125, "created_at": "2025-01-15T01:54:26.165478+00:00"} {"global_step": 9159, "acc_step": 0, "speed/wps": 12956.03984466693, "speed/FLOPS": 203492409224910.44, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.0302398204803467, "optim/lr": 0.002957892280715076, "optim/total_tokens": 2400976896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457095, "loss/out": 4.7117767333984375, "created_at": "2025-01-15T01:54:31.230297+00:00"} {"global_step": 9160, "acc_step": 0, "speed/wps": 12960.759536515823, "speed/FLOPS": 203566538470933.56, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3082338571548462, "optim/lr": 0.0029578721198002668, "optim/total_tokens": 2401239040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328221, "loss/out": 4.633719444274902, "created_at": "2025-01-15T01:54:36.291034+00:00"} {"global_step": 9161, "acc_step": 0, "speed/wps": 12947.513575141067, "speed/FLOPS": 203358492445684.66, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2281798124313354, "optim/lr": 0.0029578519541288936, "optim/total_tokens": 2401501184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375916, "loss/out": 4.656205177307129, "created_at": "2025-01-15T01:54:41.357445+00:00"} {"global_step": 9162, "acc_step": 0, "speed/wps": 12959.446202941, "speed/FLOPS": 203545910762431.5, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1702710390090942, "optim/lr": 0.0029578317837010222, "optim/total_tokens": 2401763328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 4.686676979064941, "created_at": "2025-01-15T01:54:46.428093+00:00"} {"global_step": 9163, "acc_step": 0, "speed/wps": 12956.410766535852, "speed/FLOPS": 203498235062561.9, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.525063395500183, "optim/lr": 0.0029578116085167178, "optim/total_tokens": 2402025472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472504, "loss/out": 4.688473701477051, "created_at": "2025-01-15T01:54:51.492318+00:00"} {"global_step": 9164, "acc_step": 0, "speed/wps": 12958.534727072083, "speed/FLOPS": 203531594781411.06, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.001, "optim/grad_norm": 1.3766353130340576, "optim/lr": 0.002957791428576047, "optim/total_tokens": 2402287616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483854, "loss/out": 4.714996814727783, "created_at": "2025-01-15T01:54:56.553716+00:00"} {"global_step": 9165, "acc_step": 0, "speed/wps": 12960.95023502536, "speed/FLOPS": 203569533653073.06, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9972215294837952, "optim/lr": 0.002957771243879076, "optim/total_tokens": 2402549760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439586, "loss/out": 4.728083610534668, "created_at": "2025-01-15T01:55:01.615180+00:00"} {"global_step": 9166, "acc_step": 0, "speed/wps": 12961.77071990596, "speed/FLOPS": 203582420495587.25, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8984196186065674, "optim/lr": 0.0029577510544258694, "optim/total_tokens": 2402811904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443016, "loss/out": 4.7977986335754395, "created_at": "2025-01-15T01:55:06.678102+00:00"} {"global_step": 9167, "acc_step": 0, "speed/wps": 12961.752818229423, "speed/FLOPS": 203582139325157.75, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2067971229553223, "optim/lr": 0.0029577308602164944, "optim/total_tokens": 2403074048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 4.742001533508301, "created_at": "2025-01-15T01:55:11.742567+00:00"} {"global_step": 9168, "acc_step": 0, "speed/wps": 12955.750005994374, "speed/FLOPS": 203487856910277.88, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2050846815109253, "optim/lr": 0.0029577106612510154, "optim/total_tokens": 2403336192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316000, "loss/out": 4.74898624420166, "created_at": "2025-01-15T01:55:16.806966+00:00"} {"global_step": 9169, "acc_step": 0, "speed/wps": 12958.38326795684, "speed/FLOPS": 203529215907879.53, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1365516185760498, "optim/lr": 0.0029576904575294998, "optim/total_tokens": 2403598336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 381371, "loss/out": 4.607390403747559, "created_at": "2025-01-15T01:55:21.873325+00:00"} {"global_step": 9170, "acc_step": 0, "speed/wps": 12957.814427063315, "speed/FLOPS": 203520281479977.4, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.105907678604126, "optim/lr": 0.0029576702490520124, "optim/total_tokens": 2403860480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464007, "loss/out": 4.816924095153809, "created_at": "2025-01-15T01:55:26.934754+00:00"} {"global_step": 9171, "acc_step": 0, "speed/wps": 12960.584397898181, "speed/FLOPS": 203563787678277.8, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.27590811252594, "optim/lr": 0.0029576500358186194, "optim/total_tokens": 2404122624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462672, "loss/out": 4.568273067474365, "created_at": "2025-01-15T01:55:31.995656+00:00"} {"global_step": 9172, "acc_step": 0, "speed/wps": 12956.064273793203, "speed/FLOPS": 203492792917907.7, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1048414707183838, "optim/lr": 0.002957629817829388, "optim/total_tokens": 2404384768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460578, "loss/out": 4.717113494873047, "created_at": "2025-01-15T01:55:37.062588+00:00"} {"global_step": 9173, "acc_step": 0, "speed/wps": 12955.437105848025, "speed/FLOPS": 203482942383509.6, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7985172867774963, "optim/lr": 0.002957609595084382, "optim/total_tokens": 2404646912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440108, "loss/out": 4.8167619705200195, "created_at": "2025-01-15T01:55:42.128498+00:00"} {"global_step": 9174, "acc_step": 0, "speed/wps": 12956.189524308073, "speed/FLOPS": 203494760149355.97, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.211748480796814, "optim/lr": 0.002957589367583669, "optim/total_tokens": 2404909056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290365, "loss/out": 4.848226547241211, "created_at": "2025-01-15T01:55:47.196086+00:00"} {"global_step": 9175, "acc_step": 0, "speed/wps": 12955.384481449995, "speed/FLOPS": 203482115845025.06, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.665523648262024, "optim/lr": 0.002957569135327314, "optim/total_tokens": 2405171200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446505, "loss/out": 4.695423126220703, "created_at": "2025-01-15T01:55:52.261588+00:00"} {"global_step": 9176, "acc_step": 0, "speed/wps": 12949.005221506555, "speed/FLOPS": 203381920801591.56, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1408342123031616, "optim/lr": 0.002957548898315384, "optim/total_tokens": 2405433344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 4.636119842529297, "created_at": "2025-01-15T01:55:57.327398+00:00"} {"global_step": 9177, "acc_step": 0, "speed/wps": 12958.42600641588, "speed/FLOPS": 203529887174107.56, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2186775207519531, "optim/lr": 0.0029575286565479445, "optim/total_tokens": 2405695488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467205, "loss/out": 4.750461578369141, "created_at": "2025-01-15T01:56:02.391404+00:00"} {"global_step": 9178, "acc_step": 0, "speed/wps": 12952.955675919951, "speed/FLOPS": 203443968116647.16, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0553961992263794, "optim/lr": 0.002957508410025061, "optim/total_tokens": 2405957632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436096, "loss/out": 4.614821910858154, "created_at": "2025-01-15T01:56:07.457510+00:00"} {"global_step": 9179, "acc_step": 0, "speed/wps": 12955.269066095449, "speed/FLOPS": 203480303088284.6, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.182613492012024, "optim/lr": 0.0029574881587468003, "optim/total_tokens": 2406219776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 4.6366071701049805, "created_at": "2025-01-15T01:56:12.520855+00:00"} {"global_step": 9180, "acc_step": 0, "speed/wps": 12954.90467532856, "speed/FLOPS": 203474579830565.22, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0835769176483154, "optim/lr": 0.0029574679027132285, "optim/total_tokens": 2406481920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447032, "loss/out": 4.64903450012207, "created_at": "2025-01-15T01:56:17.585741+00:00"} {"global_step": 9181, "acc_step": 0, "speed/wps": 12959.713916910538, "speed/FLOPS": 203550115578200.62, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4018025398254395, "optim/lr": 0.002957447641924412, "optim/total_tokens": 2406744064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329951, "loss/out": 4.619655609130859, "created_at": "2025-01-15T01:56:22.676549+00:00"} {"global_step": 9182, "acc_step": 0, "speed/wps": 12959.14390671502, "speed/FLOPS": 203541162784803.94, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1959953308105469, "optim/lr": 0.0029574273763804154, "optim/total_tokens": 2407006208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 4.706879615783691, "created_at": "2025-01-15T01:56:27.737274+00:00"} {"global_step": 9183, "acc_step": 0, "speed/wps": 12954.783935756715, "speed/FLOPS": 203472683449675.84, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3487142324447632, "optim/lr": 0.002957407106081306, "optim/total_tokens": 2407268352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443545, "loss/out": 4.664521217346191, "created_at": "2025-01-15T01:56:32.802434+00:00"} {"global_step": 9184, "acc_step": 0, "speed/wps": 12959.811200101904, "speed/FLOPS": 203551643544402.16, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.2573704719543457, "optim/lr": 0.00295738683102715, "optim/total_tokens": 2407530496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 4.734097480773926, "created_at": "2025-01-15T01:56:37.866648+00:00"} {"global_step": 9185, "acc_step": 0, "speed/wps": 12962.62579861792, "speed/FLOPS": 203595850681759.88, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1836307048797607, "optim/lr": 0.0029573665512180134, "optim/total_tokens": 2407792640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444641, "loss/out": 4.670145034790039, "created_at": "2025-01-15T01:56:42.932265+00:00"} {"global_step": 9186, "acc_step": 0, "speed/wps": 12958.38030896732, "speed/FLOPS": 203529169432883.06, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.305193543434143, "optim/lr": 0.002957346266653962, "optim/total_tokens": 2408054784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310904, "loss/out": 4.689994812011719, "created_at": "2025-01-15T01:56:47.993316+00:00"} {"global_step": 9187, "acc_step": 0, "speed/wps": 12947.265034174603, "speed/FLOPS": 203354588768275.25, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.189408302307129, "optim/lr": 0.002957325977335062, "optim/total_tokens": 2408316928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445691, "loss/out": 4.681837558746338, "created_at": "2025-01-15T01:56:53.063828+00:00"} {"global_step": 9188, "acc_step": 0, "speed/wps": 12954.851537714765, "speed/FLOPS": 203473745231321.8, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1561610698699951, "optim/lr": 0.00295730568326138, "optim/total_tokens": 2408579072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 4.86690616607666, "created_at": "2025-01-15T01:56:58.126351+00:00"} {"global_step": 9189, "acc_step": 0, "speed/wps": 12959.574905609816, "speed/FLOPS": 203547932214701.25, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2133815288543701, "optim/lr": 0.0029572853844329824, "optim/total_tokens": 2408841216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448077, "loss/out": 4.657848358154297, "created_at": "2025-01-15T01:57:03.195045+00:00"} {"global_step": 9190, "acc_step": 0, "speed/wps": 12961.340336519552, "speed/FLOPS": 203575660733094.94, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.0353949069976807, "optim/lr": 0.002957265080849935, "optim/total_tokens": 2409103360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441971, "loss/out": 4.676854610443115, "created_at": "2025-01-15T01:57:08.260538+00:00"} {"global_step": 9191, "acc_step": 0, "speed/wps": 12957.44247283345, "speed/FLOPS": 203514439427678.72, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0861597061157227, "optim/lr": 0.002957244772512304, "optim/total_tokens": 2409365504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 4.610942840576172, "created_at": "2025-01-15T01:57:13.326760+00:00"} {"global_step": 9192, "acc_step": 0, "speed/wps": 12952.358640613926, "speed/FLOPS": 203434590856753.56, "speed/curr_iter_time": 1.278, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2136434316635132, "optim/lr": 0.0029572244594201556, "optim/total_tokens": 2409627648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324181, "loss/out": 4.7465314865112305, "created_at": "2025-01-15T01:57:18.396027+00:00"} {"global_step": 9193, "acc_step": 0, "speed/wps": 12954.33483292459, "speed/FLOPS": 203465629672565.3, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1630256175994873, "optim/lr": 0.0029572041415735564, "optim/total_tokens": 2409889792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401696, "loss/out": 4.633219242095947, "created_at": "2025-01-15T01:57:23.463862+00:00"} {"global_step": 9194, "acc_step": 0, "speed/wps": 12960.54351189551, "speed/FLOPS": 203563145507422.6, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.886674702167511, "optim/lr": 0.0029571838189725724, "optim/total_tokens": 2410151936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 4.71805477142334, "created_at": "2025-01-15T01:57:28.524408+00:00"} {"global_step": 9195, "acc_step": 0, "speed/wps": 12955.403183658424, "speed/FLOPS": 203482409588907.94, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1960350275039673, "optim/lr": 0.00295716349161727, "optim/total_tokens": 2410414080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452847, "loss/out": 4.576724052429199, "created_at": "2025-01-15T01:57:33.588044+00:00"} {"global_step": 9196, "acc_step": 0, "speed/wps": 12958.58503366296, "speed/FLOPS": 203532384915549.28, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2266623973846436, "optim/lr": 0.002957143159507716, "optim/total_tokens": 2410676224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448890, "loss/out": 4.633397102355957, "created_at": "2025-01-15T01:57:38.654239+00:00"} {"global_step": 9197, "acc_step": 0, "speed/wps": 12956.075048782288, "speed/FLOPS": 203492962153917.66, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1371111869812012, "optim/lr": 0.002957122822643976, "optim/total_tokens": 2410938368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461622, "loss/out": 4.7158918380737305, "created_at": "2025-01-15T01:57:43.717130+00:00"} {"global_step": 9198, "acc_step": 0, "speed/wps": 12957.969771816868, "speed/FLOPS": 203522721382798.03, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 5.164292812347412, "optim/lr": 0.002957102481026116, "optim/total_tokens": 2411200512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318931, "loss/out": 4.667640209197998, "created_at": "2025-01-15T01:57:48.783653+00:00"} {"global_step": 9199, "acc_step": 0, "speed/wps": 12952.027547876116, "speed/FLOPS": 203429390590337.16, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0486186742782593, "optim/lr": 0.0029570821346542042, "optim/total_tokens": 2411462656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454182, "loss/out": 4.741687297821045, "created_at": "2025-01-15T01:57:53.848839+00:00"} {"global_step": 9200, "acc_step": 0, "speed/wps": 12944.641365236343, "speed/FLOPS": 203313380442300.94, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.283132791519165, "optim/lr": 0.002957061783528305, "optim/total_tokens": 2411724800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 4.681195259094238, "created_at": "2025-01-15T01:57:58.915116+00:00"} {"global_step": 9201, "acc_step": 0, "speed/wps": 12952.84746371688, "speed/FLOPS": 203442268495298.78, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3118592500686646, "optim/lr": 0.002957041427648486, "optim/total_tokens": 2411986944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418566, "loss/out": 4.833047866821289, "created_at": "2025-01-15T01:58:03.978367+00:00"} {"global_step": 9202, "acc_step": 0, "speed/wps": 12954.355969367145, "speed/FLOPS": 203465961649440.2, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0814927816390991, "optim/lr": 0.0029570210670148136, "optim/total_tokens": 2412249088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 4.602996349334717, "created_at": "2025-01-15T01:58:09.046318+00:00"} {"global_step": 9203, "acc_step": 0, "speed/wps": 12960.487582896989, "speed/FLOPS": 203562267065646.3, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2181403636932373, "optim/lr": 0.002957000701627353, "optim/total_tokens": 2412511232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 4.674266815185547, "created_at": "2025-01-15T01:58:14.106629+00:00"} {"global_step": 9204, "acc_step": 0, "speed/wps": 12955.54400173985, "speed/FLOPS": 203484621330381.1, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.230294108390808, "optim/lr": 0.002956980331486172, "optim/total_tokens": 2412773376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469590, "loss/out": 4.742124080657959, "created_at": "2025-01-15T01:58:19.170490+00:00"} {"global_step": 9205, "acc_step": 0, "speed/wps": 12963.678537007047, "speed/FLOPS": 203612385384775.22, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.382851243019104, "optim/lr": 0.0029569599565913364, "optim/total_tokens": 2413035520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325000, "loss/out": 4.618105888366699, "created_at": "2025-01-15T01:58:24.237920+00:00"} {"global_step": 9206, "acc_step": 0, "speed/wps": 12960.038289801289, "speed/FLOPS": 203555210300184.94, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.19705069065094, "optim/lr": 0.0029569395769429127, "optim/total_tokens": 2413297664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375384, "loss/out": 4.638434410095215, "created_at": "2025-01-15T01:58:29.302211+00:00"} {"global_step": 9207, "acc_step": 0, "speed/wps": 12947.135160897036, "speed/FLOPS": 203352548929986.5, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0751323699951172, "optim/lr": 0.0029569191925409677, "optim/total_tokens": 2413559808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 4.715578079223633, "created_at": "2025-01-15T01:58:34.372168+00:00"} {"global_step": 9208, "acc_step": 0, "speed/wps": 12959.614420288412, "speed/FLOPS": 203548552847027.47, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.043134331703186, "optim/lr": 0.0029568988033855676, "optim/total_tokens": 2413821952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431855, "loss/out": 4.792119979858398, "created_at": "2025-01-15T01:58:39.434985+00:00"} {"global_step": 9209, "acc_step": 0, "speed/wps": 12963.789755586566, "speed/FLOPS": 203614132225396.9, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2122104167938232, "optim/lr": 0.0029568784094767798, "optim/total_tokens": 2414084096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453375, "loss/out": 4.662568092346191, "created_at": "2025-01-15T01:58:44.498003+00:00"} {"global_step": 9210, "acc_step": 0, "speed/wps": 12957.945617066602, "speed/FLOPS": 203522341999253.0, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2742441892623901, "optim/lr": 0.002956858010814669, "optim/total_tokens": 2414346240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448077, "loss/out": 4.538350582122803, "created_at": "2025-01-15T01:58:49.562950+00:00"} {"global_step": 9211, "acc_step": 0, "speed/wps": 12964.666306646363, "speed/FLOPS": 203627899664298.4, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1709505319595337, "optim/lr": 0.0029568376073993035, "optim/total_tokens": 2414608384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298409, "loss/out": 4.837124347686768, "created_at": "2025-01-15T01:58:54.622500+00:00"} {"global_step": 9212, "acc_step": 0, "speed/wps": 12959.043213649556, "speed/FLOPS": 203539581261844.06, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.1007814407348633, "optim/lr": 0.002956817199230749, "optim/total_tokens": 2414870528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433975, "loss/out": 4.605578899383545, "created_at": "2025-01-15T01:58:59.688586+00:00"} {"global_step": 9213, "acc_step": 0, "speed/wps": 12961.134037471335, "speed/FLOPS": 203572420523058.1, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3164595365524292, "optim/lr": 0.0029567967863090724, "optim/total_tokens": 2415132672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475174, "loss/out": 4.710036277770996, "created_at": "2025-01-15T01:59:04.750062+00:00"} {"global_step": 9214, "acc_step": 0, "speed/wps": 12957.978934563043, "speed/FLOPS": 203522865296317.66, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.043350100517273, "optim/lr": 0.00295677636863434, "optim/total_tokens": 2415394816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448361, "loss/out": 4.69318151473999, "created_at": "2025-01-15T01:59:09.822503+00:00"} {"global_step": 9215, "acc_step": 0, "speed/wps": 12958.529563038277, "speed/FLOPS": 203531513673164.0, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.215776801109314, "optim/lr": 0.002956755946206619, "optim/total_tokens": 2415656960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 4.523231506347656, "created_at": "2025-01-15T01:59:14.883988+00:00"} {"global_step": 9216, "acc_step": 0, "speed/wps": 12965.002567724141, "speed/FLOPS": 203633181106596.06, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1359870433807373, "optim/lr": 0.0029567355190259757, "optim/total_tokens": 2415919104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 4.7629780769348145, "created_at": "2025-01-15T01:59:19.944356+00:00"} {"global_step": 9217, "acc_step": 0, "speed/wps": 12958.94834995614, "speed/FLOPS": 203538091297182.56, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.322053074836731, "optim/lr": 0.002956715087092476, "optim/total_tokens": 2416181248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453375, "loss/out": 4.858896732330322, "created_at": "2025-01-15T01:59:25.005100+00:00"} {"global_step": 9218, "acc_step": 0, "speed/wps": 12962.534404487506, "speed/FLOPS": 203594415211353.0, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.004997730255127, "optim/lr": 0.002956694650406188, "optim/total_tokens": 2416443392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441157, "loss/out": 4.800966262817383, "created_at": "2025-01-15T01:59:30.065518+00:00"} {"global_step": 9219, "acc_step": 0, "speed/wps": 12951.115935338017, "speed/FLOPS": 203415072462739.4, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1342331171035767, "optim/lr": 0.0029566742089671776, "optim/total_tokens": 2416705536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357199, "loss/out": 4.588225841522217, "created_at": "2025-01-15T01:59:35.139429+00:00"} {"global_step": 9220, "acc_step": 0, "speed/wps": 12956.91498296038, "speed/FLOPS": 203506154474375.1, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.0764992237091064, "optim/lr": 0.002956653762775511, "optim/total_tokens": 2416967680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446742, "loss/out": 4.663980484008789, "created_at": "2025-01-15T01:59:40.201078+00:00"} {"global_step": 9221, "acc_step": 0, "speed/wps": 12962.668772810346, "speed/FLOPS": 203596525650504.62, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2440344095230103, "optim/lr": 0.0029566333118312554, "optim/total_tokens": 2417229824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490287, "loss/out": 4.763978004455566, "created_at": "2025-01-15T01:59:45.266377+00:00"} {"global_step": 9222, "acc_step": 0, "speed/wps": 12955.110189391542, "speed/FLOPS": 203477807711330.25, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0776599645614624, "optim/lr": 0.0029566128561344784, "optim/total_tokens": 2417491968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433708, "loss/out": 4.680840492248535, "created_at": "2025-01-15T01:59:50.357938+00:00"} {"global_step": 9223, "acc_step": 0, "speed/wps": 12959.548539553074, "speed/FLOPS": 203547518099549.38, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3467937707901, "optim/lr": 0.0029565923956852455, "optim/total_tokens": 2417754112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314942, "loss/out": 4.676547050476074, "created_at": "2025-01-15T01:59:55.418854+00:00"} {"global_step": 9224, "acc_step": 0, "speed/wps": 12953.220792252429, "speed/FLOPS": 203448132132956.75, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.154950737953186, "optim/lr": 0.002956571930483624, "optim/total_tokens": 2418016256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 397433, "loss/out": 4.669111251831055, "created_at": "2025-01-15T02:00:00.485547+00:00"} {"global_step": 9225, "acc_step": 0, "speed/wps": 12958.435911947105, "speed/FLOPS": 203530042754087.06, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.050054907798767, "optim/lr": 0.00295655146052968, "optim/total_tokens": 2418278400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 4.692610740661621, "created_at": "2025-01-15T02:00:05.551316+00:00"} {"global_step": 9226, "acc_step": 0, "speed/wps": 12956.548624512845, "speed/FLOPS": 203500400311527.3, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.187617301940918, "optim/lr": 0.002956530985823481, "optim/total_tokens": 2418540544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 4.646415710449219, "created_at": "2025-01-15T02:00:10.614446+00:00"} {"global_step": 9227, "acc_step": 0, "speed/wps": 12955.288464432295, "speed/FLOPS": 203480607765820.88, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.02306067943573, "optim/lr": 0.002956510506365094, "optim/total_tokens": 2418802688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461913, "loss/out": 4.898184776306152, "created_at": "2025-01-15T02:00:15.683980+00:00"} {"global_step": 9228, "acc_step": 0, "speed/wps": 12957.546755774638, "speed/FLOPS": 203516077334569.84, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0358184576034546, "optim/lr": 0.0029564900221545854, "optim/total_tokens": 2419064832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440921, "loss/out": 4.53312873840332, "created_at": "2025-01-15T02:00:20.745837+00:00"} {"global_step": 9229, "acc_step": 0, "speed/wps": 12965.628165699896, "speed/FLOPS": 203643006982467.0, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1434491872787476, "optim/lr": 0.0029564695331920215, "optim/total_tokens": 2419326976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426544, "loss/out": 4.609709739685059, "created_at": "2025-01-15T02:00:25.809479+00:00"} {"global_step": 9230, "acc_step": 0, "speed/wps": 12957.061094585662, "speed/FLOPS": 203508449358228.6, "speed/curr_iter_time": 1.2771, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0276031494140625, "optim/lr": 0.0029564490394774707, "optim/total_tokens": 2419589120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296535, "loss/out": 4.723992347717285, "created_at": "2025-01-15T02:00:30.876886+00:00"} {"global_step": 9231, "acc_step": 0, "speed/wps": 12958.858520688384, "speed/FLOPS": 203536680405094.12, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2270411252975464, "optim/lr": 0.0029564285410109982, "optim/total_tokens": 2419851264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335181, "loss/out": 4.751681327819824, "created_at": "2025-01-15T02:00:35.945196+00:00"} {"global_step": 9232, "acc_step": 0, "speed/wps": 12958.610732389521, "speed/FLOPS": 203532788549363.7, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1783908605575562, "optim/lr": 0.0029564080377926715, "optim/total_tokens": 2420113408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437195, "loss/out": 4.674074172973633, "created_at": "2025-01-15T02:00:41.006380+00:00"} {"global_step": 9233, "acc_step": 0, "speed/wps": 12957.122562512162, "speed/FLOPS": 203509414796480.06, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2486817836761475, "optim/lr": 0.002956387529822558, "optim/total_tokens": 2420375552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 4.669014930725098, "created_at": "2025-01-15T02:00:46.072177+00:00"} {"global_step": 9234, "acc_step": 0, "speed/wps": 12956.146977124066, "speed/FLOPS": 203494091887365.2, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3043456077575684, "optim/lr": 0.002956367017100724, "optim/total_tokens": 2420637696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430791, "loss/out": 4.577359676361084, "created_at": "2025-01-15T02:00:51.135297+00:00"} {"global_step": 9235, "acc_step": 0, "speed/wps": 12956.115157566437, "speed/FLOPS": 203493592117487.97, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3044081926345825, "optim/lr": 0.002956346499627237, "optim/total_tokens": 2420899840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 4.626800537109375, "created_at": "2025-01-15T02:00:56.200027+00:00"} {"global_step": 9236, "acc_step": 0, "speed/wps": 12953.565566297251, "speed/FLOPS": 203453547283099.88, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2064776420593262, "optim/lr": 0.0029563259774021628, "optim/total_tokens": 2421161984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438773, "loss/out": 4.630025863647461, "created_at": "2025-01-15T02:01:01.262943+00:00"} {"global_step": 9237, "acc_step": 0, "speed/wps": 12946.086443717568, "speed/FLOPS": 203336077385594.53, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2810840606689453, "optim/lr": 0.0029563054504255695, "optim/total_tokens": 2421424128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321917, "loss/out": 4.614069938659668, "created_at": "2025-01-15T02:01:06.331212+00:00"} {"global_step": 9238, "acc_step": 0, "speed/wps": 12959.104576283271, "speed/FLOPS": 203540545046327.66, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1671195030212402, "optim/lr": 0.0029562849186975234, "optim/total_tokens": 2421686272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443307, "loss/out": 4.687018394470215, "created_at": "2025-01-15T02:01:11.399183+00:00"} {"global_step": 9239, "acc_step": 0, "speed/wps": 12962.136724291468, "speed/FLOPS": 203588169097404.38, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2444714307785034, "optim/lr": 0.002956264382218092, "optim/total_tokens": 2421948416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 4.587326526641846, "created_at": "2025-01-15T02:01:16.462191+00:00"} {"global_step": 9240, "acc_step": 0, "speed/wps": 12962.910576904123, "speed/FLOPS": 203600323516071.06, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3446658849716187, "optim/lr": 0.002956243840987342, "optim/total_tokens": 2422210560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 4.780876636505127, "created_at": "2025-01-15T02:01:21.525451+00:00"} {"global_step": 9241, "acc_step": 0, "speed/wps": 12955.313942577135, "speed/FLOPS": 203481007935096.22, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1619845628738403, "optim/lr": 0.002956223295005341, "optim/total_tokens": 2422472704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 4.739953994750977, "created_at": "2025-01-15T02:01:26.590458+00:00"} {"global_step": 9242, "acc_step": 0, "speed/wps": 12952.161469596895, "speed/FLOPS": 203431494014989.72, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0799306631088257, "optim/lr": 0.0029562027442721545, "optim/total_tokens": 2422734848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435866, "loss/out": 4.618529319763184, "created_at": "2025-01-15T02:01:31.655434+00:00"} {"global_step": 9243, "acc_step": 0, "speed/wps": 12959.328120938506, "speed/FLOPS": 203544056122320.66, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.640246629714966, "optim/lr": 0.0029561821887878516, "optim/total_tokens": 2422996992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436916, "loss/out": 4.702866554260254, "created_at": "2025-01-15T02:01:36.717439+00:00"} {"global_step": 9244, "acc_step": 0, "speed/wps": 12961.194720423091, "speed/FLOPS": 203573373632202.53, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8754198551177979, "optim/lr": 0.002956161628552498, "optim/total_tokens": 2423259136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325525, "loss/out": 4.693212032318115, "created_at": "2025-01-15T02:01:41.783073+00:00"} {"global_step": 9245, "acc_step": 0, "speed/wps": 12953.471941636526, "speed/FLOPS": 203452076779149.4, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0498511791229248, "optim/lr": 0.002956141063566161, "optim/total_tokens": 2423521280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363627, "loss/out": 4.672610282897949, "created_at": "2025-01-15T02:01:46.848682+00:00"} {"global_step": 9246, "acc_step": 0, "speed/wps": 12955.238736782216, "speed/FLOPS": 203479826724743.4, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.001, "optim/grad_norm": 1.1565346717834473, "optim/lr": 0.002956120493828908, "optim/total_tokens": 2423783424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428932, "loss/out": 4.609350204467773, "created_at": "2025-01-15T02:01:51.910883+00:00"} {"global_step": 9247, "acc_step": 0, "speed/wps": 12959.858265746727, "speed/FLOPS": 203552382775029.06, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.47416353225708, "optim/lr": 0.0029560999193408055, "optim/total_tokens": 2424045568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 4.805411338806152, "created_at": "2025-01-15T02:01:56.972129+00:00"} {"global_step": 9248, "acc_step": 0, "speed/wps": 12959.139644315677, "speed/FLOPS": 203541095837965.12, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5235297679901123, "optim/lr": 0.0029560793401019214, "optim/total_tokens": 2424307712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441687, "loss/out": 4.654559135437012, "created_at": "2025-01-15T02:02:02.034386+00:00"} {"global_step": 9249, "acc_step": 0, "speed/wps": 12949.997012138163, "speed/FLOPS": 203397498236324.1, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1569795608520508, "optim/lr": 0.0029560587561123226, "optim/total_tokens": 2424569856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467205, "loss/out": 4.637297630310059, "created_at": "2025-01-15T02:02:07.102210+00:00"} {"global_step": 9250, "acc_step": 0, "speed/wps": 12957.962171996865, "speed/FLOPS": 203522602017181.66, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3311666250228882, "optim/lr": 0.002956038167372076, "optim/total_tokens": 2424832000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434294, "loss/out": 4.612695217132568, "created_at": "2025-01-15T02:02:12.171425+00:00"} {"global_step": 9251, "acc_step": 0, "speed/wps": 12953.529374330028, "speed/FLOPS": 203452978838521.66, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2200379371643066, "optim/lr": 0.002956017573881249, "optim/total_tokens": 2425094144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 378395, "loss/out": 4.5782365798950195, "created_at": "2025-01-15T02:02:17.239599+00:00"} {"global_step": 9252, "acc_step": 0, "speed/wps": 12959.976149752929, "speed/FLOPS": 203554234305336.06, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1031514406204224, "optim/lr": 0.0029559969756399085, "optim/total_tokens": 2425356288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467490, "loss/out": 4.812941074371338, "created_at": "2025-01-15T02:02:22.303907+00:00"} {"global_step": 9253, "acc_step": 0, "speed/wps": 12957.499759408522, "speed/FLOPS": 203515339192060.22, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6685657501220703, "optim/lr": 0.0029559763726481223, "optim/total_tokens": 2425618432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448083, "loss/out": 4.594275951385498, "created_at": "2025-01-15T02:02:27.369742+00:00"} {"global_step": 9254, "acc_step": 0, "speed/wps": 12950.269114351084, "speed/FLOPS": 203401771975485.66, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9853162169456482, "optim/lr": 0.002955955764905957, "optim/total_tokens": 2425880576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 4.521655559539795, "created_at": "2025-01-15T02:02:32.434299+00:00"} {"global_step": 9255, "acc_step": 0, "speed/wps": 12955.068314728316, "speed/FLOPS": 203477150012202.47, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2734506130218506, "optim/lr": 0.0029559351524134805, "optim/total_tokens": 2426142720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440105, "loss/out": 4.797086715698242, "created_at": "2025-01-15T02:02:37.499808+00:00"} {"global_step": 9256, "acc_step": 0, "speed/wps": 12954.040682712777, "speed/FLOPS": 203461009639284.06, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1329468488693237, "optim/lr": 0.002955914535170759, "optim/total_tokens": 2426404864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309845, "loss/out": 4.608114242553711, "created_at": "2025-01-15T02:02:42.564375+00:00"} {"global_step": 9257, "acc_step": 0, "speed/wps": 12958.72464180827, "speed/FLOPS": 203534577653313.1, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3730018138885498, "optim/lr": 0.002955893913177861, "optim/total_tokens": 2426667008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 285549, "loss/out": 4.580724239349365, "created_at": "2025-01-15T02:02:47.626616+00:00"} {"global_step": 9258, "acc_step": 0, "speed/wps": 12956.281702085213, "speed/FLOPS": 203496207928011.5, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3390029668807983, "optim/lr": 0.0029558732864348523, "optim/total_tokens": 2426929152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441157, "loss/out": 4.647616863250732, "created_at": "2025-01-15T02:02:52.694022+00:00"} {"global_step": 9259, "acc_step": 0, "speed/wps": 12952.901889083783, "speed/FLOPS": 203443123320474.28, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.092185139656067, "optim/lr": 0.002955852654941802, "optim/total_tokens": 2427191296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 4.712640762329102, "created_at": "2025-01-15T02:02:57.784755+00:00"} {"global_step": 9260, "acc_step": 0, "speed/wps": 12956.548340625646, "speed/FLOPS": 203500395852688.75, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.5908708572387695, "optim/lr": 0.0029558320186987763, "optim/total_tokens": 2427453440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461889, "loss/out": 4.607147216796875, "created_at": "2025-01-15T02:03:02.846612+00:00"} {"global_step": 9261, "acc_step": 0, "speed/wps": 12961.001878132787, "speed/FLOPS": 203570344779040.25, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7809469103813171, "optim/lr": 0.0029558113777058428, "optim/total_tokens": 2427715584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322156, "loss/out": 4.695966720581055, "created_at": "2025-01-15T02:03:07.910569+00:00"} {"global_step": 9262, "acc_step": 0, "speed/wps": 12954.337835377926, "speed/FLOPS": 203465676830220.78, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.196563720703125, "optim/lr": 0.0029557907319630685, "optim/total_tokens": 2427977728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356674, "loss/out": 4.578622817993164, "created_at": "2025-01-15T02:03:13.010913+00:00"} {"global_step": 9263, "acc_step": 0, "speed/wps": 12951.291225757215, "speed/FLOPS": 203417825639647.2, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1041072607040405, "optim/lr": 0.0029557700814705212, "optim/total_tokens": 2428239872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447554, "loss/out": 4.694067001342773, "created_at": "2025-01-15T02:03:18.086478+00:00"} {"global_step": 9264, "acc_step": 0, "speed/wps": 12953.881657288912, "speed/FLOPS": 203458511926478.44, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.061599612236023, "optim/lr": 0.002955749426228268, "optim/total_tokens": 2428502016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430791, "loss/out": 4.763266563415527, "created_at": "2025-01-15T02:03:23.150407+00:00"} {"global_step": 9265, "acc_step": 0, "speed/wps": 12957.066381510616, "speed/FLOPS": 203508532396649.97, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.061659574508667, "optim/lr": 0.0029557287662363763, "optim/total_tokens": 2428764160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306907, "loss/out": 4.701072692871094, "created_at": "2025-01-15T02:03:28.217442+00:00"} {"global_step": 9266, "acc_step": 0, "speed/wps": 12957.055962741604, "speed/FLOPS": 203508368755565.75, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0612596273422241, "optim/lr": 0.002955708101494914, "optim/total_tokens": 2429026304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434816, "loss/out": 4.769042015075684, "created_at": "2025-01-15T02:03:33.279951+00:00"} {"global_step": 9267, "acc_step": 0, "speed/wps": 12952.389287830007, "speed/FLOPS": 203435072213395.78, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.713048219680786, "optim/lr": 0.0029556874320039476, "optim/total_tokens": 2429288448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 4.625565052032471, "created_at": "2025-01-15T02:03:38.352502+00:00"} {"global_step": 9268, "acc_step": 0, "speed/wps": 12961.82190159245, "speed/FLOPS": 203583224374304.56, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.3244956731796265, "optim/lr": 0.002955666757763546, "optim/total_tokens": 2429550592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425464, "loss/out": 4.732004642486572, "created_at": "2025-01-15T02:03:43.412644+00:00"} {"global_step": 9269, "acc_step": 0, "speed/wps": 12947.135289532682, "speed/FLOPS": 203352550950386.06, "speed/curr_iter_time": 1.2771, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2575210332870483, "optim/lr": 0.002955646078773775, "optim/total_tokens": 2429812736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470403, "loss/out": 4.699948310852051, "created_at": "2025-01-15T02:03:48.478054+00:00"} {"global_step": 9270, "acc_step": 0, "speed/wps": 12962.984637212057, "speed/FLOPS": 203601486734976.7, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1669703722000122, "optim/lr": 0.002955625395034703, "optim/total_tokens": 2430074880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448083, "loss/out": 4.663871765136719, "created_at": "2025-01-15T02:03:53.542442+00:00"} {"global_step": 9271, "acc_step": 0, "speed/wps": 12953.5590494391, "speed/FLOPS": 203453444926887.5, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.061498761177063, "optim/lr": 0.002955604706546397, "optim/total_tokens": 2430337024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449696, "loss/out": 4.76737117767334, "created_at": "2025-01-15T02:03:58.607505+00:00"} {"global_step": 9272, "acc_step": 0, "speed/wps": 12956.617120416431, "speed/FLOPS": 203501476133816.38, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.535731077194214, "optim/lr": 0.002955584013308925, "optim/total_tokens": 2430599168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292773, "loss/out": 4.607763290405273, "created_at": "2025-01-15T02:04:03.674266+00:00"} {"global_step": 9273, "acc_step": 0, "speed/wps": 12957.48194587781, "speed/FLOPS": 203515059406082.03, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0635008811950684, "optim/lr": 0.002955563315322354, "optim/total_tokens": 2430861312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 4.660277366638184, "created_at": "2025-01-15T02:04:08.741713+00:00"} {"global_step": 9274, "acc_step": 0, "speed/wps": 12953.70452262977, "speed/FLOPS": 203455729783248.1, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5120830535888672, "optim/lr": 0.002955542612586752, "optim/total_tokens": 2431123456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 4.653947830200195, "created_at": "2025-01-15T02:04:13.835518+00:00"} {"global_step": 9275, "acc_step": 0, "speed/wps": 12958.540905304626, "speed/FLOPS": 203531691819042.94, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2400990724563599, "optim/lr": 0.0029555219051021872, "optim/total_tokens": 2431385600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443543, "loss/out": 4.789602756500244, "created_at": "2025-01-15T02:04:18.898227+00:00"} {"global_step": 9276, "acc_step": 0, "speed/wps": 12946.566232829786, "speed/FLOPS": 203343613132901.22, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6143072843551636, "optim/lr": 0.0029555011928687255, "optim/total_tokens": 2431647744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345322, "loss/out": 4.649250030517578, "created_at": "2025-01-15T02:04:23.964332+00:00"} {"global_step": 9277, "acc_step": 0, "speed/wps": 12958.510281469684, "speed/FLOPS": 203531210829633.6, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2192999124526978, "optim/lr": 0.0029554804758864356, "optim/total_tokens": 2431909888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444067, "loss/out": 4.568865776062012, "created_at": "2025-01-15T02:04:29.031748+00:00"} {"global_step": 9278, "acc_step": 0, "speed/wps": 12954.42513607331, "speed/FLOPS": 203467048007605.53, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.1572966575622559, "optim/lr": 0.0029554597541553845, "optim/total_tokens": 2432172032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461077, "loss/out": 4.668685436248779, "created_at": "2025-01-15T02:04:34.098359+00:00"} {"global_step": 9279, "acc_step": 0, "speed/wps": 12956.025932915849, "speed/FLOPS": 203492190721742.97, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1891672611236572, "optim/lr": 0.0029554390276756402, "optim/total_tokens": 2432434176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 4.760481834411621, "created_at": "2025-01-15T02:04:39.167140+00:00"} {"global_step": 9280, "acc_step": 0, "speed/wps": 12957.19953423367, "speed/FLOPS": 203510623743136.56, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5626693964004517, "optim/lr": 0.0029554182964472704, "optim/total_tokens": 2432696320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440919, "loss/out": 4.726771831512451, "created_at": "2025-01-15T02:04:44.233886+00:00"} {"global_step": 9281, "acc_step": 0, "speed/wps": 12964.270141274026, "speed/FLOPS": 203621677342737.6, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9200838208198547, "optim/lr": 0.0029553975604703422, "optim/total_tokens": 2432958464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 283915, "loss/out": 4.558049201965332, "created_at": "2025-01-15T02:04:49.292605+00:00"} {"global_step": 9282, "acc_step": 0, "speed/wps": 12953.25729246665, "speed/FLOPS": 203448705418977.62, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7785364389419556, "optim/lr": 0.0029553768197449243, "optim/total_tokens": 2433220608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 379212, "loss/out": 4.718975067138672, "created_at": "2025-01-15T02:04:54.356760+00:00"} {"global_step": 9283, "acc_step": 0, "speed/wps": 12951.29355937168, "speed/FLOPS": 203417862292269.06, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5881731510162354, "optim/lr": 0.0029553560742710833, "optim/total_tokens": 2433482752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 398201, "loss/out": 4.643887042999268, "created_at": "2025-01-15T02:04:59.420642+00:00"} {"global_step": 9284, "acc_step": 0, "speed/wps": 12956.131552765419, "speed/FLOPS": 203493849626617.34, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.261838436126709, "optim/lr": 0.0029553353240488875, "optim/total_tokens": 2433744896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438531, "loss/out": 4.631397247314453, "created_at": "2025-01-15T02:05:04.486392+00:00"} {"global_step": 9285, "acc_step": 0, "speed/wps": 12957.548989047005, "speed/FLOPS": 203516112411181.22, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.051582932472229, "optim/lr": 0.0029553145690784043, "optim/total_tokens": 2434007040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460578, "loss/out": 4.638322353363037, "created_at": "2025-01-15T02:05:09.551915+00:00"} {"global_step": 9286, "acc_step": 0, "speed/wps": 12952.838382137781, "speed/FLOPS": 203442125856619.62, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0469248294830322, "optim/lr": 0.002955293809359701, "optim/total_tokens": 2434269184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461622, "loss/out": 4.6458306312561035, "created_at": "2025-01-15T02:05:14.616519+00:00"} {"global_step": 9287, "acc_step": 0, "speed/wps": 12958.134622100133, "speed/FLOPS": 203525310583024.75, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.001, "optim/grad_norm": 1.0712171792984009, "optim/lr": 0.0029552730448928464, "optim/total_tokens": 2434531328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453375, "loss/out": 4.62435245513916, "created_at": "2025-01-15T02:05:19.680486+00:00"} {"global_step": 9288, "acc_step": 0, "speed/wps": 12955.15786248797, "speed/FLOPS": 203478556482819.34, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.944988489151001, "optim/lr": 0.002955252275677908, "optim/total_tokens": 2434793472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291904, "loss/out": 4.675812721252441, "created_at": "2025-01-15T02:05:24.751711+00:00"} {"global_step": 9289, "acc_step": 0, "speed/wps": 12958.382361169586, "speed/FLOPS": 203529201665539.66, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3645405769348145, "optim/lr": 0.0029552315017149526, "optim/total_tokens": 2435055616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444878, "loss/out": 4.656710624694824, "created_at": "2025-01-15T02:05:29.817798+00:00"} {"global_step": 9290, "acc_step": 0, "speed/wps": 12957.217962538, "speed/FLOPS": 203510913184979.2, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.400156855583191, "optim/lr": 0.0029552107230040488, "optim/total_tokens": 2435317760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492386, "loss/out": 4.686859130859375, "created_at": "2025-01-15T02:05:34.882169+00:00"} {"global_step": 9291, "acc_step": 0, "speed/wps": 12951.2122957359, "speed/FLOPS": 203416585935201.22, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0873386859893799, "optim/lr": 0.002955189939545264, "optim/total_tokens": 2435579904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416992, "loss/out": 4.718421936035156, "created_at": "2025-01-15T02:05:39.954130+00:00"} {"global_step": 9292, "acc_step": 0, "speed/wps": 12956.444715472255, "speed/FLOPS": 203498768277258.84, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4563981294631958, "optim/lr": 0.0029551691513386665, "optim/total_tokens": 2435842048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452847, "loss/out": 4.7410478591918945, "created_at": "2025-01-15T02:05:45.016019+00:00"} {"global_step": 9293, "acc_step": 0, "speed/wps": 12967.343755375896, "speed/FLOPS": 203669952675795.84, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4791803359985352, "optim/lr": 0.002955148358384324, "optim/total_tokens": 2436104192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437717, "loss/out": 4.677400588989258, "created_at": "2025-01-15T02:05:50.078343+00:00"} {"global_step": 9294, "acc_step": 0, "speed/wps": 12960.037160801472, "speed/FLOPS": 203555192567691.4, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5540368556976318, "optim/lr": 0.0029551275606823036, "optim/total_tokens": 2436366336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351757, "loss/out": 4.531778335571289, "created_at": "2025-01-15T02:05:55.138854+00:00"} {"global_step": 9295, "acc_step": 0, "speed/wps": 12957.767345243576, "speed/FLOPS": 203519541995301.22, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3825136423110962, "optim/lr": 0.002955106758232675, "optim/total_tokens": 2436628480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 408639, "loss/out": 4.631943225860596, "created_at": "2025-01-15T02:06:00.200474+00:00"} {"global_step": 9296, "acc_step": 0, "speed/wps": 12956.901045472483, "speed/FLOPS": 203505935566975.56, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.2497448921203613, "optim/lr": 0.0029550859510355035, "optim/total_tokens": 2436890624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 406484, "loss/out": 4.739969253540039, "created_at": "2025-01-15T02:06:05.262633+00:00"} {"global_step": 9297, "acc_step": 0, "speed/wps": 12958.24684801354, "speed/FLOPS": 203527073245209.34, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1719276905059814, "optim/lr": 0.002955065139090859, "optim/total_tokens": 2437152768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426252, "loss/out": 4.572861671447754, "created_at": "2025-01-15T02:06:10.324010+00:00"} {"global_step": 9298, "acc_step": 0, "speed/wps": 12953.722201778393, "speed/FLOPS": 203456007458570.9, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8794574737548828, "optim/lr": 0.0029550443223988086, "optim/total_tokens": 2437414912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 4.7163848876953125, "created_at": "2025-01-15T02:06:15.390361+00:00"} {"global_step": 9299, "acc_step": 0, "speed/wps": 12955.18480457316, "speed/FLOPS": 203478979645289.6, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.0494887828826904, "optim/lr": 0.0029550235009594204, "optim/total_tokens": 2437677056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429217, "loss/out": 4.717444896697998, "created_at": "2025-01-15T02:06:20.457070+00:00"} {"global_step": 9300, "acc_step": 0, "speed/wps": 12957.90521300175, "speed/FLOPS": 203521707397893.5, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3829716444015503, "optim/lr": 0.002955002674772762, "optim/total_tokens": 2437939200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 4.563266754150391, "created_at": "2025-01-15T02:06:25.518833+00:00"} {"global_step": 9301, "acc_step": 0, "speed/wps": 12954.114155222374, "speed/FLOPS": 203462163626009.1, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.025452136993408, "optim/lr": 0.0029549818438389017, "optim/total_tokens": 2438201344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443306, "loss/out": 4.644496917724609, "created_at": "2025-01-15T02:06:30.581668+00:00"} {"global_step": 9302, "acc_step": 0, "speed/wps": 12960.682379264772, "speed/FLOPS": 203565326610280.62, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2807546854019165, "optim/lr": 0.0029549610081579074, "optim/total_tokens": 2438463488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288392, "loss/out": 4.5058441162109375, "created_at": "2025-01-15T02:06:35.647295+00:00"} {"global_step": 9303, "acc_step": 0, "speed/wps": 12961.33417471631, "speed/FLOPS": 203575563953507.84, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1314325332641602, "optim/lr": 0.0029549401677298475, "optim/total_tokens": 2438725632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437201, "loss/out": 4.7921905517578125, "created_at": "2025-01-15T02:06:40.716884+00:00"} {"global_step": 9304, "acc_step": 0, "speed/wps": 12956.379658960286, "speed/FLOPS": 203497746475341.38, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2382270097732544, "optim/lr": 0.0029549193225547887, "optim/total_tokens": 2438987776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 4.681025505065918, "created_at": "2025-01-15T02:06:45.783225+00:00"} {"global_step": 9305, "acc_step": 0, "speed/wps": 12952.59918522504, "speed/FLOPS": 203438368940414.06, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2195408344268799, "optim/lr": 0.0029548984726328006, "optim/total_tokens": 2439249920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465633, "loss/out": 4.664450645446777, "created_at": "2025-01-15T02:06:50.852535+00:00"} {"global_step": 9306, "acc_step": 0, "speed/wps": 12958.05191009495, "speed/FLOPS": 203524011477325.12, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1308099031448364, "optim/lr": 0.0029548776179639507, "optim/total_tokens": 2439512064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452847, "loss/out": 4.5074992179870605, "created_at": "2025-01-15T02:06:55.913744+00:00"} {"global_step": 9307, "acc_step": 0, "speed/wps": 12960.340262544001, "speed/FLOPS": 203559953197063.28, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0957484245300293, "optim/lr": 0.002954856758548306, "optim/total_tokens": 2439774208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307679, "loss/out": 4.582150459289551, "created_at": "2025-01-15T02:07:00.982442+00:00"} {"global_step": 9308, "acc_step": 0, "speed/wps": 12953.552973720685, "speed/FLOPS": 203453349499380.88, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.4203402996063232, "optim/lr": 0.0029548358943859354, "optim/total_tokens": 2440036352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456054, "loss/out": 4.646182537078857, "created_at": "2025-01-15T02:07:06.051436+00:00"} {"global_step": 9309, "acc_step": 0, "speed/wps": 12960.623400730701, "speed/FLOPS": 203564400271358.03, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3016053438186646, "optim/lr": 0.002954815025476908, "optim/total_tokens": 2440298496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 4.763706684112549, "created_at": "2025-01-15T02:07:11.111533+00:00"} {"global_step": 9310, "acc_step": 0, "speed/wps": 12958.345447877198, "speed/FLOPS": 203528621891558.53, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7141588926315308, "optim/lr": 0.00295479415182129, "optim/total_tokens": 2440560640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 4.522402286529541, "created_at": "2025-01-15T02:07:16.172494+00:00"} {"global_step": 9311, "acc_step": 0, "speed/wps": 12959.118182008167, "speed/FLOPS": 203540758742933.38, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9436807632446289, "optim/lr": 0.0029547732734191505, "optim/total_tokens": 2440822784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491573, "loss/out": 4.681389808654785, "created_at": "2025-01-15T02:07:21.237488+00:00"} {"global_step": 9312, "acc_step": 0, "speed/wps": 12964.045989019472, "speed/FLOPS": 203618156723562.34, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0722452402114868, "optim/lr": 0.0029547523902705574, "optim/total_tokens": 2441084928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439054, "loss/out": 4.609226703643799, "created_at": "2025-01-15T02:07:26.301237+00:00"} {"global_step": 9313, "acc_step": 0, "speed/wps": 12961.957394028752, "speed/FLOPS": 203585352469202.8, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0676169395446777, "optim/lr": 0.002954731502375579, "optim/total_tokens": 2441347072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 4.612393379211426, "created_at": "2025-01-15T02:07:31.360848+00:00"} {"global_step": 9314, "acc_step": 0, "speed/wps": 12958.92463067007, "speed/FLOPS": 203537718753201.12, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9981066584587097, "optim/lr": 0.0029547106097342836, "optim/total_tokens": 2441609216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456567, "loss/out": 4.619725227355957, "created_at": "2025-01-15T02:07:36.422008+00:00"} {"global_step": 9315, "acc_step": 0, "speed/wps": 12961.886399162922, "speed/FLOPS": 203584237397277.78, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4762569665908813, "optim/lr": 0.0029546897123467392, "optim/total_tokens": 2441871360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329565, "loss/out": 4.604732990264893, "created_at": "2025-01-15T02:07:41.485836+00:00"} {"global_step": 9316, "acc_step": 0, "speed/wps": 12956.656206728867, "speed/FLOPS": 203502090038063.4, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.027236938476562, "optim/lr": 0.0029546688102130136, "optim/total_tokens": 2442133504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 412087, "loss/out": 4.548164367675781, "created_at": "2025-01-15T02:07:46.554106+00:00"} {"global_step": 9317, "acc_step": 0, "speed/wps": 12955.55442046271, "speed/FLOPS": 203484784970740.4, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.002167820930481, "optim/lr": 0.002954647903333175, "optim/total_tokens": 2442395648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 4.628202438354492, "created_at": "2025-01-15T02:07:51.620022+00:00"} {"global_step": 9318, "acc_step": 0, "speed/wps": 12957.717190549454, "speed/FLOPS": 203518754246910.28, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.407799243927002, "optim/lr": 0.002954626991707293, "optim/total_tokens": 2442657792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428093, "loss/out": 4.691765308380127, "created_at": "2025-01-15T02:07:56.685003+00:00"} {"global_step": 9319, "acc_step": 0, "speed/wps": 12960.930893733594, "speed/FLOPS": 203569229871508.12, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.4419286251068115, "optim/lr": 0.002954606075335434, "optim/total_tokens": 2442919936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 4.779561996459961, "created_at": "2025-01-15T02:08:01.746097+00:00"} {"global_step": 9320, "acc_step": 0, "speed/wps": 12952.437119710352, "speed/FLOPS": 203435823478804.5, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1736435890197754, "optim/lr": 0.002954585154217667, "optim/total_tokens": 2443182080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450471, "loss/out": 4.689032554626465, "created_at": "2025-01-15T02:08:06.814202+00:00"} {"global_step": 9321, "acc_step": 0, "speed/wps": 12959.792843184867, "speed/FLOPS": 203551355223795.03, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4688650369644165, "optim/lr": 0.0029545642283540603, "optim/total_tokens": 2443444224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306907, "loss/out": 4.632195472717285, "created_at": "2025-01-15T02:08:11.876591+00:00"} {"global_step": 9322, "acc_step": 0, "speed/wps": 12952.726631357218, "speed/FLOPS": 203440370657049.22, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4442378282546997, "optim/lr": 0.002954543297744682, "optim/total_tokens": 2443706368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 405430, "loss/out": 4.807953357696533, "created_at": "2025-01-15T02:08:16.942703+00:00"} {"global_step": 9323, "acc_step": 0, "speed/wps": 12955.399515244962, "speed/FLOPS": 203482351971433.62, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.3691487312316895, "optim/lr": 0.0029545223623896003, "optim/total_tokens": 2443968512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428890, "loss/out": 4.634758949279785, "created_at": "2025-01-15T02:08:22.038002+00:00"} {"global_step": 9324, "acc_step": 0, "speed/wps": 12958.722021532656, "speed/FLOPS": 203534536498283.9, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9340290427207947, "optim/lr": 0.002954501422288885, "optim/total_tokens": 2444230656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432127, "loss/out": 4.677999019622803, "created_at": "2025-01-15T02:08:27.100797+00:00"} {"global_step": 9325, "acc_step": 0, "speed/wps": 12954.763768591982, "speed/FLOPS": 203472366696641.1, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2855345010757446, "optim/lr": 0.0029544804774426017, "optim/total_tokens": 2444492800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495515, "loss/out": 4.686345100402832, "created_at": "2025-01-15T02:08:32.163577+00:00"} {"global_step": 9326, "acc_step": 0, "speed/wps": 12953.952714727413, "speed/FLOPS": 203459627981193.25, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7808003425598145, "optim/lr": 0.0029544595278508208, "optim/total_tokens": 2444754944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427898, "loss/out": 4.579968452453613, "created_at": "2025-01-15T02:08:37.226958+00:00"} {"global_step": 9327, "acc_step": 0, "speed/wps": 12949.979705400923, "speed/FLOPS": 203397226410233.8, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0641379356384277, "optim/lr": 0.0029544385735136096, "optim/total_tokens": 2445017088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355618, "loss/out": 4.613253593444824, "created_at": "2025-01-15T02:08:42.291617+00:00"} {"global_step": 9328, "acc_step": 0, "speed/wps": 12957.568160385797, "speed/FLOPS": 203516413523401.12, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0217716693878174, "optim/lr": 0.0029544176144310377, "optim/total_tokens": 2445279232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447026, "loss/out": 4.63561487197876, "created_at": "2025-01-15T02:08:47.358988+00:00"} {"global_step": 9329, "acc_step": 0, "speed/wps": 12958.494803757005, "speed/FLOPS": 203530967730887.38, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.142235279083252, "optim/lr": 0.0029543966506031727, "optim/total_tokens": 2445541376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 4.6073126792907715, "created_at": "2025-01-15T02:08:52.423478+00:00"} {"global_step": 9330, "acc_step": 0, "speed/wps": 12955.589416808318, "speed/FLOPS": 203485334636438.2, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.9514111876487732, "optim/lr": 0.002954375682030082, "optim/total_tokens": 2445803520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451797, "loss/out": 4.592759132385254, "created_at": "2025-01-15T02:08:57.488484+00:00"} {"global_step": 9331, "acc_step": 0, "speed/wps": 12956.150801015752, "speed/FLOPS": 203494151946838.7, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.440403699874878, "optim/lr": 0.0029543547087118355, "optim/total_tokens": 2446065664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436916, "loss/out": 4.6491289138793945, "created_at": "2025-01-15T02:09:02.553646+00:00"} {"global_step": 9332, "acc_step": 0, "speed/wps": 12956.06789010368, "speed/FLOPS": 203492849717033.1, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1904525756835938, "optim/lr": 0.0029543337306485013, "optim/total_tokens": 2446327808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313070, "loss/out": 4.678897857666016, "created_at": "2025-01-15T02:09:07.620151+00:00"} {"global_step": 9333, "acc_step": 0, "speed/wps": 12956.659389181925, "speed/FLOPS": 203502140022861.8, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8507407307624817, "optim/lr": 0.0029543127478401476, "optim/total_tokens": 2446589952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432941, "loss/out": 4.507593154907227, "created_at": "2025-01-15T02:09:12.688341+00:00"} {"global_step": 9334, "acc_step": 0, "speed/wps": 12950.828419173356, "speed/FLOPS": 203410556626284.8, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.196150779724121, "optim/lr": 0.0029542917602868425, "optim/total_tokens": 2446852096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 4.589723587036133, "created_at": "2025-01-15T02:09:17.754815+00:00"} {"global_step": 9335, "acc_step": 0, "speed/wps": 12952.815147754074, "speed/FLOPS": 203441760928696.12, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.518681526184082, "optim/lr": 0.0029542707679886555, "optim/total_tokens": 2447114240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434003, "loss/out": 4.6294708251953125, "created_at": "2025-01-15T02:09:22.820942+00:00"} {"global_step": 9336, "acc_step": 0, "speed/wps": 12947.507059606372, "speed/FLOPS": 203358390110259.0, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1819837093353271, "optim/lr": 0.0029542497709456545, "optim/total_tokens": 2447376384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444878, "loss/out": 4.625277519226074, "created_at": "2025-01-15T02:09:27.887327+00:00"} {"global_step": 9337, "acc_step": 0, "speed/wps": 12956.251399041941, "speed/FLOPS": 203495731977076.16, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2803614139556885, "optim/lr": 0.0029542287691579077, "optim/total_tokens": 2447638528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 4.679966926574707, "created_at": "2025-01-15T02:09:32.949693+00:00"} {"global_step": 9338, "acc_step": 0, "speed/wps": 12954.231419646609, "speed/FLOPS": 203464005424929.9, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1236827373504639, "optim/lr": 0.002954207762625484, "optim/total_tokens": 2447900672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452330, "loss/out": 4.5972418785095215, "created_at": "2025-01-15T02:09:38.012773+00:00"} {"global_step": 9339, "acc_step": 0, "speed/wps": 12965.365061481674, "speed/FLOPS": 203638874569176.75, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1558884382247925, "optim/lr": 0.0029541867513484515, "optim/total_tokens": 2448162816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319751, "loss/out": 4.68890380859375, "created_at": "2025-01-15T02:09:43.074572+00:00"} {"global_step": 9340, "acc_step": 0, "speed/wps": 12950.29718962604, "speed/FLOPS": 203402212936258.3, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1556369066238403, "optim/lr": 0.002954165735326879, "optim/total_tokens": 2448424960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462435, "loss/out": 4.70243501663208, "created_at": "2025-01-15T02:09:48.138883+00:00"} {"global_step": 9341, "acc_step": 0, "speed/wps": 12956.936182695625, "speed/FLOPS": 203506487445348.62, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8370378017425537, "optim/lr": 0.002954144714560836, "optim/total_tokens": 2448687104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 4.675773620605469, "created_at": "2025-01-15T02:09:53.200796+00:00"} {"global_step": 9342, "acc_step": 0, "speed/wps": 12955.062721761098, "speed/FLOPS": 203477062166966.78, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2762556076049805, "optim/lr": 0.0029541236890503896, "optim/total_tokens": 2448949248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432953, "loss/out": 4.631923675537109, "created_at": "2025-01-15T02:09:58.265877+00:00"} {"global_step": 9343, "acc_step": 0, "speed/wps": 12949.599054674398, "speed/FLOPS": 203391247767504.8, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0640592575073242, "optim/lr": 0.0029541026587956098, "optim/total_tokens": 2449211392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475383, "loss/out": 4.709833145141602, "created_at": "2025-01-15T02:10:03.330411+00:00"} {"global_step": 9344, "acc_step": 0, "speed/wps": 12949.25615502038, "speed/FLOPS": 203385862057244.97, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.0746463537216187, "optim/lr": 0.0029540816237965638, "optim/total_tokens": 2449473536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321779, "loss/out": 4.710812091827393, "created_at": "2025-01-15T02:10:08.396097+00:00"} {"global_step": 9345, "acc_step": 0, "speed/wps": 12955.988901850244, "speed/FLOPS": 203491609097971.62, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8578015565872192, "optim/lr": 0.0029540605840533214, "optim/total_tokens": 2449735680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350938, "loss/out": 4.5636887550354, "created_at": "2025-01-15T02:10:13.458403+00:00"} {"global_step": 9346, "acc_step": 0, "speed/wps": 12953.761599698622, "speed/FLOPS": 203456626257046.53, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9584844708442688, "optim/lr": 0.00295403953956595, "optim/total_tokens": 2449997824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443307, "loss/out": 4.621251106262207, "created_at": "2025-01-15T02:10:18.525317+00:00"} {"global_step": 9347, "acc_step": 0, "speed/wps": 12964.694031151228, "speed/FLOPS": 203628335115744.78, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0639526844024658, "optim/lr": 0.0029540184903345196, "optim/total_tokens": 2450259968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 4.680488109588623, "created_at": "2025-01-15T02:10:23.584481+00:00"} {"global_step": 9348, "acc_step": 0, "speed/wps": 12960.325009611306, "speed/FLOPS": 203559713628795.56, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1639574766159058, "optim/lr": 0.0029539974363590985, "optim/total_tokens": 2450522112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442784, "loss/out": 4.59304666519165, "created_at": "2025-01-15T02:10:28.647099+00:00"} {"global_step": 9349, "acc_step": 0, "speed/wps": 12958.404644309992, "speed/FLOPS": 203529551652879.97, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8614176511764526, "optim/lr": 0.002953976377639755, "optim/total_tokens": 2450784256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446747, "loss/out": 4.618800640106201, "created_at": "2025-01-15T02:10:33.713838+00:00"} {"global_step": 9350, "acc_step": 0, "speed/wps": 12964.190384219082, "speed/FLOPS": 203620424648592.03, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2594785690307617, "optim/lr": 0.0029539553141765583, "optim/total_tokens": 2451046400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451523, "loss/out": 4.549930095672607, "created_at": "2025-01-15T02:10:38.773869+00:00"} {"global_step": 9351, "acc_step": 0, "speed/wps": 12964.791983896701, "speed/FLOPS": 203629873598213.6, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0965174436569214, "optim/lr": 0.0029539342459695764, "optim/total_tokens": 2451308544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481474, "loss/out": 4.541350841522217, "created_at": "2025-01-15T02:10:43.832277+00:00"} {"global_step": 9352, "acc_step": 0, "speed/wps": 12962.357599007106, "speed/FLOPS": 203591638238327.12, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6175782680511475, "optim/lr": 0.002953913173018878, "optim/total_tokens": 2451570688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322443, "loss/out": 4.612112045288086, "created_at": "2025-01-15T02:10:48.898279+00:00"} {"global_step": 9353, "acc_step": 0, "speed/wps": 12962.267669938914, "speed/FLOPS": 203590225778735.7, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0480401515960693, "optim/lr": 0.002953892095324533, "optim/total_tokens": 2451832832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 342518, "loss/out": 4.700008392333984, "created_at": "2025-01-15T02:10:53.957707+00:00"} {"global_step": 9354, "acc_step": 0, "speed/wps": 12960.975785576187, "speed/FLOPS": 203569934959583.9, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3953685760498047, "optim/lr": 0.002953871012886609, "optim/total_tokens": 2452094976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 4.595635414123535, "created_at": "2025-01-15T02:10:59.020876+00:00"} {"global_step": 9355, "acc_step": 0, "speed/wps": 12963.757537849857, "speed/FLOPS": 203613626201570.1, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0330201387405396, "optim/lr": 0.002953849925705176, "optim/total_tokens": 2452357120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 4.630832672119141, "created_at": "2025-01-15T02:11:04.082646+00:00"} {"global_step": 9356, "acc_step": 0, "speed/wps": 12968.358540437268, "speed/FLOPS": 203685891269645.47, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8760857582092285, "optim/lr": 0.0029538288337803017, "optim/total_tokens": 2452619264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462435, "loss/out": 4.62190055847168, "created_at": "2025-01-15T02:11:09.145579+00:00"} {"global_step": 9357, "acc_step": 0, "speed/wps": 12959.181442679808, "speed/FLOPS": 203541752338710.75, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.223954439163208, "optim/lr": 0.0029538077371120547, "optim/total_tokens": 2452881408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497434, "loss/out": 4.671621322631836, "created_at": "2025-01-15T02:11:14.209655+00:00"} {"global_step": 9358, "acc_step": 0, "speed/wps": 12965.989153481783, "speed/FLOPS": 203648676791630.72, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2045621871948242, "optim/lr": 0.0029537866357005047, "optim/total_tokens": 2453143552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443307, "loss/out": 4.583748817443848, "created_at": "2025-01-15T02:11:19.271393+00:00"} {"global_step": 9359, "acc_step": 0, "speed/wps": 12959.635626471325, "speed/FLOPS": 203548885919270.47, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8460394740104675, "optim/lr": 0.0029537655295457203, "optim/total_tokens": 2453405696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305561, "loss/out": 4.7176971435546875, "created_at": "2025-01-15T02:11:24.332679+00:00"} {"global_step": 9360, "acc_step": 0, "speed/wps": 12956.517215690297, "speed/FLOPS": 203499906992808.97, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1138280630111694, "optim/lr": 0.0029537444186477713, "optim/total_tokens": 2453667840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429438, "loss/out": 4.658812999725342, "created_at": "2025-01-15T02:11:29.397107+00:00"} {"global_step": 9361, "acc_step": 0, "speed/wps": 12962.537152859151, "speed/FLOPS": 203594458378306.34, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2153005599975586, "optim/lr": 0.002953723303006725, "optim/total_tokens": 2453929984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 4.480025768280029, "created_at": "2025-01-15T02:11:34.459705+00:00"} {"global_step": 9362, "acc_step": 0, "speed/wps": 12961.79914590258, "speed/FLOPS": 203582866964925.56, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.0786447525024414, "optim/lr": 0.00295370218262265, "optim/total_tokens": 2454192128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 4.576933860778809, "created_at": "2025-01-15T02:11:39.522371+00:00"} {"global_step": 9363, "acc_step": 0, "speed/wps": 12959.256307627174, "speed/FLOPS": 203542928195584.38, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0008, "optim/grad_norm": 4.3437910079956055, "optim/lr": 0.0029536810574956173, "optim/total_tokens": 2454454272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479622, "loss/out": 4.557765960693359, "created_at": "2025-01-15T02:11:44.582915+00:00"} {"global_step": 9364, "acc_step": 0, "speed/wps": 12962.469566177739, "speed/FLOPS": 203593396836600.94, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8402235507965088, "optim/lr": 0.002953659927625694, "optim/total_tokens": 2454716416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444119, "loss/out": 4.6805419921875, "created_at": "2025-01-15T02:11:49.650372+00:00"} {"global_step": 9365, "acc_step": 0, "speed/wps": 12962.165930275709, "speed/FLOPS": 203588627817520.1, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8860173225402832, "optim/lr": 0.0029536387930129498, "optim/total_tokens": 2454978560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 4.6543169021606445, "created_at": "2025-01-15T02:11:54.711114+00:00"} {"global_step": 9366, "acc_step": 0, "speed/wps": 12964.016444969086, "speed/FLOPS": 203617692693653.25, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5281342267990112, "optim/lr": 0.002953617653657453, "optim/total_tokens": 2455240704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456567, "loss/out": 4.680191993713379, "created_at": "2025-01-15T02:11:59.774170+00:00"} {"global_step": 9367, "acc_step": 0, "speed/wps": 12962.959924081068, "speed/FLOPS": 203601098581295.1, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9621877670288086, "optim/lr": 0.002953596509559274, "optim/total_tokens": 2455502848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310132, "loss/out": 4.698847770690918, "created_at": "2025-01-15T02:12:04.833858+00:00"} {"global_step": 9368, "acc_step": 0, "speed/wps": 12965.37356819813, "speed/FLOPS": 203639008178848.12, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7669997215270996, "optim/lr": 0.0029535753607184806, "optim/total_tokens": 2455764992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 346546, "loss/out": 4.5734028816223145, "created_at": "2025-01-15T02:12:09.892096+00:00"} {"global_step": 9369, "acc_step": 0, "speed/wps": 12955.962545612027, "speed/FLOPS": 203491195137033.16, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.5713639855384827, "optim/lr": 0.002953554207135142, "optim/total_tokens": 2456027136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446742, "loss/out": 4.596098899841309, "created_at": "2025-01-15T02:12:14.954162+00:00"} {"global_step": 9370, "acc_step": 0, "speed/wps": 12952.141825536875, "speed/FLOPS": 203431185478033.34, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0654371976852417, "optim/lr": 0.002953533048809327, "optim/total_tokens": 2456289280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 4.526413440704346, "created_at": "2025-01-15T02:12:20.018278+00:00"} {"global_step": 9371, "acc_step": 0, "speed/wps": 12960.189735094558, "speed/FLOPS": 203557588956623.38, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2527503967285156, "optim/lr": 0.0029535118857411057, "optim/total_tokens": 2456551424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 4.569769382476807, "created_at": "2025-01-15T02:12:25.083377+00:00"} {"global_step": 9372, "acc_step": 0, "speed/wps": 12957.191629929142, "speed/FLOPS": 203510499595172.3, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0062373876571655, "optim/lr": 0.002953490717930546, "optim/total_tokens": 2456813568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 4.623693466186523, "created_at": "2025-01-15T02:12:30.149161+00:00"} {"global_step": 9373, "acc_step": 0, "speed/wps": 12966.181011749997, "speed/FLOPS": 203651690189373.9, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.1427407264709473, "optim/lr": 0.0029534695453777176, "optim/total_tokens": 2457075712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 4.4454498291015625, "created_at": "2025-01-15T02:12:35.211780+00:00"} {"global_step": 9374, "acc_step": 0, "speed/wps": 12962.46903131333, "speed/FLOPS": 203593388435820.4, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.7737619876861572, "optim/lr": 0.002953448368082689, "optim/total_tokens": 2457337856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326727, "loss/out": 4.604393005371094, "created_at": "2025-01-15T02:12:40.276153+00:00"} {"global_step": 9375, "acc_step": 0, "speed/wps": 12958.211885034067, "speed/FLOPS": 203526524103572.56, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.795650064945221, "optim/lr": 0.0029534271860455295, "optim/total_tokens": 2457600000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 341175, "loss/out": 4.50791072845459, "created_at": "2025-01-15T02:12:45.339529+00:00"} {"global_step": 9376, "acc_step": 0, "speed/wps": 12961.512878938689, "speed/FLOPS": 203578370748885.62, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8254219889640808, "optim/lr": 0.0029534059992663082, "optim/total_tokens": 2457862144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337829, "loss/out": 4.552092552185059, "created_at": "2025-01-15T02:12:50.405542+00:00"} {"global_step": 9377, "acc_step": 0, "speed/wps": 12957.14811706047, "speed/FLOPS": 203509816165776.84, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.2958292961120605, "optim/lr": 0.002953384807745095, "optim/total_tokens": 2458124288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 412087, "loss/out": 4.673070907592773, "created_at": "2025-01-15T02:12:55.469957+00:00"} {"global_step": 9378, "acc_step": 0, "speed/wps": 12969.257518427594, "speed/FLOPS": 203700010954308.75, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4208900928497314, "optim/lr": 0.0029533636114819576, "optim/total_tokens": 2458386432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430268, "loss/out": 4.620223045349121, "created_at": "2025-01-15T02:13:00.532214+00:00"} {"global_step": 9379, "acc_step": 0, "speed/wps": 12962.844533704321, "speed/FLOPS": 203599286216865.28, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4852848052978516, "optim/lr": 0.0029533424104769665, "optim/total_tokens": 2458648576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479906, "loss/out": 4.591158866882324, "created_at": "2025-01-15T02:13:05.595077+00:00"} {"global_step": 9380, "acc_step": 0, "speed/wps": 12965.723232331105, "speed/FLOPS": 203644500134545.38, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2952524423599243, "optim/lr": 0.0029533212047301903, "optim/total_tokens": 2458910720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437432, "loss/out": 4.564552307128906, "created_at": "2025-01-15T02:13:10.657965+00:00"} {"global_step": 9381, "acc_step": 0, "speed/wps": 12959.194385096514, "speed/FLOPS": 203541955617149.88, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7490708827972412, "optim/lr": 0.002953299994241698, "optim/total_tokens": 2459172864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 4.727412700653076, "created_at": "2025-01-15T02:13:15.720136+00:00"} {"global_step": 9382, "acc_step": 0, "speed/wps": 12959.181034575258, "speed/FLOPS": 203541745928868.0, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.0385382175445557, "optim/lr": 0.00295327877901156, "optim/total_tokens": 2459435008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432127, "loss/out": 4.680583953857422, "created_at": "2025-01-15T02:13:20.782077+00:00"} {"global_step": 9383, "acc_step": 0, "speed/wps": 12958.549381591538, "speed/FLOPS": 203531824950776.53, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0381314754486084, "optim/lr": 0.002953257559039843, "optim/total_tokens": 2459697152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350526, "loss/out": 4.600529670715332, "created_at": "2025-01-15T02:13:25.846920+00:00"} {"global_step": 9384, "acc_step": 0, "speed/wps": 12962.53091112019, "speed/FLOPS": 203594360343218.9, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3166964054107666, "optim/lr": 0.0029532363343266186, "optim/total_tokens": 2459959296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429454, "loss/out": 4.531903266906738, "created_at": "2025-01-15T02:13:30.909069+00:00"} {"global_step": 9385, "acc_step": 0, "speed/wps": 12965.670237660474, "speed/FLOPS": 203643667780420.78, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2517273426055908, "optim/lr": 0.0029532151048719552, "optim/total_tokens": 2460221440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495278, "loss/out": 4.655661106109619, "created_at": "2025-01-15T02:13:35.967193+00:00"} {"global_step": 9386, "acc_step": 0, "speed/wps": 12955.270356453935, "speed/FLOPS": 203480323355137.75, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9183549284934998, "optim/lr": 0.0029531938706759218, "optim/total_tokens": 2460483584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436366, "loss/out": 4.6216139793396, "created_at": "2025-01-15T02:13:41.033112+00:00"} {"global_step": 9387, "acc_step": 0, "speed/wps": 12959.98948043062, "speed/FLOPS": 203554443681947.5, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0759738683700562, "optim/lr": 0.0029531726317385877, "optim/total_tokens": 2460745728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479662, "loss/out": 4.618454933166504, "created_at": "2025-01-15T02:13:46.098214+00:00"} {"global_step": 9388, "acc_step": 0, "speed/wps": 12959.26493520626, "speed/FLOPS": 203543063703569.44, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.0366243124008179, "optim/lr": 0.0029531513880600236, "optim/total_tokens": 2461007872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446503, "loss/out": 4.724563121795654, "created_at": "2025-01-15T02:13:51.161234+00:00"} {"global_step": 9389, "acc_step": 0, "speed/wps": 12963.852834987396, "speed/FLOPS": 203615122974067.12, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1431949138641357, "optim/lr": 0.0029531301396402967, "optim/total_tokens": 2461270016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445691, "loss/out": 4.6547322273254395, "created_at": "2025-01-15T02:13:56.223727+00:00"} {"global_step": 9390, "acc_step": 0, "speed/wps": 12963.784312712318, "speed/FLOPS": 203614046737577.56, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8931114077568054, "optim/lr": 0.0029531088864794774, "optim/total_tokens": 2461532160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327402, "loss/out": 4.525599956512451, "created_at": "2025-01-15T02:14:01.288217+00:00"} {"global_step": 9391, "acc_step": 0, "speed/wps": 12970.541615625521, "speed/FLOPS": 203720179465336.7, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7092084884643555, "optim/lr": 0.002953087628577635, "optim/total_tokens": 2461794304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290078, "loss/out": 4.52569055557251, "created_at": "2025-01-15T02:14:06.347154+00:00"} {"global_step": 9392, "acc_step": 0, "speed/wps": 12962.598780205864, "speed/FLOPS": 203595426320471.3, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9852710366249084, "optim/lr": 0.0029530663659348387, "optim/total_tokens": 2462056448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445406, "loss/out": 4.539473533630371, "created_at": "2025-01-15T02:14:11.411934+00:00"} {"global_step": 9393, "acc_step": 0, "speed/wps": 12955.695807146247, "speed/FLOPS": 203487005642891.03, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1170047521591187, "optim/lr": 0.0029530450985511586, "optim/total_tokens": 2462318592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 4.49729061126709, "created_at": "2025-01-15T02:14:16.477842+00:00"} {"global_step": 9394, "acc_step": 0, "speed/wps": 12966.193736403002, "speed/FLOPS": 203651890047534.84, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2427146434783936, "optim/lr": 0.002953023826426663, "optim/total_tokens": 2462580736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444357, "loss/out": 4.546961784362793, "created_at": "2025-01-15T02:14:21.535751+00:00"} {"global_step": 9395, "acc_step": 0, "speed/wps": 12964.119374443853, "speed/FLOPS": 203619309342494.4, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.068554162979126, "optim/lr": 0.002953002549561422, "optim/total_tokens": 2462842880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480229, "loss/out": 4.536059379577637, "created_at": "2025-01-15T02:14:26.603438+00:00"} {"global_step": 9396, "acc_step": 0, "speed/wps": 12964.03527719583, "speed/FLOPS": 203617988479652.44, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3027360439300537, "optim/lr": 0.0029529812679555045, "optim/total_tokens": 2463105024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436381, "loss/out": 4.641528129577637, "created_at": "2025-01-15T02:14:31.663793+00:00"} {"global_step": 9397, "acc_step": 0, "speed/wps": 12964.250283722486, "speed/FLOPS": 203621365452603.5, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8897423148155212, "optim/lr": 0.0029529599816089805, "optim/total_tokens": 2463367168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 280160, "loss/out": 4.589844703674316, "created_at": "2025-01-15T02:14:36.724059+00:00"} {"global_step": 9398, "acc_step": 0, "speed/wps": 12967.280704126968, "speed/FLOPS": 203668962369289.94, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0473132133483887, "optim/lr": 0.0029529386905219187, "optim/total_tokens": 2463629312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 396273, "loss/out": 4.569197177886963, "created_at": "2025-01-15T02:14:41.787311+00:00"} {"global_step": 9399, "acc_step": 0, "speed/wps": 12957.0862480763, "speed/FLOPS": 203508844428363.6, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1170583963394165, "optim/lr": 0.0029529173946943894, "optim/total_tokens": 2463891456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 4.622350692749023, "created_at": "2025-01-15T02:14:46.849092+00:00"} {"global_step": 9400, "acc_step": 0, "speed/wps": 12964.785947797825, "speed/FLOPS": 203629778792986.3, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.4766464233398438, "optim/lr": 0.002952896094126462, "optim/total_tokens": 2464153600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457902, "loss/out": 4.606375694274902, "created_at": "2025-01-15T02:14:51.917968+00:00"} {"global_step": 9401, "acc_step": 0, "speed/wps": 12960.86608340023, "speed/FLOPS": 203568211936163.9, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0372464656829834, "optim/lr": 0.002952874788818205, "optim/total_tokens": 2464415744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 4.525717735290527, "created_at": "2025-01-15T02:14:56.982129+00:00"} {"global_step": 9402, "acc_step": 0, "speed/wps": 12964.2012873731, "speed/FLOPS": 203620595897608.38, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9900413751602173, "optim/lr": 0.002952853478769689, "optim/total_tokens": 2464677888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 4.489462852478027, "created_at": "2025-01-15T02:15:02.044199+00:00"} {"global_step": 9403, "acc_step": 0, "speed/wps": 12963.941603828536, "speed/FLOPS": 203616517210698.4, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0238571166992188, "optim/lr": 0.002952832163980983, "optim/total_tokens": 2464940032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 4.588985919952393, "created_at": "2025-01-15T02:15:07.104258+00:00"} {"global_step": 9404, "acc_step": 0, "speed/wps": 12970.756448314167, "speed/FLOPS": 203723553707922.9, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0367568731307983, "optim/lr": 0.0029528108444521572, "optim/total_tokens": 2465202176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443022, "loss/out": 4.762360572814941, "created_at": "2025-01-15T02:15:12.164670+00:00"} {"global_step": 9405, "acc_step": 0, "speed/wps": 12958.403055040855, "speed/FLOPS": 203529526691224.28, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7741110324859619, "optim/lr": 0.00295278952018328, "optim/total_tokens": 2465464320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298347, "loss/out": 4.571595191955566, "created_at": "2025-01-15T02:15:17.230775+00:00"} {"global_step": 9406, "acc_step": 0, "speed/wps": 12970.913526120452, "speed/FLOPS": 203726020830718.12, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.147454023361206, "optim/lr": 0.002952768191174422, "optim/total_tokens": 2465726464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376530, "loss/out": 4.4699931144714355, "created_at": "2025-01-15T02:15:22.289948+00:00"} {"global_step": 9407, "acc_step": 0, "speed/wps": 12969.13049253932, "speed/FLOPS": 203698015838181.56, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2255010604858398, "optim/lr": 0.002952746857425652, "optim/total_tokens": 2465988608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430506, "loss/out": 4.4731974601745605, "created_at": "2025-01-15T02:15:27.347003+00:00"} {"global_step": 9408, "acc_step": 0, "speed/wps": 12964.253032821794, "speed/FLOPS": 203621408630985.8, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9094948768615723, "optim/lr": 0.0029527255189370407, "optim/total_tokens": 2466250752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491052, "loss/out": 4.569845199584961, "created_at": "2025-01-15T02:15:32.406905+00:00"} {"global_step": 9409, "acc_step": 0, "speed/wps": 12969.170861466413, "speed/FLOPS": 203698649887654.28, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.434936761856079, "optim/lr": 0.002952704175708657, "optim/total_tokens": 2466512896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460840, "loss/out": 4.581347465515137, "created_at": "2025-01-15T02:15:37.467262+00:00"} {"global_step": 9410, "acc_step": 0, "speed/wps": 12961.291338289497, "speed/FLOPS": 203574891148561.3, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.156083345413208, "optim/lr": 0.00295268282774057, "optim/total_tokens": 2466775040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 4.581536293029785, "created_at": "2025-01-15T02:15:42.529685+00:00"} {"global_step": 9411, "acc_step": 0, "speed/wps": 12960.427780221446, "speed/FLOPS": 203561327782449.44, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1722418069839478, "optim/lr": 0.0029526614750328504, "optim/total_tokens": 2467037184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452081, "loss/out": 4.627257347106934, "created_at": "2025-01-15T02:15:47.622171+00:00"} {"global_step": 9412, "acc_step": 0, "speed/wps": 12956.256279771287, "speed/FLOPS": 203495808635637.38, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.029900312423706, "optim/lr": 0.002952640117585567, "optim/total_tokens": 2467299328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 4.6517534255981445, "created_at": "2025-01-15T02:15:52.688818+00:00"} {"global_step": 9413, "acc_step": 0, "speed/wps": 12955.96711840809, "speed/FLOPS": 203491266959079.03, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4196316003799438, "optim/lr": 0.00295261875539879, "optim/total_tokens": 2467561472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 4.556980609893799, "created_at": "2025-01-15T02:15:57.755510+00:00"} {"global_step": 9414, "acc_step": 0, "speed/wps": 12959.78149634004, "speed/FLOPS": 203551177006004.84, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4227263927459717, "optim/lr": 0.002952597388472589, "optim/total_tokens": 2467823616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 4.579303741455078, "created_at": "2025-01-15T02:16:02.821554+00:00"} {"global_step": 9415, "acc_step": 0, "speed/wps": 12963.396246841832, "speed/FLOPS": 203607951629823.2, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.015673875808716, "optim/lr": 0.0029525760168070342, "optim/total_tokens": 2468085760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 4.585484504699707, "created_at": "2025-01-15T02:16:07.880846+00:00"} {"global_step": 9416, "acc_step": 0, "speed/wps": 12963.750041099247, "speed/FLOPS": 203613508454800.16, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.0522353649139404, "optim/lr": 0.002952554640402194, "optim/total_tokens": 2468347904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 4.7027716636657715, "created_at": "2025-01-15T02:16:12.941801+00:00"} {"global_step": 9417, "acc_step": 0, "speed/wps": 12957.299203736553, "speed/FLOPS": 203512189189639.6, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1137661933898926, "optim/lr": 0.00295253325925814, "optim/total_tokens": 2468610048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 4.536357879638672, "created_at": "2025-01-15T02:16:18.003264+00:00"} {"global_step": 9418, "acc_step": 0, "speed/wps": 12960.472229361352, "speed/FLOPS": 203562025917271.2, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.246261715888977, "optim/lr": 0.00295251187337494, "optim/total_tokens": 2468872192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324713, "loss/out": 4.508203983306885, "created_at": "2025-01-15T02:16:23.069467+00:00"} {"global_step": 9419, "acc_step": 0, "speed/wps": 12959.023500979074, "speed/FLOPS": 203539271647266.1, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0479180812835693, "optim/lr": 0.002952490482752665, "optim/total_tokens": 2469134336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 400071, "loss/out": 4.645040512084961, "created_at": "2025-01-15T02:16:28.132566+00:00"} {"global_step": 9420, "acc_step": 0, "speed/wps": 12949.712422040247, "speed/FLOPS": 203393028357768.56, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2297428846359253, "optim/lr": 0.0029524690873913846, "optim/total_tokens": 2469396480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467734, "loss/out": 4.617763519287109, "created_at": "2025-01-15T02:16:33.201587+00:00"} {"global_step": 9421, "acc_step": 0, "speed/wps": 12957.541294281677, "speed/FLOPS": 203515991554318.03, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2180635929107666, "optim/lr": 0.002952447687291169, "optim/total_tokens": 2469658624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447269, "loss/out": 4.6123480796813965, "created_at": "2025-01-15T02:16:38.267313+00:00"} {"global_step": 9422, "acc_step": 0, "speed/wps": 12958.324274551504, "speed/FLOPS": 203528289335383.28, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3457468748092651, "optim/lr": 0.0029524262824520866, "optim/total_tokens": 2469920768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 4.494444847106934, "created_at": "2025-01-15T02:16:43.331451+00:00"} {"global_step": 9423, "acc_step": 0, "speed/wps": 12958.614562508392, "speed/FLOPS": 203532848706643.7, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0014, "optim/grad_norm": 1.1404720544815063, "optim/lr": 0.0029524048728742083, "optim/total_tokens": 2470182912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311717, "loss/out": 4.5762529373168945, "created_at": "2025-01-15T02:16:48.396415+00:00"} {"global_step": 9424, "acc_step": 0, "speed/wps": 12958.244950959239, "speed/FLOPS": 203527043449364.75, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0637811422348022, "optim/lr": 0.002952383458557604, "optim/total_tokens": 2470445056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352115, "loss/out": 4.558959007263184, "created_at": "2025-01-15T02:16:53.459882+00:00"} {"global_step": 9425, "acc_step": 0, "speed/wps": 12958.287151600307, "speed/FLOPS": 203527706268422.47, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2819960117340088, "optim/lr": 0.0029523620395023437, "optim/total_tokens": 2470707200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432365, "loss/out": 4.509948253631592, "created_at": "2025-01-15T02:16:58.525637+00:00"} {"global_step": 9426, "acc_step": 0, "speed/wps": 12958.99133079333, "speed/FLOPS": 203538766370292.5, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0738751888275146, "optim/lr": 0.0029523406157084967, "optim/total_tokens": 2470969344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 4.5817108154296875, "created_at": "2025-01-15T02:17:03.589069+00:00"} {"global_step": 9427, "acc_step": 0, "speed/wps": 12960.608560100196, "speed/FLOPS": 203564167178862.78, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.030040979385376, "optim/lr": 0.0029523191871761332, "optim/total_tokens": 2471231488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419902, "loss/out": 4.683256149291992, "created_at": "2025-01-15T02:17:08.652828+00:00"} {"global_step": 9428, "acc_step": 0, "speed/wps": 12960.090469857372, "speed/FLOPS": 203556029859668.66, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9619171619415283, "optim/lr": 0.0029522977539053227, "optim/total_tokens": 2471493632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 4.5532097816467285, "created_at": "2025-01-15T02:17:13.716206+00:00"} {"global_step": 9429, "acc_step": 0, "speed/wps": 12955.444702706422, "speed/FLOPS": 203483061702609.9, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0810396671295166, "optim/lr": 0.0029522763158961363, "optim/total_tokens": 2471755776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 4.676608562469482, "created_at": "2025-01-15T02:17:18.779702+00:00"} {"global_step": 9430, "acc_step": 0, "speed/wps": 12955.641179402503, "speed/FLOPS": 203486147639109.12, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1198363304138184, "optim/lr": 0.0029522548731486424, "optim/total_tokens": 2472017920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455517, "loss/out": 4.535196781158447, "created_at": "2025-01-15T02:17:23.842993+00:00"} {"global_step": 9431, "acc_step": 0, "speed/wps": 12960.400813937864, "speed/FLOPS": 203560904239910.22, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.086094856262207, "optim/lr": 0.002952233425662912, "optim/total_tokens": 2472280064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 4.542003631591797, "created_at": "2025-01-15T02:17:28.903964+00:00"} {"global_step": 9432, "acc_step": 0, "speed/wps": 12953.320810673766, "speed/FLOPS": 203449703059703.56, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 251.60409545898438, "optim/lr": 0.0029522119734390144, "optim/total_tokens": 2472542208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352115, "loss/out": 4.552660942077637, "created_at": "2025-01-15T02:17:33.970593+00:00"} {"global_step": 9433, "acc_step": 0, "speed/wps": 12961.77377833284, "speed/FLOPS": 203582468532384.1, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.201677918434143, "optim/lr": 0.002952190516477021, "optim/total_tokens": 2472804352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448890, "loss/out": 4.54936408996582, "created_at": "2025-01-15T02:17:39.030881+00:00"} {"global_step": 9434, "acc_step": 0, "speed/wps": 12955.536898508462, "speed/FLOPS": 203484509764371.28, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2247159481048584, "optim/lr": 0.002952169054777, "optim/total_tokens": 2473066496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 4.593417167663574, "created_at": "2025-01-15T02:17:44.096745+00:00"} {"global_step": 9435, "acc_step": 0, "speed/wps": 12952.125082837123, "speed/FLOPS": 203430922510926.84, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1738331317901611, "optim/lr": 0.0029521475883390225, "optim/total_tokens": 2473328640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441679, "loss/out": 4.574080944061279, "created_at": "2025-01-15T02:17:49.164853+00:00"} {"global_step": 9436, "acc_step": 0, "speed/wps": 12960.294790291991, "speed/FLOPS": 203559238992860.03, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1439934968948364, "optim/lr": 0.002952126117163158, "optim/total_tokens": 2473590784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 4.678252220153809, "created_at": "2025-01-15T02:17:54.229409+00:00"} {"global_step": 9437, "acc_step": 0, "speed/wps": 12960.698695313758, "speed/FLOPS": 203565582876250.56, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0299780368804932, "optim/lr": 0.0029521046412494764, "optim/total_tokens": 2473852928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302869, "loss/out": 4.724144458770752, "created_at": "2025-01-15T02:17:59.289490+00:00"} {"global_step": 9438, "acc_step": 0, "speed/wps": 12960.00105916684, "speed/FLOPS": 203554625541910.66, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0476504564285278, "optim/lr": 0.0029520831605980486, "optim/total_tokens": 2474115072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380554, "loss/out": 4.4330525398254395, "created_at": "2025-01-15T02:18:04.350065+00:00"} {"global_step": 9439, "acc_step": 0, "speed/wps": 12964.673632477554, "speed/FLOPS": 203628014726544.0, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4100550413131714, "optim/lr": 0.0029520616752089445, "optim/total_tokens": 2474377216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434230, "loss/out": 4.531634330749512, "created_at": "2025-01-15T02:18:09.409820+00:00"} {"global_step": 9440, "acc_step": 0, "speed/wps": 12959.630447239606, "speed/FLOPS": 203548804572319.3, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1070517301559448, "optim/lr": 0.002952040185082234, "optim/total_tokens": 2474639360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 4.530429363250732, "created_at": "2025-01-15T02:18:14.472764+00:00"} {"global_step": 9441, "acc_step": 0, "speed/wps": 12954.406078965876, "speed/FLOPS": 203466748689546.06, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0011, "optim/grad_norm": 1.7994565963745117, "optim/lr": 0.0029520186902179868, "optim/total_tokens": 2474901504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439583, "loss/out": 4.578150749206543, "created_at": "2025-01-15T02:18:19.536284+00:00"} {"global_step": 9442, "acc_step": 0, "speed/wps": 12958.129770975967, "speed/FLOPS": 203525234389453.56, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2444158792495728, "optim/lr": 0.0029519971906162737, "optim/total_tokens": 2475163648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437717, "loss/out": 4.53851318359375, "created_at": "2025-01-15T02:18:24.597501+00:00"} {"global_step": 9443, "acc_step": 0, "speed/wps": 12957.819404413332, "speed/FLOPS": 203520359656099.03, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1508140563964844, "optim/lr": 0.0029519756862771643, "optim/total_tokens": 2475425792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448605, "loss/out": 4.550303936004639, "created_at": "2025-01-15T02:18:29.662926+00:00"} {"global_step": 9444, "acc_step": 0, "speed/wps": 12958.059204591334, "speed/FLOPS": 203524126047414.53, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0086584091186523, "optim/lr": 0.0029519541772007295, "optim/total_tokens": 2475687936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 4.53298807144165, "created_at": "2025-01-15T02:18:34.724018+00:00"} {"global_step": 9445, "acc_step": 0, "speed/wps": 12956.004225682173, "speed/FLOPS": 203491849779809.22, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.247292160987854, "optim/lr": 0.0029519326633870384, "optim/total_tokens": 2475950080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440390, "loss/out": 4.617069721221924, "created_at": "2025-01-15T02:18:39.787412+00:00"} {"global_step": 9446, "acc_step": 0, "speed/wps": 12955.706688860817, "speed/FLOPS": 203487176555171.38, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1937098503112793, "optim/lr": 0.002951911144836162, "optim/total_tokens": 2476212224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 285796, "loss/out": 4.509634971618652, "created_at": "2025-01-15T02:18:44.849462+00:00"} {"global_step": 9447, "acc_step": 0, "speed/wps": 12962.83163891793, "speed/FLOPS": 203599083686525.72, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4411239624023438, "optim/lr": 0.00295188962154817, "optim/total_tokens": 2476474368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 410699, "loss/out": 4.57757568359375, "created_at": "2025-01-15T02:18:49.913647+00:00"} {"global_step": 9448, "acc_step": 0, "speed/wps": 12958.147352440777, "speed/FLOPS": 203525510530517.88, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.4864070415496826, "optim/lr": 0.002951868093523134, "optim/total_tokens": 2476736512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456045, "loss/out": 4.452516078948975, "created_at": "2025-01-15T02:18:54.988705+00:00"} {"global_step": 9449, "acc_step": 0, "speed/wps": 12955.255675958053, "speed/FLOPS": 203480092777776.8, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1514190435409546, "optim/lr": 0.0029518465607611217, "optim/total_tokens": 2476998656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 4.5615949630737305, "created_at": "2025-01-15T02:19:00.056595+00:00"} {"global_step": 9450, "acc_step": 0, "speed/wps": 12963.318656993366, "speed/FLOPS": 203606732974633.4, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2519482374191284, "optim/lr": 0.0029518250232622055, "optim/total_tokens": 2477260800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 4.650249481201172, "created_at": "2025-01-15T02:19:05.117249+00:00"} {"global_step": 9451, "acc_step": 0, "speed/wps": 12957.865045411923, "speed/FLOPS": 203521076510699.22, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1537399291992188, "optim/lr": 0.0029518034810264546, "optim/total_tokens": 2477522944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334895, "loss/out": 4.444339752197266, "created_at": "2025-01-15T02:19:10.178683+00:00"} {"global_step": 9452, "acc_step": 0, "speed/wps": 12961.169555753066, "speed/FLOPS": 203572978386479.9, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.4977097511291504, "optim/lr": 0.00295178193405394, "optim/total_tokens": 2477785088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355325, "loss/out": 4.432947158813477, "created_at": "2025-01-15T02:19:15.244676+00:00"} {"global_step": 9453, "acc_step": 0, "speed/wps": 12964.7311382708, "speed/FLOPS": 203628917934048.2, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5460153818130493, "optim/lr": 0.002951760382344732, "optim/total_tokens": 2478047232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435044, "loss/out": 4.788379192352295, "created_at": "2025-01-15T02:19:20.308718+00:00"} {"global_step": 9454, "acc_step": 0, "speed/wps": 12962.455674044166, "speed/FLOPS": 203593178641553.38, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.811055064201355, "optim/lr": 0.0029517388258989, "optim/total_tokens": 2478309376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 4.625064849853516, "created_at": "2025-01-15T02:19:25.373362+00:00"} {"global_step": 9455, "acc_step": 0, "speed/wps": 12959.840658157545, "speed/FLOPS": 203552106223645.62, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9000965356826782, "optim/lr": 0.0029517172647165156, "optim/total_tokens": 2478571520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448077, "loss/out": 4.671177864074707, "created_at": "2025-01-15T02:19:30.445627+00:00"} {"global_step": 9456, "acc_step": 0, "speed/wps": 12963.563721331371, "speed/FLOPS": 203610582046815.47, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.001, "optim/grad_norm": 0.8643995523452759, "optim/lr": 0.0029516956987976476, "optim/total_tokens": 2478833664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440390, "loss/out": 4.5599446296691895, "created_at": "2025-01-15T02:19:35.507758+00:00"} {"global_step": 9457, "acc_step": 0, "speed/wps": 12958.146428983175, "speed/FLOPS": 203525496026347.28, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3437094688415527, "optim/lr": 0.002951674128142368, "optim/total_tokens": 2479095808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462963, "loss/out": 4.595737457275391, "created_at": "2025-01-15T02:19:40.569359+00:00"} {"global_step": 9458, "acc_step": 0, "speed/wps": 12952.227666226701, "speed/FLOPS": 203432533724024.53, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1734782457351685, "optim/lr": 0.0029516525527507457, "optim/total_tokens": 2479357952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 4.550297260284424, "created_at": "2025-01-15T02:19:45.641359+00:00"} {"global_step": 9459, "acc_step": 0, "speed/wps": 12961.19854968227, "speed/FLOPS": 203573433775979.9, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8016765713691711, "optim/lr": 0.0029516309726228523, "optim/total_tokens": 2479620096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290365, "loss/out": 4.534041404724121, "created_at": "2025-01-15T02:19:50.707662+00:00"} {"global_step": 9460, "acc_step": 0, "speed/wps": 12957.335672063497, "speed/FLOPS": 203512761974826.97, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8107720613479614, "optim/lr": 0.0029516093877587574, "optim/total_tokens": 2479882240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439586, "loss/out": 4.6014814376831055, "created_at": "2025-01-15T02:19:55.770511+00:00"} {"global_step": 9461, "acc_step": 0, "speed/wps": 12961.176001489803, "speed/FLOPS": 203573079625632.75, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9414114952087402, "optim/lr": 0.0029515877981585318, "optim/total_tokens": 2480144384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 4.532098770141602, "created_at": "2025-01-15T02:20:00.834057+00:00"} {"global_step": 9462, "acc_step": 0, "speed/wps": 12957.96936617476, "speed/FLOPS": 203522715011631.3, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4016262292861938, "optim/lr": 0.002951566203822246, "optim/total_tokens": 2480406528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436666, "loss/out": 4.615791320800781, "created_at": "2025-01-15T02:20:05.895463+00:00"} {"global_step": 9463, "acc_step": 0, "speed/wps": 12958.738685835431, "speed/FLOPS": 203534798234058.16, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3905717134475708, "optim/lr": 0.0029515446047499705, "optim/total_tokens": 2480668672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 4.6307549476623535, "created_at": "2025-01-15T02:20:10.958370+00:00"} {"global_step": 9464, "acc_step": 0, "speed/wps": 12959.368507624305, "speed/FLOPS": 203544690450718.25, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.0668630599975586, "optim/lr": 0.002951523000941775, "optim/total_tokens": 2480930816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457380, "loss/out": 4.58049201965332, "created_at": "2025-01-15T02:20:16.019391+00:00"} {"global_step": 9465, "acc_step": 0, "speed/wps": 12966.307470967296, "speed/FLOPS": 203653676405154.78, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4599573612213135, "optim/lr": 0.0029515013923977307, "optim/total_tokens": 2481192960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319090, "loss/out": 4.505206108093262, "created_at": "2025-01-15T02:20:21.078603+00:00"} {"global_step": 9466, "acc_step": 0, "speed/wps": 12956.544778915879, "speed/FLOPS": 203500339911142.56, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.129157543182373, "optim/lr": 0.002951479779117908, "optim/total_tokens": 2481455104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364683, "loss/out": 4.7350592613220215, "created_at": "2025-01-15T02:20:26.140191+00:00"} {"global_step": 9467, "acc_step": 0, "speed/wps": 12964.341190916526, "speed/FLOPS": 203622793275005.5, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7484077215194702, "optim/lr": 0.002951458161102378, "optim/total_tokens": 2481717248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437438, "loss/out": 4.676762580871582, "created_at": "2025-01-15T02:20:31.200024+00:00"} {"global_step": 9468, "acc_step": 0, "speed/wps": 12965.28432109124, "speed/FLOPS": 203637606430396.03, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1000971794128418, "optim/lr": 0.00295143653835121, "optim/total_tokens": 2481979392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 4.662906646728516, "created_at": "2025-01-15T02:20:36.261909+00:00"} {"global_step": 9469, "acc_step": 0, "speed/wps": 12957.20274796966, "speed/FLOPS": 203510674219276.44, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1972696781158447, "optim/lr": 0.0029514149108644757, "optim/total_tokens": 2482241536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476746, "loss/out": 4.772494792938232, "created_at": "2025-01-15T02:20:41.324549+00:00"} {"global_step": 9470, "acc_step": 0, "speed/wps": 12963.174331360591, "speed/FLOPS": 203604466142246.47, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1312167644500732, "optim/lr": 0.0029513932786422443, "optim/total_tokens": 2482503680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 4.702506065368652, "created_at": "2025-01-15T02:20:46.386048+00:00"} {"global_step": 9471, "acc_step": 0, "speed/wps": 12966.758352914901, "speed/FLOPS": 203660758125718.62, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1479616165161133, "optim/lr": 0.002951371641684588, "optim/total_tokens": 2482765824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454979, "loss/out": 4.686810493469238, "created_at": "2025-01-15T02:20:51.446564+00:00"} {"global_step": 9472, "acc_step": 0, "speed/wps": 12964.084071666266, "speed/FLOPS": 203618754863861.28, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.06400465965271, "optim/lr": 0.0029513499999915763, "optim/total_tokens": 2483027968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448368, "loss/out": 4.492157936096191, "created_at": "2025-01-15T02:20:56.510309+00:00"} {"global_step": 9473, "acc_step": 0, "speed/wps": 12957.582640928089, "speed/FLOPS": 203516640960216.2, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1488025188446045, "optim/lr": 0.0029513283535632802, "optim/total_tokens": 2483290112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459237, "loss/out": 4.5039567947387695, "created_at": "2025-01-15T02:21:01.574056+00:00"} {"global_step": 9474, "acc_step": 0, "speed/wps": 12961.334974485095, "speed/FLOPS": 203575576514975.6, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6810904145240784, "optim/lr": 0.00295130670239977, "optim/total_tokens": 2483552256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309558, "loss/out": 4.499153137207031, "created_at": "2025-01-15T02:21:06.640192+00:00"} {"global_step": 9475, "acc_step": 0, "speed/wps": 12962.739033614022, "speed/FLOPS": 203597629192974.8, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.001, "optim/grad_norm": 1.6990692615509033, "optim/lr": 0.0029512850465011173, "optim/total_tokens": 2483814400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 414236, "loss/out": 4.76345157623291, "created_at": "2025-01-15T02:21:11.705176+00:00"} {"global_step": 9476, "acc_step": 0, "speed/wps": 12958.663545274965, "speed/FLOPS": 203533618048302.1, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.001, "optim/grad_norm": 1.682326316833496, "optim/lr": 0.002951263385867391, "optim/total_tokens": 2484076544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467734, "loss/out": 4.613071441650391, "created_at": "2025-01-15T02:21:16.766607+00:00"} {"global_step": 9477, "acc_step": 0, "speed/wps": 12953.400555649687, "speed/FLOPS": 203450955564131.47, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9490986466407776, "optim/lr": 0.0029512417204986636, "optim/total_tokens": 2484338688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434807, "loss/out": 4.559239387512207, "created_at": "2025-01-15T02:21:21.834176+00:00"} {"global_step": 9478, "acc_step": 0, "speed/wps": 12965.745875184572, "speed/FLOPS": 203644855771673.6, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1760097742080688, "optim/lr": 0.0029512200503950046, "optim/total_tokens": 2484600832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493719, "loss/out": 4.695979118347168, "created_at": "2025-01-15T02:21:26.898527+00:00"} {"global_step": 9479, "acc_step": 0, "speed/wps": 12956.639452355981, "speed/FLOPS": 203501826887614.3, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0430110692977905, "optim/lr": 0.0029511983755564856, "optim/total_tokens": 2484862976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445407, "loss/out": 4.72304105758667, "created_at": "2025-01-15T02:21:31.964462+00:00"} {"global_step": 9480, "acc_step": 0, "speed/wps": 12958.762717062116, "speed/FLOPS": 203535175677495.62, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2377073764801025, "optim/lr": 0.0029511766959831767, "optim/total_tokens": 2485125120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348373, "loss/out": 4.573268413543701, "created_at": "2025-01-15T02:21:37.029585+00:00"} {"global_step": 9481, "acc_step": 0, "speed/wps": 12968.566697145787, "speed/FLOPS": 203689160656790.06, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0260463953018188, "optim/lr": 0.0029511550116751483, "optim/total_tokens": 2485387264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455003, "loss/out": 4.608864784240723, "created_at": "2025-01-15T02:21:42.091654+00:00"} {"global_step": 9482, "acc_step": 0, "speed/wps": 12961.900836831679, "speed/FLOPS": 203584464160705.03, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.993635654449463, "optim/lr": 0.002951133322632472, "optim/total_tokens": 2485649408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 4.528425693511963, "created_at": "2025-01-15T02:21:47.153609+00:00"} {"global_step": 9483, "acc_step": 0, "speed/wps": 12960.331177616072, "speed/FLOPS": 203559810505786.12, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1754392385482788, "optim/lr": 0.002951111628855218, "optim/total_tokens": 2485911552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435852, "loss/out": 4.668733596801758, "created_at": "2025-01-15T02:21:52.216570+00:00"} {"global_step": 9484, "acc_step": 0, "speed/wps": 12958.27971128638, "speed/FLOPS": 203527589408068.16, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4941455125808716, "optim/lr": 0.0029510899303434577, "optim/total_tokens": 2486173696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 4.651408672332764, "created_at": "2025-01-15T02:21:57.280298+00:00"} {"global_step": 9485, "acc_step": 0, "speed/wps": 12963.666335535829, "speed/FLOPS": 203612193743903.16, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6957229375839233, "optim/lr": 0.0029510682270972605, "optim/total_tokens": 2486435840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446220, "loss/out": 4.414008140563965, "created_at": "2025-01-15T02:22:02.342134+00:00"} {"global_step": 9486, "acc_step": 0, "speed/wps": 12961.451832508405, "speed/FLOPS": 203577411930811.44, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0245939493179321, "optim/lr": 0.002951046519116699, "optim/total_tokens": 2486697984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450746, "loss/out": 4.51541805267334, "created_at": "2025-01-15T02:22:07.405722+00:00"} {"global_step": 9487, "acc_step": 0, "speed/wps": 12956.61155712502, "speed/FLOPS": 203501388754679.97, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1665056943893433, "optim/lr": 0.002951024806401843, "optim/total_tokens": 2486960128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443306, "loss/out": 4.568462371826172, "created_at": "2025-01-15T02:22:12.467934+00:00"} {"global_step": 9488, "acc_step": 0, "speed/wps": 12960.375652350542, "speed/FLOPS": 203560509042605.7, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.001, "optim/grad_norm": 1.1492563486099243, "optim/lr": 0.0029510030889527633, "optim/total_tokens": 2487222272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324468, "loss/out": 4.494967460632324, "created_at": "2025-01-15T02:22:17.530166+00:00"} {"global_step": 9489, "acc_step": 0, "speed/wps": 12959.184781501042, "speed/FLOPS": 203541804779486.25, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2848857641220093, "optim/lr": 0.002950981366769531, "optim/total_tokens": 2487484416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354800, "loss/out": 4.432643413543701, "created_at": "2025-01-15T02:22:22.595501+00:00"} {"global_step": 9490, "acc_step": 0, "speed/wps": 12963.615698123382, "speed/FLOPS": 203611398413757.6, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2059178352355957, "optim/lr": 0.0029509596398522167, "optim/total_tokens": 2487746560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452615, "loss/out": 4.600771903991699, "created_at": "2025-01-15T02:22:27.655681+00:00"} {"global_step": 9491, "acc_step": 0, "speed/wps": 12960.527719006048, "speed/FLOPS": 203562897458391.88, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4351907968521118, "optim/lr": 0.0029509379082008924, "optim/total_tokens": 2488008704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 4.560324192047119, "created_at": "2025-01-15T02:22:32.717263+00:00"} {"global_step": 9492, "acc_step": 0, "speed/wps": 12958.249778307312, "speed/FLOPS": 203527119269499.7, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.8856631517410278, "optim/lr": 0.0029509161718156277, "optim/total_tokens": 2488270848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438240, "loss/out": 4.622900009155273, "created_at": "2025-01-15T02:22:37.779526+00:00"} {"global_step": 9493, "acc_step": 0, "speed/wps": 12961.83071881702, "speed/FLOPS": 203583362860932.6, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7416103482246399, "optim/lr": 0.002950894430696494, "optim/total_tokens": 2488532992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 4.636167526245117, "created_at": "2025-01-15T02:22:42.844092+00:00"} {"global_step": 9494, "acc_step": 0, "speed/wps": 12959.91804884552, "speed/FLOPS": 203553321750746.22, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.550013303756714, "optim/lr": 0.002950872684843562, "optim/total_tokens": 2488795136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420140, "loss/out": 4.500711917877197, "created_at": "2025-01-15T02:22:47.909842+00:00"} {"global_step": 9495, "acc_step": 0, "speed/wps": 12958.134829698738, "speed/FLOPS": 203525313843646.1, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0846697092056274, "optim/lr": 0.002950850934256903, "optim/total_tokens": 2489057280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316527, "loss/out": 4.54401159286499, "created_at": "2025-01-15T02:22:52.974431+00:00"} {"global_step": 9496, "acc_step": 0, "speed/wps": 12960.789557480284, "speed/FLOPS": 203567009991433.53, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0951045751571655, "optim/lr": 0.0029508291789365876, "optim/total_tokens": 2489319424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305807, "loss/out": 4.4555511474609375, "created_at": "2025-01-15T02:22:58.038861+00:00"} {"global_step": 9497, "acc_step": 0, "speed/wps": 12962.214705899925, "speed/FLOPS": 203589393905723.16, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.132095217704773, "optim/lr": 0.0029508074188826874, "optim/total_tokens": 2489581568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 406820, "loss/out": 4.4928693771362305, "created_at": "2025-01-15T02:23:03.101801+00:00"} {"global_step": 9498, "acc_step": 0, "speed/wps": 12968.935511458843, "speed/FLOPS": 203694953392380.6, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7590808272361755, "optim/lr": 0.0029507856540952728, "optim/total_tokens": 2489843712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430268, "loss/out": 4.7323808670043945, "created_at": "2025-01-15T02:23:08.163020+00:00"} {"global_step": 9499, "acc_step": 0, "speed/wps": 12955.607716499431, "speed/FLOPS": 203485622058233.56, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.159385323524475, "optim/lr": 0.002950763884574415, "optim/total_tokens": 2490105856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 4.5585527420043945, "created_at": "2025-01-15T02:23:13.225594+00:00"} {"global_step": 9500, "acc_step": 0, "speed/wps": 12954.32085090645, "speed/FLOPS": 203465410065757.16, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2040510177612305, "optim/lr": 0.002950742110320185, "optim/total_tokens": 2490368000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436666, "loss/out": 4.617114543914795, "created_at": "2025-01-15T02:23:18.291843+00:00"} {"global_step": 9501, "acc_step": 0, "speed/wps": 12961.372315441404, "speed/FLOPS": 203576163006006.28, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.851807177066803, "optim/lr": 0.002950720331332654, "optim/total_tokens": 2490630144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 4.490289211273193, "created_at": "2025-01-15T02:23:23.351823+00:00"} {"global_step": 9502, "acc_step": 0, "speed/wps": 12958.1706369786, "speed/FLOPS": 203525876246180.72, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8834295272827148, "optim/lr": 0.0029506985476118923, "optim/total_tokens": 2490892288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451278, "loss/out": 4.751863479614258, "created_at": "2025-01-15T02:23:28.415634+00:00"} {"global_step": 9503, "acc_step": 0, "speed/wps": 12964.097690237479, "speed/FLOPS": 203618968762236.06, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0395106077194214, "optim/lr": 0.0029506767591579723, "optim/total_tokens": 2491154432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314942, "loss/out": 4.538769245147705, "created_at": "2025-01-15T02:23:33.477761+00:00"} {"global_step": 9504, "acc_step": 0, "speed/wps": 12964.263389126449, "speed/FLOPS": 203621571290981.47, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4391425848007202, "optim/lr": 0.0029506549659709643, "optim/total_tokens": 2491416576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467205, "loss/out": 4.609155654907227, "created_at": "2025-01-15T02:23:38.538806+00:00"} {"global_step": 9505, "acc_step": 0, "speed/wps": 12957.441420635789, "speed/FLOPS": 203514422901468.6, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1110910177230835, "optim/lr": 0.0029506331680509393, "optim/total_tokens": 2491678720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 4.57056999206543, "created_at": "2025-01-15T02:23:43.600877+00:00"} {"global_step": 9506, "acc_step": 0, "speed/wps": 12959.20439442563, "speed/FLOPS": 203542112827417.88, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7376468777656555, "optim/lr": 0.002950611365397969, "optim/total_tokens": 2491940864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449411, "loss/out": 4.6405744552612305, "created_at": "2025-01-15T02:23:48.665351+00:00"} {"global_step": 9507, "acc_step": 0, "speed/wps": 12961.306056341014, "speed/FLOPS": 203575122315785.12, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2908648252487183, "optim/lr": 0.0029505895580121235, "optim/total_tokens": 2492203008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471217, "loss/out": 4.475207328796387, "created_at": "2025-01-15T02:23:53.726610+00:00"} {"global_step": 9508, "acc_step": 0, "speed/wps": 12956.485599474257, "speed/FLOPS": 203499410416690.8, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.8198830485343933, "optim/lr": 0.0029505677458934753, "optim/total_tokens": 2492465152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456840, "loss/out": 4.575065612792969, "created_at": "2025-01-15T02:23:58.789854+00:00"} {"global_step": 9509, "acc_step": 0, "speed/wps": 12961.094721218848, "speed/FLOPS": 203571803007286.6, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5247535109519958, "optim/lr": 0.0029505459290420947, "optim/total_tokens": 2492727296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435582, "loss/out": 4.559342384338379, "created_at": "2025-01-15T02:24:03.851249+00:00"} {"global_step": 9510, "acc_step": 0, "speed/wps": 12959.335209234725, "speed/FLOPS": 203544167453753.12, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0815197229385376, "optim/lr": 0.002950524107458053, "optim/total_tokens": 2492989440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430791, "loss/out": 4.705906867980957, "created_at": "2025-01-15T02:24:08.913420+00:00"} {"global_step": 9511, "acc_step": 0, "speed/wps": 12965.039112868562, "speed/FLOPS": 203633755098307.5, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0453294515609741, "optim/lr": 0.002950502281141421, "optim/total_tokens": 2493251584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305274, "loss/out": 4.61131477355957, "created_at": "2025-01-15T02:24:13.977279+00:00"} {"global_step": 9512, "acc_step": 0, "speed/wps": 12963.68532198072, "speed/FLOPS": 203612491952110.28, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6455866694450378, "optim/lr": 0.0029504804500922704, "optim/total_tokens": 2493513728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355086, "loss/out": 4.555489540100098, "created_at": "2025-01-15T02:24:19.036582+00:00"} {"global_step": 9513, "acc_step": 0, "speed/wps": 12962.84266157078, "speed/FLOPS": 203599256812435.44, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0395071506500244, "optim/lr": 0.002950458614310673, "optim/total_tokens": 2493775872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454995, "loss/out": 4.52083683013916, "created_at": "2025-01-15T02:24:24.100008+00:00"} {"global_step": 9514, "acc_step": 0, "speed/wps": 12963.304710501663, "speed/FLOPS": 203606513925816.72, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9592851400375366, "optim/lr": 0.0029504367737966993, "optim/total_tokens": 2494038016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469590, "loss/out": 4.545196533203125, "created_at": "2025-01-15T02:24:29.161509+00:00"} {"global_step": 9515, "acc_step": 0, "speed/wps": 12961.342807449657, "speed/FLOPS": 203575699542447.66, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7368320226669312, "optim/lr": 0.0029504149285504203, "optim/total_tokens": 2494300160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448605, "loss/out": 4.489638805389404, "created_at": "2025-01-15T02:24:34.221772+00:00"} {"global_step": 9516, "acc_step": 0, "speed/wps": 12957.860452187299, "speed/FLOPS": 203521004367794.72, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9385764002799988, "optim/lr": 0.002950393078571908, "optim/total_tokens": 2494562304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456567, "loss/out": 4.58457088470459, "created_at": "2025-01-15T02:24:39.285789+00:00"} {"global_step": 9517, "acc_step": 0, "speed/wps": 12955.912617186475, "speed/FLOPS": 203490410942501.03, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0351933240890503, "optim/lr": 0.002950371223861233, "optim/total_tokens": 2494824448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464820, "loss/out": 4.468402862548828, "created_at": "2025-01-15T02:24:44.347831+00:00"} {"global_step": 9518, "acc_step": 0, "speed/wps": 12959.020598973568, "speed/FLOPS": 203539226067281.84, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.633931577205658, "optim/lr": 0.0029503493644184673, "optim/total_tokens": 2495086592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316449, "loss/out": 4.476053714752197, "created_at": "2025-01-15T02:24:49.409739+00:00"} {"global_step": 9519, "acc_step": 0, "speed/wps": 12954.084913969897, "speed/FLOPS": 203461704351957.25, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.2336196899414062, "optim/lr": 0.002950327500243682, "optim/total_tokens": 2495348736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 404851, "loss/out": 4.458581924438477, "created_at": "2025-01-15T02:24:54.473884+00:00"} {"global_step": 9520, "acc_step": 0, "speed/wps": 12958.780660595256, "speed/FLOPS": 203535457505340.56, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.059691309928894, "optim/lr": 0.002950305631336948, "optim/total_tokens": 2495610880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436096, "loss/out": 4.416315078735352, "created_at": "2025-01-15T02:24:59.538804+00:00"} {"global_step": 9521, "acc_step": 0, "speed/wps": 12963.75972549837, "speed/FLOPS": 203613660561596.12, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7503490447998047, "optim/lr": 0.002950283757698337, "optim/total_tokens": 2495873024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 4.59168004989624, "created_at": "2025-01-15T02:25:04.601955+00:00"} {"global_step": 9522, "acc_step": 0, "speed/wps": 12957.52686632307, "speed/FLOPS": 203515764943402.03, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0637898445129395, "optim/lr": 0.00295026187932792, "optim/total_tokens": 2496135168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483617, "loss/out": 4.6167755126953125, "created_at": "2025-01-15T02:25:09.665629+00:00"} {"global_step": 9523, "acc_step": 0, "speed/wps": 12952.177449511515, "speed/FLOPS": 203431745001507.53, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6632949113845825, "optim/lr": 0.0029502399962257693, "optim/total_tokens": 2496397312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459743, "loss/out": 4.47865104675293, "created_at": "2025-01-15T02:25:14.729385+00:00"} {"global_step": 9524, "acc_step": 0, "speed/wps": 12955.099854860728, "speed/FLOPS": 203477645393322.7, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2127397060394287, "optim/lr": 0.0029502181083919557, "optim/total_tokens": 2496659456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452081, "loss/out": 4.471851348876953, "created_at": "2025-01-15T02:25:19.791712+00:00"} {"global_step": 9525, "acc_step": 0, "speed/wps": 12956.60855361762, "speed/FLOPS": 203501341580469.03, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3286880254745483, "optim/lr": 0.00295019621582655, "optim/total_tokens": 2496921600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 4.499431133270264, "created_at": "2025-01-15T02:25:24.853746+00:00"} {"global_step": 9526, "acc_step": 0, "speed/wps": 12965.278664356705, "speed/FLOPS": 203637517583607.0, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2216196060180664, "optim/lr": 0.002950174318529625, "optim/total_tokens": 2497183744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328746, "loss/out": 4.56416130065918, "created_at": "2025-01-15T02:25:29.916833+00:00"} {"global_step": 9527, "acc_step": 0, "speed/wps": 12961.356678098946, "speed/FLOPS": 203575917400055.0, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.0944724082946777, "optim/lr": 0.002950152416501251, "optim/total_tokens": 2497445888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351876, "loss/out": 4.55027961730957, "created_at": "2025-01-15T02:25:34.979277+00:00"} {"global_step": 9528, "acc_step": 0, "speed/wps": 12959.812949632662, "speed/FLOPS": 203551671023186.8, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1281814575195312, "optim/lr": 0.0029501305097415, "optim/total_tokens": 2497708032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438769, "loss/out": 4.6197357177734375, "created_at": "2025-01-15T02:25:40.045623+00:00"} {"global_step": 9529, "acc_step": 0, "speed/wps": 12963.070495120128, "speed/FLOPS": 203602835251404.06, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0370168685913086, "optim/lr": 0.002950108598250443, "optim/total_tokens": 2497970176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 4.618690490722656, "created_at": "2025-01-15T02:25:45.105938+00:00"} {"global_step": 9530, "acc_step": 0, "speed/wps": 12964.40326534823, "speed/FLOPS": 203623768239254.62, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9231043457984924, "optim/lr": 0.002950086682028152, "optim/total_tokens": 2498232320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450746, "loss/out": 4.55899715423584, "created_at": "2025-01-15T02:25:50.166887+00:00"} {"global_step": 9531, "acc_step": 0, "speed/wps": 12967.002981020874, "speed/FLOPS": 203664600346277.2, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7420960664749146, "optim/lr": 0.002950064761074699, "optim/total_tokens": 2498494464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467734, "loss/out": 4.5535054206848145, "created_at": "2025-01-15T02:25:55.228427+00:00"} {"global_step": 9532, "acc_step": 0, "speed/wps": 12961.709176832816, "speed/FLOPS": 203581453877054.47, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1202850341796875, "optim/lr": 0.002950042835390154, "optim/total_tokens": 2498756608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452081, "loss/out": 4.568624973297119, "created_at": "2025-01-15T02:26:00.288396+00:00"} {"global_step": 9533, "acc_step": 0, "speed/wps": 12964.025410869579, "speed/FLOPS": 203617833515440.94, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4127938747406006, "optim/lr": 0.0029500209049745894, "optim/total_tokens": 2499018752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323215, "loss/out": 4.51334285736084, "created_at": "2025-01-15T02:26:05.347472+00:00"} {"global_step": 9534, "acc_step": 0, "speed/wps": 12964.850345881821, "speed/FLOPS": 203630790253387.7, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2292534112930298, "optim/lr": 0.0029499989698280773, "optim/total_tokens": 2499280896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328606, "loss/out": 4.624020576477051, "created_at": "2025-01-15T02:26:10.407119+00:00"} {"global_step": 9535, "acc_step": 0, "speed/wps": 12957.444682210504, "speed/FLOPS": 203514474128981.34, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8957554697990417, "optim/lr": 0.002949977029950688, "optim/total_tokens": 2499543040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 391766, "loss/out": 4.612977981567383, "created_at": "2025-01-15T02:26:15.468612+00:00"} {"global_step": 9536, "acc_step": 0, "speed/wps": 12957.6946568542, "speed/FLOPS": 203518400324261.78, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.5426701307296753, "optim/lr": 0.0029499550853424946, "optim/total_tokens": 2499805184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380554, "loss/out": 4.716685771942139, "created_at": "2025-01-15T02:26:20.529838+00:00"} {"global_step": 9537, "acc_step": 0, "speed/wps": 12959.197750176823, "speed/FLOPS": 203542008470360.22, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7821400165557861, "optim/lr": 0.002949933136003567, "optim/total_tokens": 2500067328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435567, "loss/out": 4.4217424392700195, "created_at": "2025-01-15T02:26:25.590983+00:00"} {"global_step": 9538, "acc_step": 0, "speed/wps": 12955.451400352236, "speed/FLOPS": 203483166898340.97, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7587158679962158, "optim/lr": 0.002949911181933978, "optim/total_tokens": 2500329472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 4.414942264556885, "created_at": "2025-01-15T02:26:30.653645+00:00"} {"global_step": 9539, "acc_step": 0, "speed/wps": 12966.041660993162, "speed/FLOPS": 203649501494249.8, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6243428587913513, "optim/lr": 0.002949889223133799, "optim/total_tokens": 2500591616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453131, "loss/out": 4.532288074493408, "created_at": "2025-01-15T02:26:35.716677+00:00"} {"global_step": 9540, "acc_step": 0, "speed/wps": 12962.843251388304, "speed/FLOPS": 203599266076330.16, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7323442101478577, "optim/lr": 0.002949867259603102, "optim/total_tokens": 2500853760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 4.546605110168457, "created_at": "2025-01-15T02:26:40.779039+00:00"} {"global_step": 9541, "acc_step": 0, "speed/wps": 12962.606820052104, "speed/FLOPS": 203595552597304.44, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5387595891952515, "optim/lr": 0.0029498452913419575, "optim/total_tokens": 2501115904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 4.500723361968994, "created_at": "2025-01-15T02:26:45.840690+00:00"} {"global_step": 9542, "acc_step": 0, "speed/wps": 12952.109079211998, "speed/FLOPS": 203430671152002.97, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9999486207962036, "optim/lr": 0.0029498233183504383, "optim/total_tokens": 2501378048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290611, "loss/out": 4.539876937866211, "created_at": "2025-01-15T02:26:50.904424+00:00"} {"global_step": 9543, "acc_step": 0, "speed/wps": 12958.603757044895, "speed/FLOPS": 203532678991991.2, "speed/curr_iter_time": 1.277, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9761097431182861, "optim/lr": 0.0029498013406286158, "optim/total_tokens": 2501640192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357486, "loss/out": 4.562943458557129, "created_at": "2025-01-15T02:26:55.965278+00:00"} {"global_step": 9544, "acc_step": 0, "speed/wps": 12958.27016154382, "speed/FLOPS": 203527439416238.56, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8716797232627869, "optim/lr": 0.0029497793581765614, "optim/total_tokens": 2501902336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455003, "loss/out": 4.374446868896484, "created_at": "2025-01-15T02:27:01.032433+00:00"} {"global_step": 9545, "acc_step": 0, "speed/wps": 12960.515652366896, "speed/FLOPS": 203562707935242.7, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7783757448196411, "optim/lr": 0.0029497573709943476, "optim/total_tokens": 2502164480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443835, "loss/out": 4.539347171783447, "created_at": "2025-01-15T02:27:06.094143+00:00"} {"global_step": 9546, "acc_step": 0, "speed/wps": 12968.256963249978, "speed/FLOPS": 203684295860337.4, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2681233882904053, "optim/lr": 0.0029497353790820446, "optim/total_tokens": 2502426624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462150, "loss/out": 4.478237152099609, "created_at": "2025-01-15T02:27:11.151830+00:00"} {"global_step": 9547, "acc_step": 0, "speed/wps": 12964.751386641847, "speed/FLOPS": 203629235962539.66, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1342849731445312, "optim/lr": 0.002949713382439726, "optim/total_tokens": 2502688768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 4.423510551452637, "created_at": "2025-01-15T02:27:16.215777+00:00"} {"global_step": 9548, "acc_step": 0, "speed/wps": 12958.913853327203, "speed/FLOPS": 203537549480221.8, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.165531873703003, "optim/lr": 0.002949691381067462, "optim/total_tokens": 2502950912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426252, "loss/out": 4.598239898681641, "created_at": "2025-01-15T02:27:21.281283+00:00"} {"global_step": 9549, "acc_step": 0, "speed/wps": 12959.289610384518, "speed/FLOPS": 203543451261151.22, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.9514718055725098, "optim/lr": 0.002949669374965325, "optim/total_tokens": 2503213056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329272, "loss/out": 4.533756256103516, "created_at": "2025-01-15T02:27:26.342337+00:00"} {"global_step": 9550, "acc_step": 0, "speed/wps": 12965.240299978263, "speed/FLOPS": 203636915018325.38, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1164757013320923, "optim/lr": 0.002949647364133387, "optim/total_tokens": 2503475200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291617, "loss/out": 4.541693210601807, "created_at": "2025-01-15T02:27:31.402191+00:00"} {"global_step": 9551, "acc_step": 0, "speed/wps": 12953.251579458998, "speed/FLOPS": 203448615688341.97, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.148197054862976, "optim/lr": 0.00294962534857172, "optim/total_tokens": 2503737344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463770, "loss/out": 4.645519256591797, "created_at": "2025-01-15T02:27:36.465139+00:00"} {"global_step": 9552, "acc_step": 0, "speed/wps": 12961.18622396395, "speed/FLOPS": 203573240183636.38, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.001, "optim/grad_norm": 0.6945860981941223, "optim/lr": 0.0029496033282803958, "optim/total_tokens": 2503999488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461622, "loss/out": 4.5044450759887695, "created_at": "2025-01-15T02:27:41.525009+00:00"} {"global_step": 9553, "acc_step": 0, "speed/wps": 12957.4240438905, "speed/FLOPS": 203514149975805.66, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0795890092849731, "optim/lr": 0.002949581303259485, "optim/total_tokens": 2504261632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462672, "loss/out": 4.524283409118652, "created_at": "2025-01-15T02:27:46.589210+00:00"} {"global_step": 9554, "acc_step": 0, "speed/wps": 12958.473291086408, "speed/FLOPS": 203530629844834.12, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8623353242874146, "optim/lr": 0.0029495592735090614, "optim/total_tokens": 2504523776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469590, "loss/out": 4.611324310302734, "created_at": "2025-01-15T02:27:51.658510+00:00"} {"global_step": 9555, "acc_step": 0, "speed/wps": 12962.397854148423, "speed/FLOPS": 203592270500637.88, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2935009002685547, "optim/lr": 0.002949537239029195, "optim/total_tokens": 2504785920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445407, "loss/out": 4.441266059875488, "created_at": "2025-01-15T02:27:56.722431+00:00"} {"global_step": 9556, "acc_step": 0, "speed/wps": 12968.36700567303, "speed/FLOPS": 203686024227805.56, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6079104542732239, "optim/lr": 0.0029495151998199594, "optim/total_tokens": 2505048064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316287, "loss/out": 4.536630153656006, "created_at": "2025-01-15T02:28:01.782857+00:00"} {"global_step": 9557, "acc_step": 0, "speed/wps": 12952.303707242703, "speed/FLOPS": 203433728052672.12, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6711848378181458, "optim/lr": 0.002949493155881425, "optim/total_tokens": 2505310208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 410019, "loss/out": 4.385224342346191, "created_at": "2025-01-15T02:28:06.849505+00:00"} {"global_step": 9558, "acc_step": 0, "speed/wps": 12966.719951106901, "speed/FLOPS": 203660154972554.38, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9266250729560852, "optim/lr": 0.0029494711072136645, "optim/total_tokens": 2505572352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440628, "loss/out": 4.446525573730469, "created_at": "2025-01-15T02:28:11.913132+00:00"} {"global_step": 9559, "acc_step": 0, "speed/wps": 12956.07723145328, "speed/FLOPS": 203492996435764.9, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.4205129146575928, "optim/lr": 0.0029494490538167504, "optim/total_tokens": 2505834496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 4.551553726196289, "created_at": "2025-01-15T02:28:16.975323+00:00"} {"global_step": 9560, "acc_step": 0, "speed/wps": 12955.02067568155, "speed/FLOPS": 203476401775510.1, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7316729426383972, "optim/lr": 0.002949426995690753, "optim/total_tokens": 2506096640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469353, "loss/out": 4.42625093460083, "created_at": "2025-01-15T02:28:22.043298+00:00"} {"global_step": 9561, "acc_step": 0, "speed/wps": 12964.164543936718, "speed/FLOPS": 203620018791449.22, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7285561561584473, "optim/lr": 0.0029494049328357462, "optim/total_tokens": 2506358784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 4.535724639892578, "created_at": "2025-01-15T02:28:27.117680+00:00"} {"global_step": 9562, "acc_step": 0, "speed/wps": 12963.764556956168, "speed/FLOPS": 203613736446279.97, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.5702023506164551, "optim/lr": 0.0029493828652518007, "optim/total_tokens": 2506620928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441971, "loss/out": 4.4118146896362305, "created_at": "2025-01-15T02:28:32.182212+00:00"} {"global_step": 9563, "acc_step": 0, "speed/wps": 12958.302814944662, "speed/FLOPS": 203527952282769.25, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9922166466712952, "optim/lr": 0.002949360792938989, "optim/total_tokens": 2506883072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305807, "loss/out": 4.470950126647949, "created_at": "2025-01-15T02:28:37.248863+00:00"} {"global_step": 9564, "acc_step": 0, "speed/wps": 12956.402321673111, "speed/FLOPS": 203498102424388.12, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2312923669815063, "optim/lr": 0.0029493387158973827, "optim/total_tokens": 2507145216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462458, "loss/out": 4.500064849853516, "created_at": "2025-01-15T02:28:42.310749+00:00"} {"global_step": 9565, "acc_step": 0, "speed/wps": 12955.507159476527, "speed/FLOPS": 203484042672009.0, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6903807520866394, "optim/lr": 0.002949316634127054, "optim/total_tokens": 2507407360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472260, "loss/out": 4.443303108215332, "created_at": "2025-01-15T02:28:47.373140+00:00"} {"global_step": 9566, "acc_step": 0, "speed/wps": 12949.666252218853, "speed/FLOPS": 203392303197279.56, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.2492828369140625, "optim/lr": 0.0029492945476280756, "optim/total_tokens": 2507669504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446505, "loss/out": 4.588527202606201, "created_at": "2025-01-15T02:28:52.448588+00:00"} {"global_step": 9567, "acc_step": 0, "speed/wps": 12953.47640775772, "speed/FLOPS": 203452146925719.84, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8558663725852966, "optim/lr": 0.0029492724564005183, "optim/total_tokens": 2507931648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 4.478580951690674, "created_at": "2025-01-15T02:28:57.511672+00:00"} {"global_step": 9568, "acc_step": 0, "speed/wps": 12955.499228614024, "speed/FLOPS": 203483918106915.25, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6831350326538086, "optim/lr": 0.002949250360444455, "optim/total_tokens": 2508193792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439583, "loss/out": 4.615732192993164, "created_at": "2025-01-15T02:29:02.573813+00:00"} {"global_step": 9569, "acc_step": 0, "speed/wps": 12962.43307133427, "speed/FLOPS": 203592823634935.1, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.49692174792289734, "optim/lr": 0.0029492282597599583, "optim/total_tokens": 2508455936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289017, "loss/out": 4.445257186889648, "created_at": "2025-01-15T02:29:07.636957+00:00"} {"global_step": 9570, "acc_step": 0, "speed/wps": 12955.395565382078, "speed/FLOPS": 203482289933409.3, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7456792593002319, "optim/lr": 0.002949206154347099, "optim/total_tokens": 2508718080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440919, "loss/out": 4.494997501373291, "created_at": "2025-01-15T02:29:12.703683+00:00"} {"global_step": 9571, "acc_step": 0, "speed/wps": 12954.61632553448, "speed/FLOPS": 203470050900815.03, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.334844708442688, "optim/lr": 0.00294918404420595, "optim/total_tokens": 2508980224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461605, "loss/out": 4.5748066902160645, "created_at": "2025-01-15T02:29:17.770281+00:00"} {"global_step": 9572, "acc_step": 0, "speed/wps": 12957.855935320753, "speed/FLOPS": 203520933424198.8, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9904122948646545, "optim/lr": 0.0029491619293365834, "optim/total_tokens": 2509242368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463485, "loss/out": 4.546454429626465, "created_at": "2025-01-15T02:29:22.832787+00:00"} {"global_step": 9573, "acc_step": 0, "speed/wps": 12952.784782244957, "speed/FLOPS": 203441283996648.84, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1027897596359253, "optim/lr": 0.0029491398097390717, "optim/total_tokens": 2509504512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455517, "loss/out": 4.654351234436035, "created_at": "2025-01-15T02:29:27.897006+00:00"} {"global_step": 9574, "acc_step": 0, "speed/wps": 12955.742845289647, "speed/FLOPS": 203487744441570.3, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0963231325149536, "optim/lr": 0.002949117685413486, "optim/total_tokens": 2509766656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322066, "loss/out": 4.538550853729248, "created_at": "2025-01-15T02:29:32.959345+00:00"} {"global_step": 9575, "acc_step": 0, "speed/wps": 12958.97911189223, "speed/FLOPS": 203538574455660.2, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5152297616004944, "optim/lr": 0.0029490955563599, "optim/total_tokens": 2510028800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 286322, "loss/out": 4.493915557861328, "created_at": "2025-01-15T02:29:38.023483+00:00"} {"global_step": 9576, "acc_step": 0, "speed/wps": 12955.017799318959, "speed/FLOPS": 203476356598283.1, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.076980710029602, "optim/lr": 0.0029490734225783846, "optim/total_tokens": 2510290944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350820, "loss/out": 4.551323890686035, "created_at": "2025-01-15T02:29:43.086035+00:00"} {"global_step": 9577, "acc_step": 0, "speed/wps": 12947.74520129949, "speed/FLOPS": 203362130452790.4, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9668715596199036, "optim/lr": 0.0029490512840690125, "optim/total_tokens": 2510553088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423342, "loss/out": 4.444934844970703, "created_at": "2025-01-15T02:29:48.151656+00:00"} {"global_step": 9578, "acc_step": 0, "speed/wps": 12957.723095993986, "speed/FLOPS": 203518847000031.56, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.308972716331482, "optim/lr": 0.002949029140831856, "optim/total_tokens": 2510815232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 4.484354496002197, "created_at": "2025-01-15T02:29:53.213156+00:00"} {"global_step": 9579, "acc_step": 0, "speed/wps": 12963.092846923208, "speed/FLOPS": 203603186317186.03, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0028340816497803, "optim/lr": 0.0029490069928669867, "optim/total_tokens": 2511077376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436679, "loss/out": 4.44818115234375, "created_at": "2025-01-15T02:29:58.276585+00:00"} {"global_step": 9580, "acc_step": 0, "speed/wps": 12957.719605219456, "speed/FLOPS": 203518792172620.75, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.708871603012085, "optim/lr": 0.002948984840174478, "optim/total_tokens": 2511339520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 4.5944671630859375, "created_at": "2025-01-15T02:30:03.339262+00:00"} {"global_step": 9581, "acc_step": 0, "speed/wps": 12958.113246637835, "speed/FLOPS": 203524974852016.38, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.785781979560852, "optim/lr": 0.0029489626827544015, "optim/total_tokens": 2511601664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 4.399807453155518, "created_at": "2025-01-15T02:30:08.401503+00:00"} {"global_step": 9582, "acc_step": 0, "speed/wps": 12948.565940301607, "speed/FLOPS": 203375021286632.22, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.036065697669983, "optim/lr": 0.0029489405206068295, "optim/total_tokens": 2511863808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336633, "loss/out": 4.463881015777588, "created_at": "2025-01-15T02:30:13.469551+00:00"} {"global_step": 9583, "acc_step": 0, "speed/wps": 12952.706716169117, "speed/FLOPS": 203440057861653.78, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.723157525062561, "optim/lr": 0.002948918353731834, "optim/total_tokens": 2512125952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 4.5152668952941895, "created_at": "2025-01-15T02:30:18.533134+00:00"} {"global_step": 9584, "acc_step": 0, "speed/wps": 12959.043712433402, "speed/FLOPS": 203539589095929.75, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.943492889404297, "optim/lr": 0.0029488961821294885, "optim/total_tokens": 2512388096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481190, "loss/out": 4.410412311553955, "created_at": "2025-01-15T02:30:23.603735+00:00"} {"global_step": 9585, "acc_step": 0, "speed/wps": 12960.634295498616, "speed/FLOPS": 203564571388659.12, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3748383522033691, "optim/lr": 0.002948874005799864, "optim/total_tokens": 2512650240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467490, "loss/out": 4.572662830352783, "created_at": "2025-01-15T02:30:28.668319+00:00"} {"global_step": 9586, "acc_step": 0, "speed/wps": 12959.757962560301, "speed/FLOPS": 203550807375655.22, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1739169359207153, "optim/lr": 0.0029488518247430336, "optim/total_tokens": 2512912384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450994, "loss/out": 4.529233932495117, "created_at": "2025-01-15T02:30:33.730054+00:00"} {"global_step": 9587, "acc_step": 0, "speed/wps": 12954.87571909098, "speed/FLOPS": 203474125033063.97, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6105127930641174, "optim/lr": 0.0029488296389590693, "optim/total_tokens": 2513174528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292963, "loss/out": 4.565360069274902, "created_at": "2025-01-15T02:30:38.793291+00:00"} {"global_step": 9588, "acc_step": 0, "speed/wps": 12958.705579208041, "speed/FLOPS": 203534278248981.72, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1645394563674927, "optim/lr": 0.0029488074484480434, "optim/total_tokens": 2513436672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 347724, "loss/out": 4.640460968017578, "created_at": "2025-01-15T02:30:43.858894+00:00"} {"global_step": 9589, "acc_step": 0, "speed/wps": 12957.360397157177, "speed/FLOPS": 203513150316399.06, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2139540910720825, "optim/lr": 0.0029487852532100287, "optim/total_tokens": 2513698816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 394230, "loss/out": 4.507603645324707, "created_at": "2025-01-15T02:30:48.921986+00:00"} {"global_step": 9590, "acc_step": 0, "speed/wps": 12955.281058568326, "speed/FLOPS": 203480491446550.5, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5581376552581787, "optim/lr": 0.0029487630532450976, "optim/total_tokens": 2513960960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428403, "loss/out": 4.589140892028809, "created_at": "2025-01-15T02:30:53.986279+00:00"} {"global_step": 9591, "acc_step": 0, "speed/wps": 12956.416341582624, "speed/FLOPS": 203498322626332.4, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.767896294593811, "optim/lr": 0.0029487408485533227, "optim/total_tokens": 2514223104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493719, "loss/out": 4.490833282470703, "created_at": "2025-01-15T02:30:59.049741+00:00"} {"global_step": 9592, "acc_step": 0, "speed/wps": 12953.090451351545, "speed/FLOPS": 203446084950001.2, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8918343186378479, "optim/lr": 0.002948718639134776, "optim/total_tokens": 2514485248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431605, "loss/out": 4.399921894073486, "created_at": "2025-01-15T02:31:04.114135+00:00"} {"global_step": 9593, "acc_step": 0, "speed/wps": 12957.533418191808, "speed/FLOPS": 203515867849503.75, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.31164273619651794, "optim/lr": 0.00294869642498953, "optim/total_tokens": 2514747392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 4.456615447998047, "created_at": "2025-01-15T02:31:09.176230+00:00"} {"global_step": 9594, "acc_step": 0, "speed/wps": 12952.291891932193, "speed/FLOPS": 203433542476984.62, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.06088125705719, "optim/lr": 0.0029486742061176573, "optim/total_tokens": 2515009536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441157, "loss/out": 4.497986793518066, "created_at": "2025-01-15T02:31:14.239967+00:00"} {"global_step": 9595, "acc_step": 0, "speed/wps": 12956.50182384835, "speed/FLOPS": 203499665242780.56, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5803365111351013, "optim/lr": 0.0029486519825192305, "optim/total_tokens": 2515271680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325620, "loss/out": 4.605992794036865, "created_at": "2025-01-15T02:31:19.304830+00:00"} {"global_step": 9596, "acc_step": 0, "speed/wps": 12954.214442608418, "speed/FLOPS": 203463738777216.22, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5754032731056213, "optim/lr": 0.0029486297541943215, "optim/total_tokens": 2515533824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442778, "loss/out": 4.349914073944092, "created_at": "2025-01-15T02:31:24.368577+00:00"} {"global_step": 9597, "acc_step": 0, "speed/wps": 12956.076880794613, "speed/FLOPS": 203492990928188.7, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1180847883224487, "optim/lr": 0.002948607521143004, "optim/total_tokens": 2515795968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 4.493287086486816, "created_at": "2025-01-15T02:31:29.436594+00:00"} {"global_step": 9598, "acc_step": 0, "speed/wps": 12961.414264291032, "speed/FLOPS": 203576821870333.44, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7676474452018738, "optim/lr": 0.0029485852833653495, "optim/total_tokens": 2516058112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 4.467820167541504, "created_at": "2025-01-15T02:31:34.499848+00:00"} {"global_step": 9599, "acc_step": 0, "speed/wps": 12955.150640425396, "speed/FLOPS": 203478443050402.47, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4994644522666931, "optim/lr": 0.002948563040861431, "optim/total_tokens": 2516320256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453945, "loss/out": 4.571865558624268, "created_at": "2025-01-15T02:31:39.562102+00:00"} {"global_step": 9600, "acc_step": 0, "speed/wps": 12954.83340710804, "speed/FLOPS": 203473460465229.16, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9192550182342529, "optim/lr": 0.0029485407936313207, "optim/total_tokens": 2516582400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315474, "loss/out": 4.503850936889648, "created_at": "2025-01-15T02:31:44.627327+00:00"} {"global_step": 9601, "acc_step": 0, "speed/wps": 12956.231597058992, "speed/FLOPS": 203495420959723.2, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4646878242492676, "optim/lr": 0.0029485185416750917, "optim/total_tokens": 2516844544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361473, "loss/out": 4.629238128662109, "created_at": "2025-01-15T02:31:49.692714+00:00"} {"global_step": 9602, "acc_step": 0, "speed/wps": 12946.31431505518, "speed/FLOPS": 203339656418079.75, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.090203881263733, "optim/lr": 0.0029484962849928163, "optim/total_tokens": 2517106688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441157, "loss/out": 4.540429592132568, "created_at": "2025-01-15T02:31:54.758950+00:00"} {"global_step": 9603, "acc_step": 0, "speed/wps": 12959.691925225865, "speed/FLOPS": 203549770168573.1, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2323168516159058, "optim/lr": 0.002948474023584567, "optim/total_tokens": 2517368832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461622, "loss/out": 4.481071949005127, "created_at": "2025-01-15T02:31:59.825179+00:00"} {"global_step": 9604, "acc_step": 0, "speed/wps": 12957.361399237652, "speed/FLOPS": 203513166055449.94, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.879762589931488, "optim/lr": 0.002948451757450417, "optim/total_tokens": 2517630976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 4.579875946044922, "created_at": "2025-01-15T02:32:04.897050+00:00"} {"global_step": 9605, "acc_step": 0, "speed/wps": 12962.037436390116, "speed/FLOPS": 203586609644477.75, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1305934190750122, "optim/lr": 0.002948429486590438, "optim/total_tokens": 2517893120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 4.4858903884887695, "created_at": "2025-01-15T02:32:09.960842+00:00"} {"global_step": 9606, "acc_step": 0, "speed/wps": 12964.245024371538, "speed/FLOPS": 203621282847269.78, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7574778199195862, "optim/lr": 0.0029484072110047043, "optim/total_tokens": 2518155264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433179, "loss/out": 4.530972480773926, "created_at": "2025-01-15T02:32:15.025253+00:00"} {"global_step": 9607, "acc_step": 0, "speed/wps": 12956.018347302846, "speed/FLOPS": 203492071579267.03, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8154858350753784, "optim/lr": 0.0029483849306932864, "optim/total_tokens": 2518417408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361998, "loss/out": 4.539583206176758, "created_at": "2025-01-15T02:32:20.088513+00:00"} {"global_step": 9608, "acc_step": 0, "speed/wps": 12950.038429812417, "speed/FLOPS": 203398148757812.1, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0026, "optim/grad_norm": 1.5796390771865845, "optim/lr": 0.0029483626456562584, "optim/total_tokens": 2518679552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 4.492231845855713, "created_at": "2025-01-15T02:32:25.153781+00:00"} {"global_step": 9609, "acc_step": 0, "speed/wps": 12956.83640903468, "speed/FLOPS": 203504920362900.12, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.049389123916626, "optim/lr": 0.0029483403558936932, "optim/total_tokens": 2518941696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451559, "loss/out": 4.562046051025391, "created_at": "2025-01-15T02:32:30.218979+00:00"} {"global_step": 9610, "acc_step": 0, "speed/wps": 12965.360238386234, "speed/FLOPS": 203638798815835.22, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0080348253250122, "optim/lr": 0.0029483180614056624, "optim/total_tokens": 2519203840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463485, "loss/out": 4.419437408447266, "created_at": "2025-01-15T02:32:35.277337+00:00"} {"global_step": 9611, "acc_step": 0, "speed/wps": 12961.03907393421, "speed/FLOPS": 203570928990213.97, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4661650061607361, "optim/lr": 0.0029482957621922395, "optim/total_tokens": 2519465984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452847, "loss/out": 4.391522407531738, "created_at": "2025-01-15T02:32:40.339410+00:00"} {"global_step": 9612, "acc_step": 0, "speed/wps": 12960.501688035094, "speed/FLOPS": 203562488606222.75, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0383301973342896, "optim/lr": 0.0029482734582534976, "optim/total_tokens": 2519728128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322976, "loss/out": 4.548273086547852, "created_at": "2025-01-15T02:32:45.400676+00:00"} {"global_step": 9613, "acc_step": 0, "speed/wps": 12962.400122521974, "speed/FLOPS": 203592306128561.56, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5207656621932983, "optim/lr": 0.002948251149589508, "optim/total_tokens": 2519990272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367469, "loss/out": 4.528905868530273, "created_at": "2025-01-15T02:32:50.462924+00:00"} {"global_step": 9614, "acc_step": 0, "speed/wps": 12959.195518722665, "speed/FLOPS": 203541973422306.34, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6351327300071716, "optim/lr": 0.0029482288362003457, "optim/total_tokens": 2520252416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445166, "loss/out": 4.5899763107299805, "created_at": "2025-01-15T02:32:55.523531+00:00"} {"global_step": 9615, "acc_step": 0, "speed/wps": 12960.512654210386, "speed/FLOPS": 203562660845074.8, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.0286887884140015, "optim/lr": 0.0029482065180860815, "optim/total_tokens": 2520514560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433179, "loss/out": 4.506818771362305, "created_at": "2025-01-15T02:33:00.583763+00:00"} {"global_step": 9616, "acc_step": 0, "speed/wps": 12951.812600912494, "speed/FLOPS": 203426014552905.38, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.46636179089546204, "optim/lr": 0.002948184195246789, "optim/total_tokens": 2520776704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457665, "loss/out": 4.545133590698242, "created_at": "2025-01-15T02:33:05.650887+00:00"} {"global_step": 9617, "acc_step": 0, "speed/wps": 12966.150616988982, "speed/FLOPS": 203651212797885.38, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9509137272834778, "optim/lr": 0.002948161867682541, "optim/total_tokens": 2521038848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 4.48890495300293, "created_at": "2025-01-15T02:33:10.712943+00:00"} {"global_step": 9618, "acc_step": 0, "speed/wps": 12958.857093586983, "speed/FLOPS": 203536657990505.56, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0249319076538086, "optim/lr": 0.002948139535393411, "optim/total_tokens": 2521300992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424678, "loss/out": 4.417885780334473, "created_at": "2025-01-15T02:33:15.777627+00:00"} {"global_step": 9619, "acc_step": 0, "speed/wps": 12958.395309149104, "speed/FLOPS": 203529405031350.38, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0431849956512451, "optim/lr": 0.0029481171983794705, "optim/total_tokens": 2521563136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 4.483330726623535, "created_at": "2025-01-15T02:33:20.843045+00:00"} {"global_step": 9620, "acc_step": 0, "speed/wps": 12955.929095361107, "speed/FLOPS": 203490669754877.0, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7611943483352661, "optim/lr": 0.0029480948566407935, "optim/total_tokens": 2521825280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309845, "loss/out": 4.424050807952881, "created_at": "2025-01-15T02:33:25.905666+00:00"} {"global_step": 9621, "acc_step": 0, "speed/wps": 12965.232053816477, "speed/FLOPS": 203636785501023.12, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.20786714553833, "optim/lr": 0.0029480725101774523, "optim/total_tokens": 2522087424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362577, "loss/out": 4.537321090698242, "created_at": "2025-01-15T02:33:30.969562+00:00"} {"global_step": 9622, "acc_step": 0, "speed/wps": 12957.836276369862, "speed/FLOPS": 203520624653360.84, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.4294944107532501, "optim/lr": 0.0029480501589895206, "optim/total_tokens": 2522349568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 4.605838775634766, "created_at": "2025-01-15T02:33:36.030718+00:00"} {"global_step": 9623, "acc_step": 0, "speed/wps": 12964.141811069907, "speed/FLOPS": 203619661740537.72, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8255149126052856, "optim/lr": 0.00294802780307707, "optim/total_tokens": 2522611712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 4.543432712554932, "created_at": "2025-01-15T02:33:41.092223+00:00"} {"global_step": 9624, "acc_step": 0, "speed/wps": 12962.966536368647, "speed/FLOPS": 203601202436357.75, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6101099252700806, "optim/lr": 0.002948005442440175, "optim/total_tokens": 2522873856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451275, "loss/out": 4.467290878295898, "created_at": "2025-01-15T02:33:46.156158+00:00"} {"global_step": 9625, "acc_step": 0, "speed/wps": 12961.819705047714, "speed/FLOPS": 203583189874551.12, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5997468829154968, "optim/lr": 0.002947983077078907, "optim/total_tokens": 2523136000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 4.425054550170898, "created_at": "2025-01-15T02:33:51.215854+00:00"} {"global_step": 9626, "acc_step": 0, "speed/wps": 12964.416657682088, "speed/FLOPS": 203623978584260.84, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9233887791633606, "optim/lr": 0.00294796070699334, "optim/total_tokens": 2523398144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433993, "loss/out": 4.551741123199463, "created_at": "2025-01-15T02:33:56.301476+00:00"} {"global_step": 9627, "acc_step": 0, "speed/wps": 12959.870577005324, "speed/FLOPS": 203552576140262.78, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6366773247718811, "optim/lr": 0.002947938332183547, "optim/total_tokens": 2523660288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427351, "loss/out": 4.638908863067627, "created_at": "2025-01-15T02:34:01.367204+00:00"} {"global_step": 9628, "acc_step": 0, "speed/wps": 12964.20460967406, "speed/FLOPS": 203620648078910.25, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.41535237431526184, "optim/lr": 0.0029479159526496, "optim/total_tokens": 2523922432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474123, "loss/out": 4.450541973114014, "created_at": "2025-01-15T02:34:06.428992+00:00"} {"global_step": 9629, "acc_step": 0, "speed/wps": 12958.118484311568, "speed/FLOPS": 203525057116879.7, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9070483446121216, "optim/lr": 0.0029478935683915734, "optim/total_tokens": 2524184576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470403, "loss/out": 4.470956325531006, "created_at": "2025-01-15T02:34:11.490701+00:00"} {"global_step": 9630, "acc_step": 0, "speed/wps": 12947.14188806825, "speed/FLOPS": 203352654589454.5, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8600171804428101, "optim/lr": 0.0029478711794095392, "optim/total_tokens": 2524446720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457095, "loss/out": 4.509713649749756, "created_at": "2025-01-15T02:34:16.558233+00:00"} {"global_step": 9631, "acc_step": 0, "speed/wps": 12945.445416398792, "speed/FLOPS": 203326009170689.3, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.001, "optim/grad_norm": 1.043888807296753, "optim/lr": 0.0029478487857035702, "optim/total_tokens": 2524708864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 4.51915979385376, "created_at": "2025-01-15T02:34:21.625474+00:00"} {"global_step": 9632, "acc_step": 0, "speed/wps": 12947.22822230748, "speed/FLOPS": 203354010587317.28, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.795800507068634, "optim/lr": 0.002947826387273741, "optim/total_tokens": 2524971008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450227, "loss/out": 4.493577003479004, "created_at": "2025-01-15T02:34:26.692050+00:00"} {"global_step": 9633, "acc_step": 0, "speed/wps": 12960.16463161334, "speed/FLOPS": 203557194671955.03, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.721433162689209, "optim/lr": 0.0029478039841201236, "optim/total_tokens": 2525233152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289838, "loss/out": 4.450984477996826, "created_at": "2025-01-15T02:34:31.755099+00:00"} {"global_step": 9634, "acc_step": 0, "speed/wps": 12953.708588305077, "speed/FLOPS": 203455793640265.6, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9356093406677246, "optim/lr": 0.002947781576242791, "optim/total_tokens": 2525495296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452615, "loss/out": 4.455358505249023, "created_at": "2025-01-15T02:34:36.832173+00:00"} {"global_step": 9635, "acc_step": 0, "speed/wps": 12948.514779854937, "speed/FLOPS": 203374217741515.44, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0494604110717773, "optim/lr": 0.0029477591636418166, "optim/total_tokens": 2525757440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432656, "loss/out": 4.413180351257324, "created_at": "2025-01-15T02:34:41.899376+00:00"} {"global_step": 9636, "acc_step": 0, "speed/wps": 12965.527515178572, "speed/FLOPS": 203641426127721.28, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2613577842712402, "optim/lr": 0.0029477367463172735, "optim/total_tokens": 2526019584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 4.49678897857666, "created_at": "2025-01-15T02:34:46.957887+00:00"} {"global_step": 9637, "acc_step": 0, "speed/wps": 12953.446096375264, "speed/FLOPS": 203451670843806.2, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2420392036437988, "optim/lr": 0.002947714324269235, "optim/total_tokens": 2526281728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447269, "loss/out": 4.3949408531188965, "created_at": "2025-01-15T02:34:52.028135+00:00"} {"global_step": 9638, "acc_step": 0, "speed/wps": 12966.959703300068, "speed/FLOPS": 203663920610202.1, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8079549670219421, "optim/lr": 0.0029476918974977737, "optim/total_tokens": 2526543872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292485, "loss/out": 4.396736145019531, "created_at": "2025-01-15T02:34:57.089320+00:00"} {"global_step": 9639, "acc_step": 0, "speed/wps": 12963.440876351184, "speed/FLOPS": 203608652597594.94, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2369670867919922, "optim/lr": 0.002947669466002963, "optim/total_tokens": 2526806016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 410549, "loss/out": 4.462978363037109, "created_at": "2025-01-15T02:35:02.148647+00:00"} {"global_step": 9640, "acc_step": 0, "speed/wps": 12959.88498154173, "speed/FLOPS": 203552802383300.97, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.8551300764083862, "optim/lr": 0.002947647029784877, "optim/total_tokens": 2527068160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433464, "loss/out": 4.532615661621094, "created_at": "2025-01-15T02:35:07.213059+00:00"} {"global_step": 9641, "acc_step": 0, "speed/wps": 12957.418172121526, "speed/FLOPS": 203514057751605.28, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2857944965362549, "optim/lr": 0.0029476245888435874, "optim/total_tokens": 2527330304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486233, "loss/out": 4.417996406555176, "created_at": "2025-01-15T02:35:12.279634+00:00"} {"global_step": 9642, "acc_step": 0, "speed/wps": 12956.113754923666, "speed/FLOPS": 203493570087055.8, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9742262363433838, "optim/lr": 0.0029476021431791687, "optim/total_tokens": 2527592448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455808, "loss/out": 4.508174896240234, "created_at": "2025-01-15T02:35:17.350664+00:00"} {"global_step": 9643, "acc_step": 0, "speed/wps": 12957.576642558151, "speed/FLOPS": 203516546747573.7, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6791694164276123, "optim/lr": 0.0029475796927916935, "optim/total_tokens": 2527854592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457665, "loss/out": 4.437092304229736, "created_at": "2025-01-15T02:35:22.412432+00:00"} {"global_step": 9644, "acc_step": 0, "speed/wps": 12957.752432413376, "speed/FLOPS": 203519307768810.56, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8576335906982422, "optim/lr": 0.0029475572376812345, "optim/total_tokens": 2528116736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 4.367664337158203, "created_at": "2025-01-15T02:35:27.474900+00:00"} {"global_step": 9645, "acc_step": 0, "speed/wps": 12964.103704199992, "speed/FLOPS": 203619063219781.38, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0772837400436401, "optim/lr": 0.002947534777847866, "optim/total_tokens": 2528378880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443307, "loss/out": 4.409482002258301, "created_at": "2025-01-15T02:35:32.533787+00:00"} {"global_step": 9646, "acc_step": 0, "speed/wps": 12957.476799384685, "speed/FLOPS": 203514978573335.5, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.626361608505249, "optim/lr": 0.0029475123132916606, "optim/total_tokens": 2528641024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 340118, "loss/out": 4.371796607971191, "created_at": "2025-01-15T02:35:37.596427+00:00"} {"global_step": 9647, "acc_step": 0, "speed/wps": 12955.091905416328, "speed/FLOPS": 203477520536374.72, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8359551429748535, "optim/lr": 0.0029474898440126924, "optim/total_tokens": 2528903168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449696, "loss/out": 4.467934608459473, "created_at": "2025-01-15T02:35:42.660080+00:00"} {"global_step": 9648, "acc_step": 0, "speed/wps": 12964.531722544543, "speed/FLOPS": 203625785836041.4, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4846587181091309, "optim/lr": 0.002947467370011034, "optim/total_tokens": 2529165312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 4.474678993225098, "created_at": "2025-01-15T02:35:47.720768+00:00"} {"global_step": 9649, "acc_step": 0, "speed/wps": 12954.617715928563, "speed/FLOPS": 203470072738864.72, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2486422061920166, "optim/lr": 0.002947444891286759, "optim/total_tokens": 2529427456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434230, "loss/out": 4.545748710632324, "created_at": "2025-01-15T02:35:52.784657+00:00"} {"global_step": 9650, "acc_step": 0, "speed/wps": 12956.982466400981, "speed/FLOPS": 203507214394541.72, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7033544182777405, "optim/lr": 0.00294742240783994, "optim/total_tokens": 2529689600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496913, "loss/out": 4.4657111167907715, "created_at": "2025-01-15T02:35:57.846258+00:00"} {"global_step": 9651, "acc_step": 0, "speed/wps": 12956.709595329274, "speed/FLOPS": 203502928579396.4, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.374356746673584, "optim/lr": 0.0029473999196706516, "optim/total_tokens": 2529951744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 4.529267311096191, "created_at": "2025-01-15T02:36:02.912698+00:00"} {"global_step": 9652, "acc_step": 0, "speed/wps": 12963.676284903662, "speed/FLOPS": 203612350012396.84, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8098872900009155, "optim/lr": 0.0029473774267789665, "optim/total_tokens": 2530213888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293067, "loss/out": 4.534016132354736, "created_at": "2025-01-15T02:36:07.971828+00:00"} {"global_step": 9653, "acc_step": 0, "speed/wps": 12954.095487736546, "speed/FLOPS": 203461870427492.34, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9838635921478271, "optim/lr": 0.002947354929164958, "optim/total_tokens": 2530476032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 389150, "loss/out": 4.561811447143555, "created_at": "2025-01-15T02:36:13.041066+00:00"} {"global_step": 9654, "acc_step": 0, "speed/wps": 12954.547214023569, "speed/FLOPS": 203468965409567.0, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0878338813781738, "optim/lr": 0.0029473324268286995, "optim/total_tokens": 2530738176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450990, "loss/out": 4.473459243774414, "created_at": "2025-01-15T02:36:18.109835+00:00"} {"global_step": 9655, "acc_step": 0, "speed/wps": 12953.320863130968, "speed/FLOPS": 203449703883616.0, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6263697147369385, "optim/lr": 0.002947309919770265, "optim/total_tokens": 2531000320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465083, "loss/out": 4.371230602264404, "created_at": "2025-01-15T02:36:23.174406+00:00"} {"global_step": 9656, "acc_step": 0, "speed/wps": 12957.101195160296, "speed/FLOPS": 203509079192857.16, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.6172397136688232, "optim/lr": 0.002947287407989727, "optim/total_tokens": 2531262464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441971, "loss/out": 4.553781509399414, "created_at": "2025-01-15T02:36:28.238042+00:00"} {"global_step": 9657, "acc_step": 0, "speed/wps": 12952.320386034908, "speed/FLOPS": 203433990016022.78, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.9592998027801514, "optim/lr": 0.00294726489148716, "optim/total_tokens": 2531524608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 4.464776992797852, "created_at": "2025-01-15T02:36:33.301652+00:00"} {"global_step": 9658, "acc_step": 0, "speed/wps": 12961.45104227485, "speed/FLOPS": 203577399519107.56, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8878425359725952, "optim/lr": 0.0029472423702626366, "optim/total_tokens": 2531786752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456312, "loss/out": 4.508382797241211, "created_at": "2025-01-15T02:36:38.362350+00:00"} {"global_step": 9659, "acc_step": 0, "speed/wps": 12959.83411827623, "speed/FLOPS": 203552003505822.8, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.9287524819374084, "optim/lr": 0.002947219844316231, "optim/total_tokens": 2532048896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318319, "loss/out": 4.538400173187256, "created_at": "2025-01-15T02:36:43.423879+00:00"} {"global_step": 9660, "acc_step": 0, "speed/wps": 12959.09462199887, "speed/FLOPS": 203540388700612.5, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.782496929168701, "optim/lr": 0.002947197313648016, "optim/total_tokens": 2532311040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338751, "loss/out": 4.429615020751953, "created_at": "2025-01-15T02:36:48.489472+00:00"} {"global_step": 9661, "acc_step": 0, "speed/wps": 12962.164320982438, "speed/FLOPS": 203588602541357.88, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6940683126449585, "optim/lr": 0.002947174778258065, "optim/total_tokens": 2532573184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436381, "loss/out": 4.529855728149414, "created_at": "2025-01-15T02:36:53.548959+00:00"} {"global_step": 9662, "acc_step": 0, "speed/wps": 12954.888674373762, "speed/FLOPS": 203474328513582.5, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.48151695728302, "optim/lr": 0.002947152238146453, "optim/total_tokens": 2532835328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 4.520836353302002, "created_at": "2025-01-15T02:36:58.611480+00:00"} {"global_step": 9663, "acc_step": 0, "speed/wps": 12953.141006571499, "speed/FLOPS": 203446878989200.72, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6156591176986694, "optim/lr": 0.0029471296933132516, "optim/total_tokens": 2533097472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425730, "loss/out": 4.304786682128906, "created_at": "2025-01-15T02:37:03.684174+00:00"} {"global_step": 9664, "acc_step": 0, "speed/wps": 12960.875257404501, "speed/FLOPS": 203568356026507.4, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3751397132873535, "optim/lr": 0.0029471071437585355, "optim/total_tokens": 2533359616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447263, "loss/out": 4.540234088897705, "created_at": "2025-01-15T02:37:08.747837+00:00"} {"global_step": 9665, "acc_step": 0, "speed/wps": 12962.28589299209, "speed/FLOPS": 203590511996826.9, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0511226654052734, "optim/lr": 0.002947084589482378, "optim/total_tokens": 2533621760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316688, "loss/out": 4.564840316772461, "created_at": "2025-01-15T02:37:13.813175+00:00"} {"global_step": 9666, "acc_step": 0, "speed/wps": 12966.431182627017, "speed/FLOPS": 203655619466768.3, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0614749193191528, "optim/lr": 0.0029470620304848523, "optim/total_tokens": 2533883904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359360, "loss/out": 4.48013973236084, "created_at": "2025-01-15T02:37:18.876452+00:00"} {"global_step": 9667, "acc_step": 0, "speed/wps": 12962.847470853698, "speed/FLOPS": 203599332348832.3, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.213284969329834, "optim/lr": 0.0029470394667660328, "optim/total_tokens": 2534146048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448083, "loss/out": 4.490350246429443, "created_at": "2025-01-15T02:37:23.936253+00:00"} {"global_step": 9668, "acc_step": 0, "speed/wps": 12957.070115284674, "speed/FLOPS": 203508591040702.3, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1110498905181885, "optim/lr": 0.0029470168983259924, "optim/total_tokens": 2534408192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448605, "loss/out": 4.561210632324219, "created_at": "2025-01-15T02:37:29.000618+00:00"} {"global_step": 9669, "acc_step": 0, "speed/wps": 12961.63418804346, "speed/FLOPS": 203580276075072.22, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5974943041801453, "optim/lr": 0.0029469943251648053, "optim/total_tokens": 2534670336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453131, "loss/out": 4.526932239532471, "created_at": "2025-01-15T02:37:34.063039+00:00"} {"global_step": 9670, "acc_step": 0, "speed/wps": 12957.699106799457, "speed/FLOPS": 203518470216766.88, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3169832229614258, "optim/lr": 0.002946971747282545, "optim/total_tokens": 2534932480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454995, "loss/out": 4.520780086517334, "created_at": "2025-01-15T02:37:39.127522+00:00"} {"global_step": 9671, "acc_step": 0, "speed/wps": 12962.490103571728, "speed/FLOPS": 203593719404594.88, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8026665449142456, "optim/lr": 0.002946949164679285, "optim/total_tokens": 2535194624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 4.655296325683594, "created_at": "2025-01-15T02:37:44.189091+00:00"} {"global_step": 9672, "acc_step": 0, "speed/wps": 12961.068738102924, "speed/FLOPS": 203571394906746.47, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.004847764968872, "optim/lr": 0.002946926577355099, "optim/total_tokens": 2535456768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451797, "loss/out": 4.378232479095459, "created_at": "2025-01-15T02:37:49.253168+00:00"} {"global_step": 9673, "acc_step": 0, "speed/wps": 12965.670916127125, "speed/FLOPS": 203643678436671.88, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7709794640541077, "optim/lr": 0.002946903985310061, "optim/total_tokens": 2535718912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295894, "loss/out": 4.3450846672058105, "created_at": "2025-01-15T02:37:54.315061+00:00"} {"global_step": 9674, "acc_step": 0, "speed/wps": 12959.249104860402, "speed/FLOPS": 203542815066234.56, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1747486591339111, "optim/lr": 0.002946881388544244, "optim/total_tokens": 2535981056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 412372, "loss/out": 4.462889671325684, "created_at": "2025-01-15T02:37:59.380372+00:00"} {"global_step": 9675, "acc_step": 0, "speed/wps": 12958.598781478651, "speed/FLOPS": 203532600843886.2, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.001, "optim/grad_norm": 0.9492745995521545, "optim/lr": 0.0029468587870577225, "optim/total_tokens": 2536243200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 4.540073394775391, "created_at": "2025-01-15T02:38:04.445359+00:00"} {"global_step": 9676, "acc_step": 0, "speed/wps": 12961.239184373102, "speed/FLOPS": 203574071999637.34, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2944388389587402, "optim/lr": 0.0029468361808505693, "optim/total_tokens": 2536505344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439820, "loss/out": 4.5055341720581055, "created_at": "2025-01-15T02:38:09.508776+00:00"} {"global_step": 9677, "acc_step": 0, "speed/wps": 12950.21197510762, "speed/FLOPS": 203400874525149.5, "speed/curr_iter_time": 1.2774, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8482297658920288, "optim/lr": 0.0029468135699228593, "optim/total_tokens": 2536767488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 4.436336994171143, "created_at": "2025-01-15T02:38:14.575782+00:00"} {"global_step": 9678, "acc_step": 0, "speed/wps": 12961.750920148454, "speed/FLOPS": 203582109513187.88, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4293642044067383, "optim/lr": 0.0029467909542746653, "optim/total_tokens": 2537029632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439583, "loss/out": 4.2738566398620605, "created_at": "2025-01-15T02:38:19.639580+00:00"} {"global_step": 9679, "acc_step": 0, "speed/wps": 12964.334978490253, "speed/FLOPS": 203622695700314.1, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0216246843338013, "optim/lr": 0.0029467683339060627, "optim/total_tokens": 2537291776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311028, "loss/out": 4.477447986602783, "created_at": "2025-01-15T02:38:24.704146+00:00"} {"global_step": 9680, "acc_step": 0, "speed/wps": 12963.59211936986, "speed/FLOPS": 203611028077032.94, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.348000168800354, "optim/lr": 0.002946745708817123, "optim/total_tokens": 2537553920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401352, "loss/out": 4.4853973388671875, "created_at": "2025-01-15T02:38:29.767080+00:00"} {"global_step": 9681, "acc_step": 0, "speed/wps": 12964.513118175339, "speed/FLOPS": 203625493628857.97, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0771801471710205, "optim/lr": 0.0029467230790079213, "optim/total_tokens": 2537816064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442778, "loss/out": 4.401467323303223, "created_at": "2025-01-15T02:38:34.830147+00:00"} {"global_step": 9682, "acc_step": 0, "speed/wps": 12964.260900363533, "speed/FLOPS": 203621532201539.97, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7614575028419495, "optim/lr": 0.0029467004444785313, "optim/total_tokens": 2538078208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447269, "loss/out": 4.437076568603516, "created_at": "2025-01-15T02:38:39.888942+00:00"} {"global_step": 9683, "acc_step": 0, "speed/wps": 12960.43274048305, "speed/FLOPS": 203561405690174.03, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1435983180999756, "optim/lr": 0.0029466778052290267, "optim/total_tokens": 2538340352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 4.45140266418457, "created_at": "2025-01-15T02:38:44.949555+00:00"} {"global_step": 9684, "acc_step": 0, "speed/wps": 12954.732633150523, "speed/FLOPS": 203471877671748.4, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8281976580619812, "optim/lr": 0.0029466551612594826, "optim/total_tokens": 2538602496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450746, "loss/out": 4.533107757568359, "created_at": "2025-01-15T02:38:50.011876+00:00"} {"global_step": 9685, "acc_step": 0, "speed/wps": 12962.435492537752, "speed/FLOPS": 203592861663262.84, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1639561653137207, "optim/lr": 0.0029466325125699707, "optim/total_tokens": 2538864640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470403, "loss/out": 4.478238105773926, "created_at": "2025-01-15T02:38:55.071318+00:00"} {"global_step": 9686, "acc_step": 0, "speed/wps": 12955.52079104183, "speed/FLOPS": 203484256774473.84, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.049761176109314, "optim/lr": 0.002946609859160566, "optim/total_tokens": 2539126784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 4.480737686157227, "created_at": "2025-01-15T02:39:00.136362+00:00"} {"global_step": 9687, "acc_step": 0, "speed/wps": 12947.753844529223, "speed/FLOPS": 203362266206590.38, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.138375997543335, "optim/lr": 0.0029465872010313427, "optim/total_tokens": 2539388928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307800, "loss/out": 4.538235187530518, "created_at": "2025-01-15T02:39:05.204906+00:00"} {"global_step": 9688, "acc_step": 0, "speed/wps": 12963.432243210971, "speed/FLOPS": 203608517002264.72, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1874319314956665, "optim/lr": 0.002946564538182374, "optim/total_tokens": 2539651072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 4.40932559967041, "created_at": "2025-01-15T02:39:10.269445+00:00"} {"global_step": 9689, "acc_step": 0, "speed/wps": 12958.574808132451, "speed/FLOPS": 203532224309541.28, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5389809608459473, "optim/lr": 0.002946541870613735, "optim/total_tokens": 2539913216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 4.553049564361572, "created_at": "2025-01-15T02:39:15.332223+00:00"} {"global_step": 9690, "acc_step": 0, "speed/wps": 12956.306811996074, "speed/FLOPS": 203496602313666.2, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.028833270072937, "optim/lr": 0.0029465191983254983, "optim/total_tokens": 2540175360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 4.3258056640625, "created_at": "2025-01-15T02:39:20.394383+00:00"} {"global_step": 9691, "acc_step": 0, "speed/wps": 12965.617639988333, "speed/FLOPS": 203642841661703.06, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1262569427490234, "optim/lr": 0.002946496521317739, "optim/total_tokens": 2540437504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461913, "loss/out": 4.456568717956543, "created_at": "2025-01-15T02:39:25.457971+00:00"} {"global_step": 9692, "acc_step": 0, "speed/wps": 12963.932392046321, "speed/FLOPS": 203616372527000.4, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9652661085128784, "optim/lr": 0.0029464738395905297, "optim/total_tokens": 2540699648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452847, "loss/out": 4.566622734069824, "created_at": "2025-01-15T02:39:30.523075+00:00"} {"global_step": 9693, "acc_step": 0, "speed/wps": 12957.490043798773, "speed/FLOPS": 203515186595058.47, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0423011779785156, "optim/lr": 0.0029464511531439455, "optim/total_tokens": 2540961792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304748, "loss/out": 4.447624206542969, "created_at": "2025-01-15T02:39:35.584369+00:00"} {"global_step": 9694, "acc_step": 0, "speed/wps": 12959.875660955693, "speed/FLOPS": 203552655990689.4, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.9995798468589783, "optim/lr": 0.002946428461978061, "optim/total_tokens": 2541223936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 347485, "loss/out": 4.396729469299316, "created_at": "2025-01-15T02:39:40.645620+00:00"} {"global_step": 9695, "acc_step": 0, "speed/wps": 12959.228355781795, "speed/FLOPS": 203542489173442.8, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1198066473007202, "optim/lr": 0.0029464057660929487, "optim/total_tokens": 2541486080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 380792, "loss/out": 4.410913467407227, "created_at": "2025-01-15T02:39:45.710962+00:00"} {"global_step": 9696, "acc_step": 0, "speed/wps": 12955.35970426735, "speed/FLOPS": 203481726685324.2, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.57671058177948, "optim/lr": 0.002946383065488684, "optim/total_tokens": 2541748224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436666, "loss/out": 4.546915531158447, "created_at": "2025-01-15T02:39:50.774304+00:00"} {"global_step": 9697, "acc_step": 0, "speed/wps": 12958.76451642468, "speed/FLOPS": 203535203938957.28, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.43364429473876953, "optim/lr": 0.0029463603601653393, "optim/total_tokens": 2542010368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 4.563467502593994, "created_at": "2025-01-15T02:39:55.841722+00:00"} {"global_step": 9698, "acc_step": 0, "speed/wps": 12962.843282431333, "speed/FLOPS": 203599266563903.6, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9034437537193298, "optim/lr": 0.0029463376501229904, "optim/total_tokens": 2542272512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424156, "loss/out": 4.430749416351318, "created_at": "2025-01-15T02:40:00.906574+00:00"} {"global_step": 9699, "acc_step": 0, "speed/wps": 12958.853781185611, "speed/FLOPS": 203536605964690.3, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0091640949249268, "optim/lr": 0.0029463149353617106, "optim/total_tokens": 2542534656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 4.377330303192139, "created_at": "2025-01-15T02:40:05.975866+00:00"} {"global_step": 9700, "acc_step": 0, "speed/wps": 12956.737574481056, "speed/FLOPS": 203503368030422.53, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5313705801963806, "optim/lr": 0.0029462922158815744, "optim/total_tokens": 2542796800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306094, "loss/out": 4.373834609985352, "created_at": "2025-01-15T02:40:11.044129+00:00"} {"global_step": 9701, "acc_step": 0, "speed/wps": 12961.233738865703, "speed/FLOPS": 203573986470460.72, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6722735166549683, "optim/lr": 0.0029462694916826556, "optim/total_tokens": 2543058944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429723, "loss/out": 4.3842291831970215, "created_at": "2025-01-15T02:40:16.106565+00:00"} {"global_step": 9702, "acc_step": 0, "speed/wps": 12965.529669979687, "speed/FLOPS": 203641459971833.8, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7835021018981934, "optim/lr": 0.0029462467627650282, "optim/total_tokens": 2543321088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462963, "loss/out": 4.4583892822265625, "created_at": "2025-01-15T02:40:21.167334+00:00"} {"global_step": 9703, "acc_step": 0, "speed/wps": 12956.864768064132, "speed/FLOPS": 203505365780427.03, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9965649247169495, "optim/lr": 0.0029462240291287666, "optim/total_tokens": 2543583232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460815, "loss/out": 4.530088424682617, "created_at": "2025-01-15T02:40:26.229880+00:00"} {"global_step": 9704, "acc_step": 0, "speed/wps": 12955.807489899857, "speed/FLOPS": 203488759774004.66, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1105719804763794, "optim/lr": 0.0029462012907739447, "optim/total_tokens": 2543845376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 4.415056228637695, "created_at": "2025-01-15T02:40:31.297270+00:00"} {"global_step": 9705, "acc_step": 0, "speed/wps": 12968.820067256936, "speed/FLOPS": 203693140182552.7, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1087263822555542, "optim/lr": 0.0029461785477006373, "optim/total_tokens": 2544107520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446220, "loss/out": 4.4755754470825195, "created_at": "2025-01-15T02:40:36.355640+00:00"} {"global_step": 9706, "acc_step": 0, "speed/wps": 12958.054823597427, "speed/FLOPS": 203524057237885.12, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2681597471237183, "optim/lr": 0.002946155799908918, "optim/total_tokens": 2544369664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 4.396557331085205, "created_at": "2025-01-15T02:40:41.422332+00:00"} {"global_step": 9707, "acc_step": 0, "speed/wps": 12960.82635817798, "speed/FLOPS": 203567587996960.28, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7995635867118835, "optim/lr": 0.002946133047398861, "optim/total_tokens": 2544631808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 403566, "loss/out": 4.449746131896973, "created_at": "2025-01-15T02:40:46.482689+00:00"} {"global_step": 9708, "acc_step": 0, "speed/wps": 12951.817189850504, "speed/FLOPS": 203426086628482.7, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7781249284744263, "optim/lr": 0.002946110290170541, "optim/total_tokens": 2544893952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445695, "loss/out": 4.409385681152344, "created_at": "2025-01-15T02:40:51.546264+00:00"} {"global_step": 9709, "acc_step": 0, "speed/wps": 12954.620489562936, "speed/FLOPS": 203470116302603.94, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0617979764938354, "optim/lr": 0.0029460875282240322, "optim/total_tokens": 2545156096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450990, "loss/out": 4.384609222412109, "created_at": "2025-01-15T02:40:56.608938+00:00"} {"global_step": 9710, "acc_step": 0, "speed/wps": 12955.411865737567, "speed/FLOPS": 203482545952891.34, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7401699423789978, "optim/lr": 0.002946064761559408, "optim/total_tokens": 2545418240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477242, "loss/out": 4.4671220779418945, "created_at": "2025-01-15T02:41:01.671306+00:00"} {"global_step": 9711, "acc_step": 0, "speed/wps": 12955.720836071207, "speed/FLOPS": 203487398756550.88, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9005693793296814, "optim/lr": 0.002946041990176744, "optim/total_tokens": 2545680384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424156, "loss/out": 4.450266361236572, "created_at": "2025-01-15T02:41:06.743605+00:00"} {"global_step": 9712, "acc_step": 0, "speed/wps": 12952.546080827353, "speed/FLOPS": 203437534862875.28, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6774407625198364, "optim/lr": 0.0029460192140761134, "optim/total_tokens": 2545942528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441971, "loss/out": 4.300308704376221, "created_at": "2025-01-15T02:41:11.806906+00:00"} {"global_step": 9713, "acc_step": 0, "speed/wps": 12957.623947276325, "speed/FLOPS": 203517289733176.1, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3226059675216675, "optim/lr": 0.002945996433257592, "optim/total_tokens": 2546204672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302050, "loss/out": 4.4334259033203125, "created_at": "2025-01-15T02:41:16.873859+00:00"} {"global_step": 9714, "acc_step": 0, "speed/wps": 12963.213158737755, "speed/FLOPS": 203605075979559.25, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7099988460540771, "optim/lr": 0.0029459736477212522, "optim/total_tokens": 2546466816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429723, "loss/out": 4.491396903991699, "created_at": "2025-01-15T02:41:21.933616+00:00"} {"global_step": 9715, "acc_step": 0, "speed/wps": 12955.891299146946, "speed/FLOPS": 203490076113396.16, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9582438468933105, "optim/lr": 0.002945950857467169, "optim/total_tokens": 2546728960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452081, "loss/out": 4.393762588500977, "created_at": "2025-01-15T02:41:27.001334+00:00"} {"global_step": 9716, "acc_step": 0, "speed/wps": 12950.282503715756, "speed/FLOPS": 203401982273856.75, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.389961004257202, "optim/lr": 0.0029459280624954177, "optim/total_tokens": 2546991104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448361, "loss/out": 4.431193828582764, "created_at": "2025-01-15T02:41:32.066672+00:00"} {"global_step": 9717, "acc_step": 0, "speed/wps": 12962.471376121019, "speed/FLOPS": 203593425264247.28, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.355614185333252, "optim/lr": 0.0029459052628060716, "optim/total_tokens": 2547253248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450990, "loss/out": 4.429819107055664, "created_at": "2025-01-15T02:41:37.131693+00:00"} {"global_step": 9718, "acc_step": 0, "speed/wps": 12950.58438427732, "speed/FLOPS": 203406723722903.44, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2678101062774658, "optim/lr": 0.0029458824583992052, "optim/total_tokens": 2547515392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444641, "loss/out": 4.391577243804932, "created_at": "2025-01-15T02:41:42.196841+00:00"} {"global_step": 9719, "acc_step": 0, "speed/wps": 12955.627607215547, "speed/FLOPS": 203485934469262.84, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3284298181533813, "optim/lr": 0.0029458596492748937, "optim/total_tokens": 2547777536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305561, "loss/out": 4.495809078216553, "created_at": "2025-01-15T02:41:47.259640+00:00"} {"global_step": 9720, "acc_step": 0, "speed/wps": 12951.056278706505, "speed/FLOPS": 203414135473364.7, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8344480395317078, "optim/lr": 0.002945836835433211, "optim/total_tokens": 2548039680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 396624, "loss/out": 4.401845455169678, "created_at": "2025-01-15T02:41:52.323681+00:00"} {"global_step": 9721, "acc_step": 0, "speed/wps": 12956.984057707452, "speed/FLOPS": 203507239388196.53, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1697076559066772, "optim/lr": 0.0029458140168742313, "optim/total_tokens": 2548301824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437723, "loss/out": 4.502469062805176, "created_at": "2025-01-15T02:41:57.385503+00:00"} {"global_step": 9722, "acc_step": 0, "speed/wps": 12958.162430827877, "speed/FLOPS": 203525747357307.16, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1628925800323486, "optim/lr": 0.0029457911935980292, "optim/total_tokens": 2548563968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 4.503461837768555, "created_at": "2025-01-15T02:42:02.450346+00:00"} {"global_step": 9723, "acc_step": 0, "speed/wps": 12952.645979157525, "speed/FLOPS": 203439103903425.3, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5354803800582886, "optim/lr": 0.002945768365604679, "optim/total_tokens": 2548826112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500274, "loss/out": 4.504710674285889, "created_at": "2025-01-15T02:42:07.513646+00:00"} {"global_step": 9724, "acc_step": 0, "speed/wps": 12957.457368248617, "speed/FLOPS": 203514673380642.38, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5466681718826294, "optim/lr": 0.0029457455328942557, "optim/total_tokens": 2549088256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432941, "loss/out": 4.510143756866455, "created_at": "2025-01-15T02:42:12.580005+00:00"} {"global_step": 9725, "acc_step": 0, "speed/wps": 12956.29051415951, "speed/FLOPS": 203496346333747.38, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.684395432472229, "optim/lr": 0.002945722695466834, "optim/total_tokens": 2549350400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 4.446974754333496, "created_at": "2025-01-15T02:42:17.647049+00:00"} {"global_step": 9726, "acc_step": 0, "speed/wps": 12952.274450322742, "speed/FLOPS": 203433268532540.94, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.149167776107788, "optim/lr": 0.0029456998533224873, "optim/total_tokens": 2549612544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305035, "loss/out": 4.449804782867432, "created_at": "2025-01-15T02:42:22.715352+00:00"} {"global_step": 9727, "acc_step": 0, "speed/wps": 12949.650643109355, "speed/FLOPS": 203392058034765.75, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6207976937294006, "optim/lr": 0.0029456770064612905, "optim/total_tokens": 2549874688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448605, "loss/out": 4.451015472412109, "created_at": "2025-01-15T02:42:27.783911+00:00"} {"global_step": 9728, "acc_step": 0, "speed/wps": 12955.169110604102, "speed/FLOPS": 203478733149939.75, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8922701478004456, "optim/lr": 0.0029456541548833187, "optim/total_tokens": 2550136832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 4.40533971786499, "created_at": "2025-01-15T02:42:32.847109+00:00"} {"global_step": 9729, "acc_step": 0, "speed/wps": 12954.196590414822, "speed/FLOPS": 203463458383984.3, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8913045525550842, "optim/lr": 0.0029456312985886466, "optim/total_tokens": 2550398976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430567, "loss/out": 4.586801528930664, "created_at": "2025-01-15T02:42:37.917324+00:00"} {"global_step": 9730, "acc_step": 0, "speed/wps": 12957.334524447331, "speed/FLOPS": 203512743949938.1, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4854770004749298, "optim/lr": 0.0029456084375773474, "optim/total_tokens": 2550661120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 4.398014545440674, "created_at": "2025-01-15T02:42:42.979446+00:00"} {"global_step": 9731, "acc_step": 0, "speed/wps": 12960.492368949039, "speed/FLOPS": 203562342237170.5, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7925384640693665, "optim/lr": 0.0029455855718494976, "optim/total_tokens": 2550923264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420140, "loss/out": 4.3670477867126465, "created_at": "2025-01-15T02:42:48.044268+00:00"} {"global_step": 9732, "acc_step": 0, "speed/wps": 12960.138134262816, "speed/FLOPS": 203556778494653.88, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0144155025482178, "optim/lr": 0.0029455627014051697, "optim/total_tokens": 2551185408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353982, "loss/out": 4.538283824920654, "created_at": "2025-01-15T02:42:53.108776+00:00"} {"global_step": 9733, "acc_step": 0, "speed/wps": 12965.387246811377, "speed/FLOPS": 203639223020265.56, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6025627255439758, "optim/lr": 0.0029455398262444405, "optim/total_tokens": 2551447552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445166, "loss/out": 4.437045574188232, "created_at": "2025-01-15T02:42:58.167065+00:00"} {"global_step": 9734, "acc_step": 0, "speed/wps": 12961.533637720911, "speed/FLOPS": 203578696794086.0, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.692627489566803, "optim/lr": 0.002945516946367383, "optim/total_tokens": 2551709696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425201, "loss/out": 4.441096305847168, "created_at": "2025-01-15T02:43:03.232281+00:00"} {"global_step": 9735, "acc_step": 0, "speed/wps": 12956.008850986635, "speed/FLOPS": 203491922426571.62, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8589169979095459, "optim/lr": 0.0029454940617740723, "optim/total_tokens": 2551971840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 4.550358295440674, "created_at": "2025-01-15T02:43:08.294559+00:00"} {"global_step": 9736, "acc_step": 0, "speed/wps": 12959.106817252306, "speed/FLOPS": 203540580243825.75, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7237023115158081, "optim/lr": 0.0029454711724645833, "optim/total_tokens": 2552233984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 4.4357099533081055, "created_at": "2025-01-15T02:43:13.359523+00:00"} {"global_step": 9737, "acc_step": 0, "speed/wps": 12962.127078131994, "speed/FLOPS": 203588017591214.62, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1207774877548218, "optim/lr": 0.0029454482784389904, "optim/total_tokens": 2552496128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432365, "loss/out": 4.517004489898682, "created_at": "2025-01-15T02:43:18.422504+00:00"} {"global_step": 9738, "acc_step": 0, "speed/wps": 12958.701190617858, "speed/FLOPS": 203534209320142.34, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0013, "optim/grad_norm": 6.747145652770996, "optim/lr": 0.0029454253796973684, "optim/total_tokens": 2552758272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476431, "loss/out": 4.478111267089844, "created_at": "2025-01-15T02:43:23.489148+00:00"} {"global_step": 9739, "acc_step": 0, "speed/wps": 12960.543060737766, "speed/FLOPS": 203563138421370.25, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.409607172012329, "optim/lr": 0.0029454024762397926, "optim/total_tokens": 2553020416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338641, "loss/out": 4.465146064758301, "created_at": "2025-01-15T02:43:28.570670+00:00"} {"global_step": 9740, "acc_step": 0, "speed/wps": 12961.829078568811, "speed/FLOPS": 203583337098580.53, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5015393495559692, "optim/lr": 0.0029453795680663367, "optim/total_tokens": 2553282560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 4.474489688873291, "created_at": "2025-01-15T02:43:33.630366+00:00"} {"global_step": 9741, "acc_step": 0, "speed/wps": 12963.737137346525, "speed/FLOPS": 203613305783632.03, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.803903341293335, "optim/lr": 0.0029453566551770763, "optim/total_tokens": 2553544704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 4.499738693237305, "created_at": "2025-01-15T02:43:38.694872+00:00"} {"global_step": 9742, "acc_step": 0, "speed/wps": 12960.913669958423, "speed/FLOPS": 203568959348451.12, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9886746406555176, "optim/lr": 0.0029453337375720855, "optim/total_tokens": 2553806848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417752, "loss/out": 4.391331672668457, "created_at": "2025-01-15T02:43:43.760546+00:00"} {"global_step": 9743, "acc_step": 0, "speed/wps": 12958.210021397064, "speed/FLOPS": 203526494832592.53, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6753947138786316, "optim/lr": 0.0029453108152514393, "optim/total_tokens": 2554068992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 4.379818916320801, "created_at": "2025-01-15T02:43:48.826848+00:00"} {"global_step": 9744, "acc_step": 0, "speed/wps": 12953.3385150037, "speed/FLOPS": 203449981130533.47, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.04843008518219, "optim/lr": 0.0029452878882152127, "optim/total_tokens": 2554331136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297348, "loss/out": 4.426502227783203, "created_at": "2025-01-15T02:43:53.895306+00:00"} {"global_step": 9745, "acc_step": 0, "speed/wps": 12964.331173650804, "speed/FLOPS": 203622635940082.16, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.127290964126587, "optim/lr": 0.0029452649564634804, "optim/total_tokens": 2554593280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439048, "loss/out": 4.456218719482422, "created_at": "2025-01-15T02:43:58.958316+00:00"} {"global_step": 9746, "acc_step": 0, "speed/wps": 12962.01983720482, "speed/FLOPS": 203586333225088.88, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.610343873500824, "optim/lr": 0.0029452420199963174, "optim/total_tokens": 2554855424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 4.361804485321045, "created_at": "2025-01-15T02:44:04.022770+00:00"} {"global_step": 9747, "acc_step": 0, "speed/wps": 12959.338758159012, "speed/FLOPS": 203544223194485.78, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0354963541030884, "optim/lr": 0.0029452190788137977, "optim/total_tokens": 2555117568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439064, "loss/out": 4.395730018615723, "created_at": "2025-01-15T02:44:09.086751+00:00"} {"global_step": 9748, "acc_step": 0, "speed/wps": 12954.007290078422, "speed/FLOPS": 203460485162075.22, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0263664722442627, "optim/lr": 0.0029451961329159965, "optim/total_tokens": 2555379712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438769, "loss/out": 4.444502830505371, "created_at": "2025-01-15T02:44:14.150503+00:00"} {"global_step": 9749, "acc_step": 0, "speed/wps": 12960.408521647902, "speed/FLOPS": 203561025300087.88, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7558602690696716, "optim/lr": 0.0029451731823029897, "optim/total_tokens": 2555641856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460287, "loss/out": 4.625457763671875, "created_at": "2025-01-15T02:44:19.211008+00:00"} {"global_step": 9750, "acc_step": 0, "speed/wps": 12958.422548670775, "speed/FLOPS": 203529832865469.25, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0928593873977661, "optim/lr": 0.002945150226974852, "optim/total_tokens": 2555904000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 4.510815143585205, "created_at": "2025-01-15T02:44:24.276093+00:00"} {"global_step": 9751, "acc_step": 0, "speed/wps": 12958.174777044045, "speed/FLOPS": 203525941271597.56, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0102640390396118, "optim/lr": 0.0029451272669316565, "optim/total_tokens": 2556166144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317214, "loss/out": 4.38082218170166, "created_at": "2025-01-15T02:44:29.342496+00:00"} {"global_step": 9752, "acc_step": 0, "speed/wps": 12957.72107501903, "speed/FLOPS": 203518815257842.78, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8794046640396118, "optim/lr": 0.0029451043021734794, "optim/total_tokens": 2556428288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423841, "loss/out": 4.475934028625488, "created_at": "2025-01-15T02:44:34.403871+00:00"} {"global_step": 9753, "acc_step": 0, "speed/wps": 12961.591168186307, "speed/FLOPS": 203579600389100.16, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3784809112548828, "optim/lr": 0.002945081332700396, "optim/total_tokens": 2556690432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445404, "loss/out": 4.508978843688965, "created_at": "2025-01-15T02:44:39.464979+00:00"} {"global_step": 9754, "acc_step": 0, "speed/wps": 12962.51394097506, "speed/FLOPS": 203594093803770.25, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.8295329213142395, "optim/lr": 0.0029450583585124805, "optim/total_tokens": 2556952576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 4.455994129180908, "created_at": "2025-01-15T02:44:44.531939+00:00"} {"global_step": 9755, "acc_step": 0, "speed/wps": 12961.25652121845, "speed/FLOPS": 203574344298616.88, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1531867980957031, "optim/lr": 0.002945035379609809, "optim/total_tokens": 2557214720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447791, "loss/out": 4.373480796813965, "created_at": "2025-01-15T02:44:49.597233+00:00"} {"global_step": 9756, "acc_step": 0, "speed/wps": 12956.313920842782, "speed/FLOPS": 203496713967872.34, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6344708204269409, "optim/lr": 0.002945012395992454, "optim/total_tokens": 2557476864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448890, "loss/out": 4.393316268920898, "created_at": "2025-01-15T02:44:54.663702+00:00"} {"global_step": 9757, "acc_step": 0, "speed/wps": 12969.230011039237, "speed/FLOPS": 203699578912975.5, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7591887712478638, "optim/lr": 0.0029449894076604933, "optim/total_tokens": 2557739008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315181, "loss/out": 4.524290561676025, "created_at": "2025-01-15T02:44:59.722978+00:00"} {"global_step": 9758, "acc_step": 0, "speed/wps": 12958.849292263976, "speed/FLOPS": 203536535460007.62, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0588961839675903, "optim/lr": 0.0029449664146139996, "optim/total_tokens": 2558001152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 397433, "loss/out": 4.345015048980713, "created_at": "2025-01-15T02:45:04.786223+00:00"} {"global_step": 9759, "acc_step": 0, "speed/wps": 12950.435518140179, "speed/FLOPS": 203404385575656.16, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0291370153427124, "optim/lr": 0.00294494341685305, "optim/total_tokens": 2558263296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483142, "loss/out": 4.451090335845947, "created_at": "2025-01-15T02:45:09.858808+00:00"} {"global_step": 9760, "acc_step": 0, "speed/wps": 12956.465281490959, "speed/FLOPS": 203499091294843.28, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0009, "optim/grad_norm": 2.7305822372436523, "optim/lr": 0.0029449204143777185, "optim/total_tokens": 2558525440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428688, "loss/out": 4.230138301849365, "created_at": "2025-01-15T02:45:14.921269+00:00"} {"global_step": 9761, "acc_step": 0, "speed/wps": 12967.467133145774, "speed/FLOPS": 203671890493210.6, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0234105587005615, "optim/lr": 0.00294489740718808, "optim/total_tokens": 2558787584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437195, "loss/out": 4.460502624511719, "created_at": "2025-01-15T02:45:19.983086+00:00"} {"global_step": 9762, "acc_step": 0, "speed/wps": 12957.813739874351, "speed/FLOPS": 203520270686730.44, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.49738603830337524, "optim/lr": 0.0029448743952842095, "optim/total_tokens": 2559049728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446218, "loss/out": 4.507580757141113, "created_at": "2025-01-15T02:45:25.046864+00:00"} {"global_step": 9763, "acc_step": 0, "speed/wps": 12960.505259081867, "speed/FLOPS": 203562544694419.47, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6684598922729492, "optim/lr": 0.0029448513786661827, "optim/total_tokens": 2559311872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 4.454451560974121, "created_at": "2025-01-15T02:45:30.109446+00:00"} {"global_step": 9764, "acc_step": 0, "speed/wps": 12961.128085946058, "speed/FLOPS": 203572327046175.38, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4532237648963928, "optim/lr": 0.0029448283573340743, "optim/total_tokens": 2559574016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433464, "loss/out": 4.415366172790527, "created_at": "2025-01-15T02:45:35.169315+00:00"} {"global_step": 9765, "acc_step": 0, "speed/wps": 12956.540167542378, "speed/FLOPS": 203500267483185.03, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0669132471084595, "optim/lr": 0.0029448053312879593, "optim/total_tokens": 2559836160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324322, "loss/out": 4.358713626861572, "created_at": "2025-01-15T02:45:40.231645+00:00"} {"global_step": 9766, "acc_step": 0, "speed/wps": 12958.872371702259, "speed/FLOPS": 203536897954300.3, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.832790732383728, "optim/lr": 0.0029447823005279133, "optim/total_tokens": 2560098304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450462, "loss/out": 4.4062066078186035, "created_at": "2025-01-15T02:45:45.296061+00:00"} {"global_step": 9767, "acc_step": 0, "speed/wps": 12962.676798416225, "speed/FLOPS": 203596651703673.3, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6100541353225708, "optim/lr": 0.0029447592650540108, "optim/total_tokens": 2560360448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 4.46216344833374, "created_at": "2025-01-15T02:45:50.359507+00:00"} {"global_step": 9768, "acc_step": 0, "speed/wps": 12958.272349726967, "speed/FLOPS": 203527473784661.75, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6159512400627136, "optim/lr": 0.0029447362248663273, "optim/total_tokens": 2560622592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432431, "loss/out": 4.479986190795898, "created_at": "2025-01-15T02:45:55.424224+00:00"} {"global_step": 9769, "acc_step": 0, "speed/wps": 12959.90344612769, "speed/FLOPS": 203553092394996.44, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6634578108787537, "optim/lr": 0.0029447131799649378, "optim/total_tokens": 2560884736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 4.426486015319824, "created_at": "2025-01-15T02:46:00.487606+00:00"} {"global_step": 9770, "acc_step": 0, "speed/wps": 12944.651054443912, "speed/FLOPS": 203313532624620.2, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2200080156326294, "optim/lr": 0.002944690130349918, "optim/total_tokens": 2561146880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 321917, "loss/out": 4.395244598388672, "created_at": "2025-01-15T02:46:05.561783+00:00"} {"global_step": 9771, "acc_step": 0, "speed/wps": 12957.161048258604, "speed/FLOPS": 203510019268012.88, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1551707983016968, "optim/lr": 0.0029446670760213426, "optim/total_tokens": 2561409024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453393, "loss/out": 4.364224910736084, "created_at": "2025-01-15T02:46:10.625557+00:00"} {"global_step": 9772, "acc_step": 0, "speed/wps": 12959.160039898063, "speed/FLOPS": 203541416178612.94, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6319208741188049, "optim/lr": 0.002944644016979287, "optim/total_tokens": 2561671168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 4.328975200653076, "created_at": "2025-01-15T02:46:15.690957+00:00"} {"global_step": 9773, "acc_step": 0, "speed/wps": 12959.108642962477, "speed/FLOPS": 203540608919112.78, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6855013370513916, "optim/lr": 0.0029446209532238265, "optim/total_tokens": 2561933312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449411, "loss/out": 4.447653770446777, "created_at": "2025-01-15T02:46:20.752843+00:00"} {"global_step": 9774, "acc_step": 0, "speed/wps": 12954.149768584859, "speed/FLOPS": 203462722982804.8, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.039884090423584, "optim/lr": 0.0029445978847550355, "optim/total_tokens": 2562195456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 4.373804569244385, "created_at": "2025-01-15T02:46:25.820342+00:00"} {"global_step": 9775, "acc_step": 0, "speed/wps": 12965.685462564774, "speed/FLOPS": 203643906908464.03, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4528943300247192, "optim/lr": 0.002944574811572991, "optim/total_tokens": 2562457600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 4.445738792419434, "created_at": "2025-01-15T02:46:30.878465+00:00"} {"global_step": 9776, "acc_step": 0, "speed/wps": 12964.106381585416, "speed/FLOPS": 203619105271798.56, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8610659241676331, "optim/lr": 0.0029445517336777667, "optim/total_tokens": 2562719744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 4.4408040046691895, "created_at": "2025-01-15T02:46:35.941626+00:00"} {"global_step": 9777, "acc_step": 0, "speed/wps": 12957.775904030588, "speed/FLOPS": 203519676422812.25, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8938918709754944, "optim/lr": 0.002944528651069439, "optim/total_tokens": 2562981888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317879, "loss/out": 4.5101799964904785, "created_at": "2025-01-15T02:46:41.002771+00:00"} {"global_step": 9778, "acc_step": 0, "speed/wps": 12959.80831206444, "speed/FLOPS": 203551598183805.22, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3596152067184448, "optim/lr": 0.0029445055637480817, "optim/total_tokens": 2563244032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330091, "loss/out": 4.620973587036133, "created_at": "2025-01-15T02:46:46.063793+00:00"} {"global_step": 9779, "acc_step": 0, "speed/wps": 12960.757049098002, "speed/FLOPS": 203566499402618.6, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.175247311592102, "optim/lr": 0.002944482471713772, "optim/total_tokens": 2563506176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453911, "loss/out": 4.412409782409668, "created_at": "2025-01-15T02:46:51.128049+00:00"} {"global_step": 9780, "acc_step": 0, "speed/wps": 12958.80184734459, "speed/FLOPS": 203535790272352.5, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2283868789672852, "optim/lr": 0.0029444593749665842, "optim/total_tokens": 2563768320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488954, "loss/out": 4.384407043457031, "created_at": "2025-01-15T02:46:56.192037+00:00"} {"global_step": 9781, "acc_step": 0, "speed/wps": 12962.401712771696, "speed/FLOPS": 203592331105618.7, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6081833839416504, "optim/lr": 0.0029444362735065936, "optim/total_tokens": 2564030464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436388, "loss/out": 4.453858375549316, "created_at": "2025-01-15T02:47:01.256547+00:00"} {"global_step": 9782, "acc_step": 0, "speed/wps": 12963.344610921169, "speed/FLOPS": 203607140616734.0, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.9674575924873352, "optim/lr": 0.0029444131673338762, "optim/total_tokens": 2564292608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498247, "loss/out": 4.457614898681641, "created_at": "2025-01-15T02:47:06.317922+00:00"} {"global_step": 9783, "acc_step": 0, "speed/wps": 12958.197472285545, "speed/FLOPS": 203526297731551.84, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.395092487335205, "optim/lr": 0.0029443900564485067, "optim/total_tokens": 2564554752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428711, "loss/out": 4.449831962585449, "created_at": "2025-01-15T02:47:11.379301+00:00"} {"global_step": 9784, "acc_step": 0, "speed/wps": 12956.728282278347, "speed/FLOPS": 203503222083610.16, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9593513011932373, "optim/lr": 0.002944366940850561, "optim/total_tokens": 2564816896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329951, "loss/out": 4.51360559463501, "created_at": "2025-01-15T02:47:16.446192+00:00"} {"global_step": 9785, "acc_step": 0, "speed/wps": 12958.990772350862, "speed/FLOPS": 203538757599186.2, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1056537628173828, "optim/lr": 0.002944343820540114, "optim/total_tokens": 2565079040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450994, "loss/out": 4.645742416381836, "created_at": "2025-01-15T02:47:21.506983+00:00"} {"global_step": 9786, "acc_step": 0, "speed/wps": 12957.142145336906, "speed/FLOPS": 203509722371652.28, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8645938634872437, "optim/lr": 0.0029443206955172417, "optim/total_tokens": 2565341184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 4.412102699279785, "created_at": "2025-01-15T02:47:26.568578+00:00"} {"global_step": 9787, "acc_step": 0, "speed/wps": 12962.077157098473, "speed/FLOPS": 203587233512784.47, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.48593902587890625, "optim/lr": 0.002944297565782019, "optim/total_tokens": 2565603328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 4.463627815246582, "created_at": "2025-01-15T02:47:31.632925+00:00"} {"global_step": 9788, "acc_step": 0, "speed/wps": 12966.100385431315, "speed/FLOPS": 203650423842246.84, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4262290000915527, "optim/lr": 0.002944274431334522, "optim/total_tokens": 2565865472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448605, "loss/out": 4.397623062133789, "created_at": "2025-01-15T02:47:36.695064+00:00"} {"global_step": 9789, "acc_step": 0, "speed/wps": 12962.677199577387, "speed/FLOPS": 203596658004460.6, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.659947872161865, "optim/lr": 0.0029442512921748257, "optim/total_tokens": 2566127616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427351, "loss/out": 4.402010440826416, "created_at": "2025-01-15T02:47:41.757517+00:00"} {"global_step": 9790, "acc_step": 0, "speed/wps": 12963.31647426991, "speed/FLOPS": 203606698691962.16, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5536408424377441, "optim/lr": 0.002944228148303006, "optim/total_tokens": 2566389760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 4.477751731872559, "created_at": "2025-01-15T02:47:46.821957+00:00"} {"global_step": 9791, "acc_step": 0, "speed/wps": 12959.917962919004, "speed/FLOPS": 203553320401152.2, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0671626329421997, "optim/lr": 0.0029442049997191375, "optim/total_tokens": 2566651904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326344, "loss/out": 4.49052095413208, "created_at": "2025-01-15T02:47:51.884668+00:00"} {"global_step": 9792, "acc_step": 0, "speed/wps": 12963.187019068255, "speed/FLOPS": 203604665420130.06, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1086596250534058, "optim/lr": 0.002944181846423296, "optim/total_tokens": 2566914048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363341, "loss/out": 4.395925998687744, "created_at": "2025-01-15T02:47:56.944762+00:00"} {"global_step": 9793, "acc_step": 0, "speed/wps": 12962.891669142708, "speed/FLOPS": 203600026543696.03, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.184950590133667, "optim/lr": 0.002944158688415559, "optim/total_tokens": 2567176192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441157, "loss/out": 4.510429382324219, "created_at": "2025-01-15T02:48:02.006566+00:00"} {"global_step": 9794, "acc_step": 0, "speed/wps": 12963.112003626033, "speed/FLOPS": 203603487199527.97, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5131986141204834, "optim/lr": 0.0029441355256959994, "optim/total_tokens": 2567438336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481758, "loss/out": 4.417473316192627, "created_at": "2025-01-15T02:48:07.073250+00:00"} {"global_step": 9795, "acc_step": 0, "speed/wps": 12960.306875615815, "speed/FLOPS": 203559428809477.66, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.02463698387146, "optim/lr": 0.002944112358264694, "optim/total_tokens": 2567700480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452615, "loss/out": 4.350518226623535, "created_at": "2025-01-15T02:48:12.138375+00:00"} {"global_step": 9796, "acc_step": 0, "speed/wps": 12958.505709264557, "speed/FLOPS": 203531139016869.22, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.077222466468811, "optim/lr": 0.002944089186121718, "optim/total_tokens": 2567962624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325819, "loss/out": 4.370977401733398, "created_at": "2025-01-15T02:48:17.213424+00:00"} {"global_step": 9797, "acc_step": 0, "speed/wps": 12958.505609038803, "speed/FLOPS": 203531137442686.03, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7526975274085999, "optim/lr": 0.0029440660092671478, "optim/total_tokens": 2568224768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365215, "loss/out": 4.458841800689697, "created_at": "2025-01-15T02:48:22.274710+00:00"} {"global_step": 9798, "acc_step": 0, "speed/wps": 12956.359200562565, "speed/FLOPS": 203497425148092.38, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.001, "optim/grad_norm": 0.546948254108429, "optim/lr": 0.002944042827701058, "optim/total_tokens": 2568486912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436096, "loss/out": 4.417059898376465, "created_at": "2025-01-15T02:48:27.347230+00:00"} {"global_step": 9799, "acc_step": 0, "speed/wps": 12964.975482074175, "speed/FLOPS": 203632755689243.75, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8948010206222534, "optim/lr": 0.002944019641423525, "optim/total_tokens": 2568749056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445976, "loss/out": 4.398373603820801, "created_at": "2025-01-15T02:48:32.411206+00:00"} {"global_step": 9800, "acc_step": 0, "speed/wps": 12959.84840816413, "speed/FLOPS": 203552227948148.66, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.131207823753357, "optim/lr": 0.0029439964504346236, "optim/total_tokens": 2569011200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 4.413566589355469, "created_at": "2025-01-15T02:48:37.472067+00:00"} {"global_step": 9801, "acc_step": 0, "speed/wps": 12960.472093299399, "speed/FLOPS": 203562023780231.22, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.772450864315033, "optim/lr": 0.0029439732547344305, "optim/total_tokens": 2569273344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 4.420136451721191, "created_at": "2025-01-15T02:48:42.532539+00:00"} {"global_step": 9802, "acc_step": 0, "speed/wps": 12962.46496968817, "speed/FLOPS": 203593324642416.0, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1875299215316772, "optim/lr": 0.0029439500543230203, "optim/total_tokens": 2569535488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 4.519291877746582, "created_at": "2025-01-15T02:48:47.592403+00:00"} {"global_step": 9803, "acc_step": 0, "speed/wps": 12960.723213626432, "speed/FLOPS": 203565967970042.78, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.093192458152771, "optim/lr": 0.00294392684920047, "optim/total_tokens": 2569797632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441157, "loss/out": 4.380394458770752, "created_at": "2025-01-15T02:48:52.652630+00:00"} {"global_step": 9804, "acc_step": 0, "speed/wps": 12969.37919245399, "speed/FLOPS": 203701922012090.94, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1228382587432861, "optim/lr": 0.0029439036393668545, "optim/total_tokens": 2570059776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324181, "loss/out": 4.495132923126221, "created_at": "2025-01-15T02:48:57.715696+00:00"} {"global_step": 9805, "acc_step": 0, "speed/wps": 12961.578032372523, "speed/FLOPS": 203579394073094.2, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9177379012107849, "optim/lr": 0.0029438804248222493, "optim/total_tokens": 2570321920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 410019, "loss/out": 4.563268184661865, "created_at": "2025-01-15T02:49:02.782343+00:00"} {"global_step": 9806, "acc_step": 0, "speed/wps": 12964.907074531136, "speed/FLOPS": 203631681254778.53, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.564978837966919, "optim/lr": 0.0029438572055667305, "optim/total_tokens": 2570584064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476509, "loss/out": 4.37969446182251, "created_at": "2025-01-15T02:49:07.841017+00:00"} {"global_step": 9807, "acc_step": 0, "speed/wps": 12967.911401935662, "speed/FLOPS": 203678868345045.06, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4811370372772217, "optim/lr": 0.0029438339816003738, "optim/total_tokens": 2570846208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 4.294875144958496, "created_at": "2025-01-15T02:49:12.903746+00:00"} {"global_step": 9808, "acc_step": 0, "speed/wps": 12939.547042012066, "speed/FLOPS": 203233367095730.0, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6836750507354736, "optim/lr": 0.002943810752923255, "optim/total_tokens": 2571108352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 4.4116530418396, "created_at": "2025-01-15T02:49:17.972087+00:00"} {"global_step": 9809, "acc_step": 0, "speed/wps": 12961.977936962647, "speed/FLOPS": 203585675124208.56, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6645395159721375, "optim/lr": 0.0029437875195354496, "optim/total_tokens": 2571370496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438536, "loss/out": 4.356050491333008, "created_at": "2025-01-15T02:49:23.033343+00:00"} {"global_step": 9810, "acc_step": 0, "speed/wps": 12962.165927888032, "speed/FLOPS": 203588627780018.38, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7044923305511475, "optim/lr": 0.002943764281437034, "optim/total_tokens": 2571632640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 285023, "loss/out": 4.331808090209961, "created_at": "2025-01-15T02:49:28.097636+00:00"} {"global_step": 9811, "acc_step": 0, "speed/wps": 12967.541799928795, "speed/FLOPS": 203673063237640.84, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1430658102035522, "optim/lr": 0.002943741038628084, "optim/total_tokens": 2571894784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443307, "loss/out": 4.351292133331299, "created_at": "2025-01-15T02:49:33.156424+00:00"} {"global_step": 9812, "acc_step": 0, "speed/wps": 12961.973698962469, "speed/FLOPS": 203585608560592.16, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.920444130897522, "optim/lr": 0.002943717791108675, "optim/total_tokens": 2572156928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464535, "loss/out": 4.437091827392578, "created_at": "2025-01-15T02:49:38.218520+00:00"} {"global_step": 9813, "acc_step": 0, "speed/wps": 12963.447338676113, "speed/FLOPS": 203608754097288.12, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0898195505142212, "optim/lr": 0.0029436945388788822, "optim/total_tokens": 2572419072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 4.276798725128174, "created_at": "2025-01-15T02:49:43.288320+00:00"} {"global_step": 9814, "acc_step": 0, "speed/wps": 12958.053041135068, "speed/FLOPS": 203524029241864.4, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6285470724105835, "optim/lr": 0.002943671281938783, "optim/total_tokens": 2572681216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 4.420321941375732, "created_at": "2025-01-15T02:49:48.350011+00:00"} {"global_step": 9815, "acc_step": 0, "speed/wps": 12962.401445342286, "speed/FLOPS": 203592326905272.3, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.740108847618103, "optim/lr": 0.0029436480202884524, "optim/total_tokens": 2572943360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474123, "loss/out": 4.390476226806641, "created_at": "2025-01-15T02:49:53.409383+00:00"} {"global_step": 9816, "acc_step": 0, "speed/wps": 12959.73751019984, "speed/FLOPS": 203550486143229.72, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.833999514579773, "optim/lr": 0.0029436247539279662, "optim/total_tokens": 2573205504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496563, "loss/out": 4.447112560272217, "created_at": "2025-01-15T02:49:58.474553+00:00"} {"global_step": 9817, "acc_step": 0, "speed/wps": 12959.55064224634, "speed/FLOPS": 203547551125236.5, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.2635393142700195, "optim/lr": 0.0029436014828574004, "optim/total_tokens": 2573467648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324561, "loss/out": 4.301883697509766, "created_at": "2025-01-15T02:50:03.538015+00:00"} {"global_step": 9818, "acc_step": 0, "speed/wps": 12966.575905734568, "speed/FLOPS": 203657892542042.78, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.48176008462905884, "optim/lr": 0.0029435782070768315, "optim/total_tokens": 2573729792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361759, "loss/out": 4.411169528961182, "created_at": "2025-01-15T02:50:08.598696+00:00"} {"global_step": 9819, "acc_step": 0, "speed/wps": 12962.719350292093, "speed/FLOPS": 203597320039356.22, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8810918927192688, "optim/lr": 0.0029435549265863347, "optim/total_tokens": 2573991936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430268, "loss/out": 4.414830207824707, "created_at": "2025-01-15T02:50:13.661843+00:00"} {"global_step": 9820, "acc_step": 0, "speed/wps": 12960.948614097804, "speed/FLOPS": 203569508194178.38, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.560178816318512, "optim/lr": 0.002943531641385986, "optim/total_tokens": 2574254080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 4.454631328582764, "created_at": "2025-01-15T02:50:18.723040+00:00"} {"global_step": 9821, "acc_step": 0, "speed/wps": 12966.24420999099, "speed/FLOPS": 203652682804592.2, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1598827838897705, "optim/lr": 0.0029435083514758624, "optim/total_tokens": 2574516224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435852, "loss/out": 4.433498382568359, "created_at": "2025-01-15T02:50:23.785580+00:00"} {"global_step": 9822, "acc_step": 0, "speed/wps": 12965.978046637541, "speed/FLOPS": 203648502343379.5, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7586285471916199, "optim/lr": 0.0029434850568560386, "optim/total_tokens": 2574778368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 4.393654823303223, "created_at": "2025-01-15T02:50:28.847753+00:00"} {"global_step": 9823, "acc_step": 0, "speed/wps": 12961.120241304567, "speed/FLOPS": 203572203835301.1, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.49626660346984863, "optim/lr": 0.0029434617575265914, "optim/total_tokens": 2575040512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451275, "loss/out": 4.550508499145508, "created_at": "2025-01-15T02:50:33.913355+00:00"} {"global_step": 9824, "acc_step": 0, "speed/wps": 12962.808796105608, "speed/FLOPS": 203598724908768.88, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6671860814094543, "optim/lr": 0.002943438453487596, "optim/total_tokens": 2575302656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356961, "loss/out": 4.389578819274902, "created_at": "2025-01-15T02:50:38.978837+00:00"} {"global_step": 9825, "acc_step": 0, "speed/wps": 12963.941675478438, "speed/FLOPS": 203616518336058.56, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8406968116760254, "optim/lr": 0.0029434151447391295, "optim/total_tokens": 2575564800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 4.384556770324707, "created_at": "2025-01-15T02:50:44.041722+00:00"} {"global_step": 9826, "acc_step": 0, "speed/wps": 12962.33054828476, "speed/FLOPS": 203591213369561.12, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2667455673217773, "optim/lr": 0.002943391831281267, "optim/total_tokens": 2575826944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 4.223355293273926, "created_at": "2025-01-15T02:50:49.105278+00:00"} {"global_step": 9827, "acc_step": 0, "speed/wps": 12967.193531157774, "speed/FLOPS": 203667593198020.53, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9452187418937683, "optim/lr": 0.002943368513114085, "optim/total_tokens": 2576089088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437195, "loss/out": 4.422084808349609, "created_at": "2025-01-15T02:50:54.168150+00:00"} {"global_step": 9828, "acc_step": 0, "speed/wps": 12967.437767059388, "speed/FLOPS": 203671429258470.66, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0784006118774414, "optim/lr": 0.00294334519023766, "optim/total_tokens": 2576351232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436096, "loss/out": 4.438409805297852, "created_at": "2025-01-15T02:50:59.226925+00:00"} {"global_step": 9829, "acc_step": 0, "speed/wps": 12962.975593947956, "speed/FLOPS": 203601344698087.22, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3902078866958618, "optim/lr": 0.002943321862652067, "optim/total_tokens": 2576613376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 4.304427146911621, "created_at": "2025-01-15T02:51:04.294671+00:00"} {"global_step": 9830, "acc_step": 0, "speed/wps": 12961.016822315905, "speed/FLOPS": 203570579497971.56, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.704058825969696, "optim/lr": 0.0029432985303573835, "optim/total_tokens": 2576875520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436388, "loss/out": 4.444232940673828, "created_at": "2025-01-15T02:51:09.354637+00:00"} {"global_step": 9831, "acc_step": 0, "speed/wps": 12969.754275378536, "speed/FLOPS": 203707813204838.9, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6695675849914551, "optim/lr": 0.0029432751933536846, "optim/total_tokens": 2577137664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307966, "loss/out": 4.386694431304932, "created_at": "2025-01-15T02:51:14.416854+00:00"} {"global_step": 9832, "acc_step": 0, "speed/wps": 12961.002412876125, "speed/FLOPS": 203570353177919.2, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.138105034828186, "optim/lr": 0.0029432518516410463, "optim/total_tokens": 2577399808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419791, "loss/out": 4.550069808959961, "created_at": "2025-01-15T02:51:19.483029+00:00"} {"global_step": 9833, "acc_step": 0, "speed/wps": 12958.88853046038, "speed/FLOPS": 203537151749801.06, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7986124753952026, "optim/lr": 0.002943228505219545, "optim/total_tokens": 2577661952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 4.510380744934082, "created_at": "2025-01-15T02:51:24.547717+00:00"} {"global_step": 9834, "acc_step": 0, "speed/wps": 12966.722796827902, "speed/FLOPS": 203660199668513.12, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1193032264709473, "optim/lr": 0.0029432051540892582, "optim/total_tokens": 2577924096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452847, "loss/out": 4.5669450759887695, "created_at": "2025-01-15T02:51:29.611090+00:00"} {"global_step": 9835, "acc_step": 0, "speed/wps": 12959.627461426333, "speed/FLOPS": 203548757676018.9, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7165536284446716, "optim/lr": 0.0029431817982502597, "optim/total_tokens": 2578186240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 4.334338188171387, "created_at": "2025-01-15T02:51:34.671600+00:00"} {"global_step": 9836, "acc_step": 0, "speed/wps": 12958.995714807017, "speed/FLOPS": 203538835227251.72, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.37580376863479614, "optim/lr": 0.002943158437702628, "optim/total_tokens": 2578448384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454995, "loss/out": 4.417649269104004, "created_at": "2025-01-15T02:51:39.763270+00:00"} {"global_step": 9837, "acc_step": 0, "speed/wps": 12960.157251203585, "speed/FLOPS": 203557078752478.1, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5839844942092896, "optim/lr": 0.0029431350724464377, "optim/total_tokens": 2578710528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457391, "loss/out": 4.525738716125488, "created_at": "2025-01-15T02:51:44.827874+00:00"} {"global_step": 9838, "acc_step": 0, "speed/wps": 12964.139640011255, "speed/FLOPS": 203619627641078.25, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7663723230361938, "optim/lr": 0.0029431117024817655, "optim/total_tokens": 2578972672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 4.402865886688232, "created_at": "2025-01-15T02:51:49.892053+00:00"} {"global_step": 9839, "acc_step": 0, "speed/wps": 12960.066992241147, "speed/FLOPS": 203555661111444.2, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5204067826271057, "optim/lr": 0.0029430883278086877, "optim/total_tokens": 2579234816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473310, "loss/out": 4.375044822692871, "created_at": "2025-01-15T02:51:54.955425+00:00"} {"global_step": 9840, "acc_step": 0, "speed/wps": 12962.30650621135, "speed/FLOPS": 203590835755760.9, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.43206003308296204, "optim/lr": 0.002943064948427281, "optim/total_tokens": 2579496960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462963, "loss/out": 4.420053482055664, "created_at": "2025-01-15T02:52:00.015235+00:00"} {"global_step": 9841, "acc_step": 0, "speed/wps": 12961.823833119905, "speed/FLOPS": 203583254711597.44, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0946406126022339, "optim/lr": 0.0029430415643376213, "optim/total_tokens": 2579759104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435866, "loss/out": 4.406518936157227, "created_at": "2025-01-15T02:52:05.075501+00:00"} {"global_step": 9842, "acc_step": 0, "speed/wps": 12959.325751014934, "speed/FLOPS": 203544018899414.34, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9020366668701172, "optim/lr": 0.0029430181755397853, "optim/total_tokens": 2580021248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441450, "loss/out": 4.556281566619873, "created_at": "2025-01-15T02:52:10.163364+00:00"} {"global_step": 9843, "acc_step": 0, "speed/wps": 12965.549990123869, "speed/FLOPS": 203641779127621.0, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2005094289779663, "optim/lr": 0.002942994782033848, "optim/total_tokens": 2580283392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429454, "loss/out": 4.377917289733887, "created_at": "2025-01-15T02:52:15.223473+00:00"} {"global_step": 9844, "acc_step": 0, "speed/wps": 12959.983339006676, "speed/FLOPS": 203554347222445.25, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9396175146102905, "optim/lr": 0.002942971383819887, "optim/total_tokens": 2580545536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 4.381953239440918, "created_at": "2025-01-15T02:52:20.286367+00:00"} {"global_step": 9845, "acc_step": 0, "speed/wps": 12961.391622239784, "speed/FLOPS": 203576466245805.16, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.038564682006836, "optim/lr": 0.0029429479808979783, "optim/total_tokens": 2580807680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444881, "loss/out": 4.423187732696533, "created_at": "2025-01-15T02:52:25.351119+00:00"} {"global_step": 9846, "acc_step": 0, "speed/wps": 12962.241548408554, "speed/FLOPS": 203589815504206.62, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5260443687438965, "optim/lr": 0.002942924573268198, "optim/total_tokens": 2581069824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482522, "loss/out": 4.466701507568359, "created_at": "2025-01-15T02:52:30.414563+00:00"} {"global_step": 9847, "acc_step": 0, "speed/wps": 12956.734222606654, "speed/FLOPS": 203503315384629.1, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9771596193313599, "optim/lr": 0.0029429011609306225, "optim/total_tokens": 2581331968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444359, "loss/out": 4.4831390380859375, "created_at": "2025-01-15T02:52:35.476554+00:00"} {"global_step": 9848, "acc_step": 0, "speed/wps": 12959.359722388934, "speed/FLOPS": 203544552466524.56, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2763385772705078, "optim/lr": 0.0029428777438853288, "optim/total_tokens": 2581594112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 343445, "loss/out": 4.429272651672363, "created_at": "2025-01-15T02:52:40.537239+00:00"} {"global_step": 9849, "acc_step": 0, "speed/wps": 12961.601381765478, "speed/FLOPS": 203579760807395.97, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0616800785064697, "optim/lr": 0.002942854322132393, "optim/total_tokens": 2581856256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460321, "loss/out": 4.387053489685059, "created_at": "2025-01-15T02:52:45.597207+00:00"} {"global_step": 9850, "acc_step": 0, "speed/wps": 12961.814839226263, "speed/FLOPS": 203583113450138.9, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.056427240371704, "optim/lr": 0.0029428308956718913, "optim/total_tokens": 2582118400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457380, "loss/out": 4.281970977783203, "created_at": "2025-01-15T02:52:50.656882+00:00"} {"global_step": 9851, "acc_step": 0, "speed/wps": 12960.784616042889, "speed/FLOPS": 203566932379369.0, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9977927207946777, "optim/lr": 0.0029428074645039, "optim/total_tokens": 2582380544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464508, "loss/out": 4.435310363769531, "created_at": "2025-01-15T02:52:55.722460+00:00"} {"global_step": 9852, "acc_step": 0, "speed/wps": 12962.87784774251, "speed/FLOPS": 203599809459613.3, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.155395269393921, "optim/lr": 0.0029427840286284958, "optim/total_tokens": 2582642688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447839, "loss/out": 4.476301193237305, "created_at": "2025-01-15T02:53:00.799571+00:00"} {"global_step": 9853, "acc_step": 0, "speed/wps": 12956.101887337554, "speed/FLOPS": 203493383690308.16, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9282046556472778, "optim/lr": 0.0029427605880457553, "optim/total_tokens": 2582904832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325239, "loss/out": 4.410811901092529, "created_at": "2025-01-15T02:53:05.861517+00:00"} {"global_step": 9854, "acc_step": 0, "speed/wps": 12957.093892199226, "speed/FLOPS": 203508964489818.25, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8962608575820923, "optim/lr": 0.0029427371427557542, "optim/total_tokens": 2583166976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301877, "loss/out": 4.353689193725586, "created_at": "2025-01-15T02:53:10.922923+00:00"} {"global_step": 9855, "acc_step": 0, "speed/wps": 12951.74870893001, "speed/FLOPS": 203425011041523.7, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.6780805587768555, "optim/lr": 0.0029427136927585706, "optim/total_tokens": 2583429120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332882, "loss/out": 4.396992206573486, "created_at": "2025-01-15T02:53:15.989062+00:00"} {"global_step": 9856, "acc_step": 0, "speed/wps": 12961.88318549074, "speed/FLOPS": 203584186922140.12, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9644030332565308, "optim/lr": 0.0029426902380542794, "optim/total_tokens": 2583691264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436610, "loss/out": 4.323566913604736, "created_at": "2025-01-15T02:53:21.049997+00:00"} {"global_step": 9857, "acc_step": 0, "speed/wps": 12961.265755459057, "speed/FLOPS": 203574489335054.75, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7762577533721924, "optim/lr": 0.002942666778642958, "optim/total_tokens": 2583953408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453945, "loss/out": 4.383309364318848, "created_at": "2025-01-15T02:53:26.110046+00:00"} {"global_step": 9858, "acc_step": 0, "speed/wps": 12954.475215204178, "speed/FLOPS": 203467834569171.75, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7903392314910889, "optim/lr": 0.0029426433145246828, "optim/total_tokens": 2584215552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445976, "loss/out": 4.482324123382568, "created_at": "2025-01-15T02:53:31.179237+00:00"} {"global_step": 9859, "acc_step": 0, "speed/wps": 12949.844832641698, "speed/FLOPS": 203395108048215.56, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7671613693237305, "optim/lr": 0.00294261984569953, "optim/total_tokens": 2584477696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 4.4050469398498535, "created_at": "2025-01-15T02:53:36.247547+00:00"} {"global_step": 9860, "acc_step": 0, "speed/wps": 12954.800711565758, "speed/FLOPS": 203472946936809.22, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.185398817062378, "optim/lr": 0.0029425963721675763, "optim/total_tokens": 2584739840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454481, "loss/out": 4.356876373291016, "created_at": "2025-01-15T02:53:41.336700+00:00"} {"global_step": 9861, "acc_step": 0, "speed/wps": 12962.307716789337, "speed/FLOPS": 203590854769551.72, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5120479464530945, "optim/lr": 0.0029425728939288984, "optim/total_tokens": 2585001984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444881, "loss/out": 4.448772430419922, "created_at": "2025-01-15T02:53:46.399302+00:00"} {"global_step": 9862, "acc_step": 0, "speed/wps": 12962.722714815407, "speed/FLOPS": 203597372883818.2, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0201481580734253, "optim/lr": 0.0029425494109835736, "optim/total_tokens": 2585264128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465841, "loss/out": 4.393882751464844, "created_at": "2025-01-15T02:53:51.460011+00:00"} {"global_step": 9863, "acc_step": 0, "speed/wps": 12962.763846194423, "speed/FLOPS": 203598018908646.06, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5109590888023376, "optim/lr": 0.002942525923331677, "optim/total_tokens": 2585526272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 4.390686988830566, "created_at": "2025-01-15T02:53:56.519208+00:00"} {"global_step": 9864, "acc_step": 0, "speed/wps": 12964.903376846672, "speed/FLOPS": 203631623177562.88, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.761009931564331, "optim/lr": 0.002942502430973286, "optim/total_tokens": 2585788416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 4.274524211883545, "created_at": "2025-01-15T02:54:01.577881+00:00"} {"global_step": 9865, "acc_step": 0, "speed/wps": 12967.586479493883, "speed/FLOPS": 203673764991606.7, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.2060136795043945, "optim/lr": 0.002942478933908478, "optim/total_tokens": 2586050560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 4.354579925537109, "created_at": "2025-01-15T02:54:06.636194+00:00"} {"global_step": 9866, "acc_step": 0, "speed/wps": 12964.587774829333, "speed/FLOPS": 203626666214197.94, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6368535757064819, "optim/lr": 0.002942455432137328, "optim/total_tokens": 2586312704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319464, "loss/out": 4.440009117126465, "created_at": "2025-01-15T02:54:11.695141+00:00"} {"global_step": 9867, "acc_step": 0, "speed/wps": 12957.88377625774, "speed/FLOPS": 203521370704371.66, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3949421644210815, "optim/lr": 0.0029424319256599147, "optim/total_tokens": 2586574848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348659, "loss/out": 4.438073635101318, "created_at": "2025-01-15T02:54:16.766870+00:00"} {"global_step": 9868, "acc_step": 0, "speed/wps": 12955.918974207478, "speed/FLOPS": 203490510788251.34, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.723820686340332, "optim/lr": 0.0029424084144763136, "optim/total_tokens": 2586836992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452340, "loss/out": 4.410589218139648, "created_at": "2025-01-15T02:54:21.834714+00:00"} {"global_step": 9869, "acc_step": 0, "speed/wps": 12960.07857588968, "speed/FLOPS": 203555843048562.0, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5933863520622253, "optim/lr": 0.002942384898586601, "optim/total_tokens": 2587099136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462963, "loss/out": 4.3451995849609375, "created_at": "2025-01-15T02:54:26.894960+00:00"} {"global_step": 9870, "acc_step": 0, "speed/wps": 12960.16677747144, "speed/FLOPS": 203557228375605.25, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1495146751403809, "optim/lr": 0.0029423613779908544, "optim/total_tokens": 2587361280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 4.468873023986816, "created_at": "2025-01-15T02:54:31.957149+00:00"} {"global_step": 9871, "acc_step": 0, "speed/wps": 12959.505376017463, "speed/FLOPS": 203546840156909.7, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8861417770385742, "optim/lr": 0.0029423378526891504, "optim/total_tokens": 2587623424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448083, "loss/out": 4.32095193862915, "created_at": "2025-01-15T02:54:37.021894+00:00"} {"global_step": 9872, "acc_step": 0, "speed/wps": 12958.808178563933, "speed/FLOPS": 203535889712852.3, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5654304027557373, "optim/lr": 0.0029423143226815656, "optim/total_tokens": 2587885568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453124, "loss/out": 4.413994312286377, "created_at": "2025-01-15T02:54:42.082674+00:00"} {"global_step": 9873, "acc_step": 0, "speed/wps": 12954.365439412468, "speed/FLOPS": 203466110389515.2, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8245363831520081, "optim/lr": 0.0029422907879681766, "optim/total_tokens": 2588147712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306907, "loss/out": 4.425204277038574, "created_at": "2025-01-15T02:54:47.145538+00:00"} {"global_step": 9874, "acc_step": 0, "speed/wps": 12964.527805327947, "speed/FLOPS": 203625724310771.94, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0382505655288696, "optim/lr": 0.0029422672485490613, "optim/total_tokens": 2588409856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354562, "loss/out": 4.453078269958496, "created_at": "2025-01-15T02:54:52.206653+00:00"} {"global_step": 9875, "acc_step": 0, "speed/wps": 12964.171604086827, "speed/FLOPS": 203620129680808.34, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5352337956428528, "optim/lr": 0.0029422437044242947, "optim/total_tokens": 2588672000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450227, "loss/out": 4.43787956237793, "created_at": "2025-01-15T02:54:57.270142+00:00"} {"global_step": 9876, "acc_step": 0, "speed/wps": 12963.90698746613, "speed/FLOPS": 203615973513158.97, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0994746685028076, "optim/lr": 0.002942220155593955, "optim/total_tokens": 2588934144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465320, "loss/out": 4.404280662536621, "created_at": "2025-01-15T02:55:02.333588+00:00"} {"global_step": 9877, "acc_step": 0, "speed/wps": 12960.874729833255, "speed/FLOPS": 203568347740276.0, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.001, "optim/grad_norm": 0.9829708933830261, "optim/lr": 0.0029421966020581184, "optim/total_tokens": 2589196288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465057, "loss/out": 4.450932502746582, "created_at": "2025-01-15T02:55:07.393581+00:00"} {"global_step": 9878, "acc_step": 0, "speed/wps": 12959.344908518615, "speed/FLOPS": 203544319794334.8, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7804807424545288, "optim/lr": 0.0029421730438168618, "optim/total_tokens": 2589458432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 4.293715000152588, "created_at": "2025-01-15T02:55:12.455343+00:00"} {"global_step": 9879, "acc_step": 0, "speed/wps": 12961.511207734271, "speed/FLOPS": 203578344500323.78, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.39534881711006165, "optim/lr": 0.0029421494808702623, "optim/total_tokens": 2589720576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320196, "loss/out": 4.459495544433594, "created_at": "2025-01-15T02:55:17.517003+00:00"} {"global_step": 9880, "acc_step": 0, "speed/wps": 12956.67796863173, "speed/FLOPS": 203502431838651.84, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0082951784133911, "optim/lr": 0.002942125913218397, "optim/total_tokens": 2589982720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345847, "loss/out": 4.241750717163086, "created_at": "2025-01-15T02:55:22.583423+00:00"} {"global_step": 9881, "acc_step": 0, "speed/wps": 12962.607763248263, "speed/FLOPS": 203595567411496.2, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9586324691772461, "optim/lr": 0.002942102340861342, "optim/total_tokens": 2590244864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457920, "loss/out": 4.420132637023926, "created_at": "2025-01-15T02:55:27.642860+00:00"} {"global_step": 9882, "acc_step": 0, "speed/wps": 12962.650660909767, "speed/FLOPS": 203596241178217.7, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5361658334732056, "optim/lr": 0.002942078763799175, "optim/total_tokens": 2590507008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444067, "loss/out": 4.2859601974487305, "created_at": "2025-01-15T02:55:32.705691+00:00"} {"global_step": 9883, "acc_step": 0, "speed/wps": 12968.226905292307, "speed/FLOPS": 203683823758808.25, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.34928378462791443, "optim/lr": 0.002942055182031972, "optim/total_tokens": 2590769152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479141, "loss/out": 4.4972052574157715, "created_at": "2025-01-15T02:55:37.766881+00:00"} {"global_step": 9884, "acc_step": 0, "speed/wps": 12961.298292654199, "speed/FLOPS": 203575000376415.25, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6112025380134583, "optim/lr": 0.0029420315955598114, "optim/total_tokens": 2591031296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437432, "loss/out": 4.442810535430908, "created_at": "2025-01-15T02:55:42.840525+00:00"} {"global_step": 9885, "acc_step": 0, "speed/wps": 12959.61205742033, "speed/FLOPS": 203548515734937.3, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0321416854858398, "optim/lr": 0.0029420080043827688, "optim/total_tokens": 2591293440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453945, "loss/out": 4.417365550994873, "created_at": "2025-01-15T02:55:47.902258+00:00"} {"global_step": 9886, "acc_step": 0, "speed/wps": 12957.616288211968, "speed/FLOPS": 203517169437045.72, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5075119733810425, "optim/lr": 0.002941984408500922, "optim/total_tokens": 2591555584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303928, "loss/out": 4.403454780578613, "created_at": "2025-01-15T02:55:52.968424+00:00"} {"global_step": 9887, "acc_step": 0, "speed/wps": 12965.607682862486, "speed/FLOPS": 203642685271359.1, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1946570873260498, "optim/lr": 0.002941960807914347, "optim/total_tokens": 2591817728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 382133, "loss/out": 4.351067543029785, "created_at": "2025-01-15T02:55:58.028739+00:00"} {"global_step": 9888, "acc_step": 0, "speed/wps": 12956.235261172704, "speed/FLOPS": 203495478509664.03, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4731528759002686, "optim/lr": 0.0029419372026231218, "optim/total_tokens": 2592079872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463248, "loss/out": 4.439557075500488, "created_at": "2025-01-15T02:56:03.094756+00:00"} {"global_step": 9889, "acc_step": 0, "speed/wps": 12961.238119622063, "speed/FLOPS": 203574055276259.16, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2732813358306885, "optim/lr": 0.0029419135926273226, "optim/total_tokens": 2592342016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443306, "loss/out": 4.488181114196777, "created_at": "2025-01-15T02:56:08.158972+00:00"} {"global_step": 9890, "acc_step": 0, "speed/wps": 12952.91204844089, "speed/FLOPS": 203443282887138.06, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1789941787719727, "optim/lr": 0.0029418899779270276, "optim/total_tokens": 2592604160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 4.364366054534912, "created_at": "2025-01-15T02:56:13.227141+00:00"} {"global_step": 9891, "acc_step": 0, "speed/wps": 12961.04381264833, "speed/FLOPS": 203571003418230.9, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3340741395950317, "optim/lr": 0.002941866358522313, "optim/total_tokens": 2592866304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326058, "loss/out": 4.376626968383789, "created_at": "2025-01-15T02:56:18.291009+00:00"} {"global_step": 9892, "acc_step": 0, "speed/wps": 12962.375320885003, "speed/FLOPS": 203591916584771.9, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.287216305732727, "optim/lr": 0.002941842734413256, "optim/total_tokens": 2593128448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345083, "loss/out": 4.329880714416504, "created_at": "2025-01-15T02:56:23.353446+00:00"} {"global_step": 9893, "acc_step": 0, "speed/wps": 12960.285769876358, "speed/FLOPS": 203559097314837.16, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2779935598373413, "optim/lr": 0.0029418191055999338, "optim/total_tokens": 2593390592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435338, "loss/out": 4.314022541046143, "created_at": "2025-01-15T02:56:28.413687+00:00"} {"global_step": 9894, "acc_step": 0, "speed/wps": 12965.764784106233, "speed/FLOPS": 203645152762271.88, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.652516782283783, "optim/lr": 0.0029417954720824227, "optim/total_tokens": 2593652736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436381, "loss/out": 4.4559149742126465, "created_at": "2025-01-15T02:56:33.475032+00:00"} {"global_step": 9895, "acc_step": 0, "speed/wps": 12959.885919568162, "speed/FLOPS": 203552817116295.03, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.309908151626587, "optim/lr": 0.002941771833860801, "optim/total_tokens": 2593914880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461913, "loss/out": 4.466020584106445, "created_at": "2025-01-15T02:56:38.538346+00:00"} {"global_step": 9896, "acc_step": 0, "speed/wps": 12958.539647703317, "speed/FLOPS": 203531672066686.25, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.1546324491500854, "optim/lr": 0.0029417481909351454, "optim/total_tokens": 2594177024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 4.296104907989502, "created_at": "2025-01-15T02:56:43.599405+00:00"} {"global_step": 9897, "acc_step": 0, "speed/wps": 12958.334638089967, "speed/FLOPS": 203528452108995.88, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0270916223526, "optim/lr": 0.0029417245433055327, "optim/total_tokens": 2594439168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432365, "loss/out": 4.418805122375488, "created_at": "2025-01-15T02:56:48.664163+00:00"} {"global_step": 9898, "acc_step": 0, "speed/wps": 12955.590137154932, "speed/FLOPS": 203485345950471.62, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.2158186435699463, "optim/lr": 0.0029417008909720407, "optim/total_tokens": 2594701312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 4.348015785217285, "created_at": "2025-01-15T02:56:53.727929+00:00"} {"global_step": 9899, "acc_step": 0, "speed/wps": 12963.70615220474, "speed/FLOPS": 203612819119402.28, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5064042806625366, "optim/lr": 0.002941677233934745, "optim/total_tokens": 2594963456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448077, "loss/out": 4.373625755310059, "created_at": "2025-01-15T02:56:58.788502+00:00"} {"global_step": 9900, "acc_step": 0, "speed/wps": 12960.866692135472, "speed/FLOPS": 203568221497187.4, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.542148232460022, "optim/lr": 0.002941653572193726, "optim/total_tokens": 2595225600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299996, "loss/out": 4.341588973999023, "created_at": "2025-01-15T02:57:03.852407+00:00"} {"global_step": 9901, "acc_step": 0, "speed/wps": 12957.256231737234, "speed/FLOPS": 203511514255340.56, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7265362739562988, "optim/lr": 0.0029416299057490568, "optim/total_tokens": 2595487744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448890, "loss/out": 4.375190734863281, "created_at": "2025-01-15T02:57:08.916394+00:00"} {"global_step": 9902, "acc_step": 0, "speed/wps": 12965.458913087763, "speed/FLOPS": 203640348637615.97, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6239196062088013, "optim/lr": 0.0029416062346008178, "optim/total_tokens": 2595749888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 4.4497246742248535, "created_at": "2025-01-15T02:57:13.979138+00:00"} {"global_step": 9903, "acc_step": 0, "speed/wps": 12955.346118359144, "speed/FLOPS": 203481513299966.84, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0172494649887085, "optim/lr": 0.0029415825587490846, "optim/total_tokens": 2596012032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427066, "loss/out": 4.386899471282959, "created_at": "2025-01-15T02:57:19.045013+00:00"} {"global_step": 9904, "acc_step": 0, "speed/wps": 12962.518009794327, "speed/FLOPS": 203594157710167.94, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4698149561882019, "optim/lr": 0.002941558878193935, "optim/total_tokens": 2596274176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 4.348273277282715, "created_at": "2025-01-15T02:57:24.104592+00:00"} {"global_step": 9905, "acc_step": 0, "speed/wps": 12963.409006627911, "speed/FLOPS": 203608152039797.34, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5292445421218872, "optim/lr": 0.002941535192935447, "optim/total_tokens": 2596536320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446742, "loss/out": 4.434234142303467, "created_at": "2025-01-15T02:57:29.166193+00:00"} {"global_step": 9906, "acc_step": 0, "speed/wps": 12957.739290062831, "speed/FLOPS": 203519101350135.78, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3635791540145874, "optim/lr": 0.0029415115029736965, "optim/total_tokens": 2596798464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293250, "loss/out": 4.35959529876709, "created_at": "2025-01-15T02:57:34.227254+00:00"} {"global_step": 9907, "acc_step": 0, "speed/wps": 12956.224066071365, "speed/FLOPS": 203495302675213.84, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9352173209190369, "optim/lr": 0.002941487808308761, "optim/total_tokens": 2597060608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356035, "loss/out": 4.37136697769165, "created_at": "2025-01-15T02:57:39.291961+00:00"} {"global_step": 9908, "acc_step": 0, "speed/wps": 12959.360927643285, "speed/FLOPS": 203544571396700.34, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1758116483688354, "optim/lr": 0.002941464108940719, "optim/total_tokens": 2597322752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415616, "loss/out": 4.338245391845703, "created_at": "2025-01-15T02:57:44.362153+00:00"} {"global_step": 9909, "acc_step": 0, "speed/wps": 12961.04273599308, "speed/FLOPS": 203570986507880.72, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.041345238685608, "optim/lr": 0.0029414404048696465, "optim/total_tokens": 2597584896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 4.386871337890625, "created_at": "2025-01-15T02:57:49.424737+00:00"} {"global_step": 9910, "acc_step": 0, "speed/wps": 12951.319691620494, "speed/FLOPS": 203418272735146.2, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0947084426879883, "optim/lr": 0.0029414166960956213, "optim/total_tokens": 2597847040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429454, "loss/out": 4.378453254699707, "created_at": "2025-01-15T02:57:54.493106+00:00"} {"global_step": 9911, "acc_step": 0, "speed/wps": 12955.128280226578, "speed/FLOPS": 203478091852753.9, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5843754410743713, "optim/lr": 0.002941392982618721, "optim/total_tokens": 2598109184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 4.3320770263671875, "created_at": "2025-01-15T02:57:59.560248+00:00"} {"global_step": 9912, "acc_step": 0, "speed/wps": 12964.741824969544, "speed/FLOPS": 203629085783337.03, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0588805675506592, "optim/lr": 0.002941369264439023, "optim/total_tokens": 2598371328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326870, "loss/out": 4.295610427856445, "created_at": "2025-01-15T02:58:04.623155+00:00"} {"global_step": 9913, "acc_step": 0, "speed/wps": 12966.006192454699, "speed/FLOPS": 203648944412114.38, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8546446561813354, "optim/lr": 0.002941345541556604, "optim/total_tokens": 2598633472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345608, "loss/out": 4.35245943069458, "created_at": "2025-01-15T02:58:09.681190+00:00"} {"global_step": 9914, "acc_step": 0, "speed/wps": 12959.373357289356, "speed/FLOPS": 203544766621372.03, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.6680451035499573, "optim/lr": 0.002941321813971542, "optim/total_tokens": 2598895616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452859, "loss/out": 4.435263156890869, "created_at": "2025-01-15T02:58:14.746386+00:00"} {"global_step": 9915, "acc_step": 0, "speed/wps": 12957.986734838161, "speed/FLOPS": 203522987810357.1, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0512326955795288, "optim/lr": 0.002941298081683914, "optim/total_tokens": 2599157760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 4.3975911140441895, "created_at": "2025-01-15T02:58:19.813897+00:00"} {"global_step": 9916, "acc_step": 0, "speed/wps": 12965.125067267965, "speed/FLOPS": 203635105130263.84, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0750843286514282, "optim/lr": 0.002941274344693798, "optim/total_tokens": 2599419904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 4.3398566246032715, "created_at": "2025-01-15T02:58:24.874146+00:00"} {"global_step": 9917, "acc_step": 0, "speed/wps": 12956.132912473877, "speed/FLOPS": 203493870982707.12, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1558316946029663, "optim/lr": 0.0029412506030012716, "optim/total_tokens": 2599682048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444357, "loss/out": 4.489160537719727, "created_at": "2025-01-15T02:58:29.962199+00:00"} {"global_step": 9918, "acc_step": 0, "speed/wps": 12967.81723282976, "speed/FLOPS": 203677389289834.97, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.193529486656189, "optim/lr": 0.002941226856606411, "optim/total_tokens": 2599944192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449135, "loss/out": 4.3981852531433105, "created_at": "2025-01-15T02:58:35.019740+00:00"} {"global_step": 9919, "acc_step": 0, "speed/wps": 12964.565241117232, "speed/FLOPS": 203626312291284.8, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 5.869935035705566, "optim/lr": 0.002941203105509295, "optim/total_tokens": 2600206336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470641, "loss/out": 4.371214389801025, "created_at": "2025-01-15T02:58:40.078357+00:00"} {"global_step": 9920, "acc_step": 0, "speed/wps": 12961.46019800791, "speed/FLOPS": 203577543322476.56, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.922662615776062, "optim/lr": 0.0029411793497100005, "optim/total_tokens": 2600468480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 4.411249160766602, "created_at": "2025-01-15T02:58:45.141213+00:00"} {"global_step": 9921, "acc_step": 0, "speed/wps": 12958.151790765984, "speed/FLOPS": 203525580240514.16, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.001, "optim/grad_norm": 1.2445063591003418, "optim/lr": 0.002941155589208605, "optim/total_tokens": 2600730624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448605, "loss/out": 4.445533275604248, "created_at": "2025-01-15T02:58:50.211600+00:00"} {"global_step": 9922, "acc_step": 0, "speed/wps": 12964.951685779439, "speed/FLOPS": 203632381935735.38, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8716971278190613, "optim/lr": 0.0029411318240051866, "optim/total_tokens": 2600992768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 4.348245143890381, "created_at": "2025-01-15T02:58:55.272649+00:00"} {"global_step": 9923, "acc_step": 0, "speed/wps": 12964.706621512181, "speed/FLOPS": 203628532864664.53, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6425663232803345, "optim/lr": 0.002941108054099822, "optim/total_tokens": 2601254912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 4.4071044921875, "created_at": "2025-01-15T02:59:00.336399+00:00"} {"global_step": 9924, "acc_step": 0, "speed/wps": 12960.575143144606, "speed/FLOPS": 203563642319655.53, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6595219373703003, "optim/lr": 0.002941084279492589, "optim/total_tokens": 2601517056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311191, "loss/out": 4.396797180175781, "created_at": "2025-01-15T02:59:05.401018+00:00"} {"global_step": 9925, "acc_step": 0, "speed/wps": 12965.392834363582, "speed/FLOPS": 203639310780451.03, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.00235915184021, "optim/lr": 0.0029410605001835653, "optim/total_tokens": 2601779200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416344, "loss/out": 4.362646102905273, "created_at": "2025-01-15T02:59:10.465775+00:00"} {"global_step": 9926, "acc_step": 0, "speed/wps": 12960.802951897893, "speed/FLOPS": 203567220369167.94, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1282010078430176, "optim/lr": 0.0029410367161728287, "optim/total_tokens": 2602041344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 4.312504291534424, "created_at": "2025-01-15T02:59:15.555266+00:00"} {"global_step": 9927, "acc_step": 0, "speed/wps": 12961.683747540279, "speed/FLOPS": 203581054475071.66, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4525151252746582, "optim/lr": 0.0029410129274604563, "optim/total_tokens": 2602303488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453660, "loss/out": 4.375069618225098, "created_at": "2025-01-15T02:59:20.619187+00:00"} {"global_step": 9928, "acc_step": 0, "speed/wps": 12958.327787128192, "speed/FLOPS": 203528344505226.97, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6640117168426514, "optim/lr": 0.0029409891340465254, "optim/total_tokens": 2602565632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289017, "loss/out": 4.396852493286133, "created_at": "2025-01-15T02:59:25.685788+00:00"} {"global_step": 9929, "acc_step": 0, "speed/wps": 12959.690240170403, "speed/FLOPS": 203549743702461.56, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5791724920272827, "optim/lr": 0.002940965335931115, "optim/total_tokens": 2602827776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439586, "loss/out": 4.332003593444824, "created_at": "2025-01-15T02:59:30.748262+00:00"} {"global_step": 9930, "acc_step": 0, "speed/wps": 12961.767425115155, "speed/FLOPS": 203582368746370.12, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.196565866470337, "optim/lr": 0.002940941533114302, "optim/total_tokens": 2603089920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 4.391486167907715, "created_at": "2025-01-15T02:59:35.809157+00:00"} {"global_step": 9931, "acc_step": 0, "speed/wps": 12959.193019972234, "speed/FLOPS": 203541934175997.16, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1508729457855225, "optim/lr": 0.002940917725596164, "optim/total_tokens": 2603352064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 4.439603805541992, "created_at": "2025-01-15T02:59:40.871081+00:00"} {"global_step": 9932, "acc_step": 0, "speed/wps": 12955.587313015025, "speed/FLOPS": 203485301593473.62, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.8580099940299988, "optim/lr": 0.0029408939133767783, "optim/total_tokens": 2603614208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 4.507192611694336, "created_at": "2025-01-15T02:59:45.936713+00:00"} {"global_step": 9933, "acc_step": 0, "speed/wps": 12952.192013367792, "speed/FLOPS": 203431973746883.2, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0023, "optim/grad_norm": 1.0184448957443237, "optim/lr": 0.002940870096456223, "optim/total_tokens": 2603876352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456851, "loss/out": 4.422364711761475, "created_at": "2025-01-15T02:59:51.006172+00:00"} {"global_step": 9934, "acc_step": 0, "speed/wps": 12963.760336893722, "speed/FLOPS": 203613670164400.28, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5758772492408752, "optim/lr": 0.002940846274834576, "optim/total_tokens": 2604138496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302630, "loss/out": 4.444586753845215, "created_at": "2025-01-15T02:59:56.072821+00:00"} {"global_step": 9935, "acc_step": 0, "speed/wps": 12966.587828356094, "speed/FLOPS": 203658079803197.28, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.629184365272522, "optim/lr": 0.002940822448511915, "optim/total_tokens": 2604400640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335427, "loss/out": 4.421247482299805, "created_at": "2025-01-15T03:00:01.130786+00:00"} {"global_step": 9936, "acc_step": 0, "speed/wps": 12952.963985154522, "speed/FLOPS": 203444098624594.2, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.001631259918213, "optim/lr": 0.0029407986174883175, "optim/total_tokens": 2604662784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446220, "loss/out": 4.2533063888549805, "created_at": "2025-01-15T03:00:06.195428+00:00"} {"global_step": 9937, "acc_step": 0, "speed/wps": 12959.865652976443, "speed/FLOPS": 203552498801622.88, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1599568128585815, "optim/lr": 0.0029407747817638604, "optim/total_tokens": 2604924928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 4.325954914093018, "created_at": "2025-01-15T03:00:11.260201+00:00"} {"global_step": 9938, "acc_step": 0, "speed/wps": 12960.564716344961, "speed/FLOPS": 203563478552439.2, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5910964608192444, "optim/lr": 0.0029407509413386236, "optim/total_tokens": 2605187072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451797, "loss/out": 4.3544111251831055, "created_at": "2025-01-15T03:00:16.321207+00:00"} {"global_step": 9939, "acc_step": 0, "speed/wps": 12953.602969803858, "speed/FLOPS": 203454134756568.94, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5117460489273071, "optim/lr": 0.0029407270962126834, "optim/total_tokens": 2605449216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 4.3647685050964355, "created_at": "2025-01-15T03:00:21.387896+00:00"} {"global_step": 9940, "acc_step": 0, "speed/wps": 12959.869512479132, "speed/FLOPS": 203552559420416.12, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.9075135588645935, "optim/lr": 0.002940703246386117, "optim/total_tokens": 2605711360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442208, "loss/out": 4.39423942565918, "created_at": "2025-01-15T03:00:26.449033+00:00"} {"global_step": 9941, "acc_step": 0, "speed/wps": 12952.331297840874, "speed/FLOPS": 203434161400929.88, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0969102382659912, "optim/lr": 0.002940679391859004, "optim/total_tokens": 2605973504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350126, "loss/out": 4.33682918548584, "created_at": "2025-01-15T03:00:31.512644+00:00"} {"global_step": 9942, "acc_step": 0, "speed/wps": 12957.853976342634, "speed/FLOPS": 203520902655755.53, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6371727585792542, "optim/lr": 0.0029406555326314207, "optim/total_tokens": 2606235648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447839, "loss/out": 4.4089460372924805, "created_at": "2025-01-15T03:00:36.578308+00:00"} {"global_step": 9943, "acc_step": 0, "speed/wps": 12956.166759764625, "speed/FLOPS": 203494402600919.38, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2272905111312866, "optim/lr": 0.002940631668703446, "optim/total_tokens": 2606497792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458410, "loss/out": 4.360795021057129, "created_at": "2025-01-15T03:00:41.644604+00:00"} {"global_step": 9944, "acc_step": 0, "speed/wps": 12963.864262999956, "speed/FLOPS": 203615302466708.12, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.48589637875556946, "optim/lr": 0.002940607800075157, "optim/total_tokens": 2606759936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466392, "loss/out": 4.388236045837402, "created_at": "2025-01-15T03:00:46.708261+00:00"} {"global_step": 9945, "acc_step": 0, "speed/wps": 12957.517620707486, "speed/FLOPS": 203515619728304.44, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6006050705909729, "optim/lr": 0.002940583926746632, "optim/total_tokens": 2607022080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456567, "loss/out": 4.460268974304199, "created_at": "2025-01-15T03:00:51.770821+00:00"} {"global_step": 9946, "acc_step": 0, "speed/wps": 12957.263701892207, "speed/FLOPS": 203511631584389.47, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.8581532835960388, "optim/lr": 0.0029405600487179486, "optim/total_tokens": 2607284224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309845, "loss/out": 4.374258995056152, "created_at": "2025-01-15T03:00:56.835558+00:00"} {"global_step": 9947, "acc_step": 0, "speed/wps": 12962.926065167065, "speed/FLOPS": 203600566780523.66, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1281777620315552, "optim/lr": 0.002940536165989185, "optim/total_tokens": 2607546368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354691, "loss/out": 4.4498515129089355, "created_at": "2025-01-15T03:01:01.900769+00:00"} {"global_step": 9948, "acc_step": 0, "speed/wps": 12962.395074794964, "speed/FLOPS": 203592226847072.56, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.705643355846405, "optim/lr": 0.0029405122785604192, "optim/total_tokens": 2607808512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 400828, "loss/out": 4.392489433288574, "created_at": "2025-01-15T03:01:06.961500+00:00"} {"global_step": 9949, "acc_step": 0, "speed/wps": 12962.845250082095, "speed/FLOPS": 203599297468562.62, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.074249029159546, "optim/lr": 0.0029404883864317284, "optim/total_tokens": 2608070656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435567, "loss/out": 4.419674873352051, "created_at": "2025-01-15T03:01:12.022218+00:00"} {"global_step": 9950, "acc_step": 0, "speed/wps": 12959.355512353712, "speed/FLOPS": 203544486342136.22, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0493417978286743, "optim/lr": 0.0029404644896031916, "optim/total_tokens": 2608332800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 4.418173789978027, "created_at": "2025-01-15T03:01:17.082856+00:00"} {"global_step": 9951, "acc_step": 0, "speed/wps": 12962.278994872773, "speed/FLOPS": 203590403652384.03, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5214099287986755, "optim/lr": 0.002940440588074886, "optim/total_tokens": 2608594944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429438, "loss/out": 4.378477573394775, "created_at": "2025-01-15T03:01:22.146297+00:00"} {"global_step": 9952, "acc_step": 0, "speed/wps": 12953.186934124215, "speed/FLOPS": 203447600344527.2, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7442745566368103, "optim/lr": 0.00294041668184689, "optim/total_tokens": 2608857088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453375, "loss/out": 4.428183078765869, "created_at": "2025-01-15T03:01:27.224402+00:00"} {"global_step": 9953, "acc_step": 0, "speed/wps": 12957.226997770576, "speed/FLOPS": 203511055095722.53, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8962256908416748, "optim/lr": 0.0029403927709192815, "optim/total_tokens": 2609119232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350126, "loss/out": 4.480392932891846, "created_at": "2025-01-15T03:01:32.286107+00:00"} {"global_step": 9954, "acc_step": 0, "speed/wps": 12952.248075818561, "speed/FLOPS": 203432854284710.4, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9302362203598022, "optim/lr": 0.0029403688552921384, "optim/total_tokens": 2609381376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460815, "loss/out": 4.322231292724609, "created_at": "2025-01-15T03:01:37.349890+00:00"} {"global_step": 9955, "acc_step": 0, "speed/wps": 12964.91781884848, "speed/FLOPS": 203631850009046.66, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2033476829528809, "optim/lr": 0.0029403449349655383, "optim/total_tokens": 2609643520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 4.372183799743652, "created_at": "2025-01-15T03:01:42.411961+00:00"} {"global_step": 9956, "acc_step": 0, "speed/wps": 12958.274108387239, "speed/FLOPS": 203527501406837.97, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.284348487854004, "optim/lr": 0.00294032100993956, "optim/total_tokens": 2609905664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430506, "loss/out": 4.351484775543213, "created_at": "2025-01-15T03:01:47.473746+00:00"} {"global_step": 9957, "acc_step": 0, "speed/wps": 12942.486304436012, "speed/FLOPS": 203279532251067.88, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8977013230323792, "optim/lr": 0.002940297080214282, "optim/total_tokens": 2610167808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498415, "loss/out": 4.404407978057861, "created_at": "2025-01-15T03:01:52.541176+00:00"} {"global_step": 9958, "acc_step": 0, "speed/wps": 12956.4741486833, "speed/FLOPS": 203499230566283.8, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5664403438568115, "optim/lr": 0.0029402731457897806, "optim/total_tokens": 2610429952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444067, "loss/out": 4.341362953186035, "created_at": "2025-01-15T03:01:57.603563+00:00"} {"global_step": 9959, "acc_step": 0, "speed/wps": 12958.47385187059, "speed/FLOPS": 203530638652720.34, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7224808931350708, "optim/lr": 0.002940249206666135, "optim/total_tokens": 2610692096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306907, "loss/out": 4.473730564117432, "created_at": "2025-01-15T03:02:02.665924+00:00"} {"global_step": 9960, "acc_step": 0, "speed/wps": 12963.139005073319, "speed/FLOPS": 203603911294361.47, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2301745414733887, "optim/lr": 0.002940225262843424, "optim/total_tokens": 2610954240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348659, "loss/out": 4.353012561798096, "created_at": "2025-01-15T03:02:07.730029+00:00"} {"global_step": 9961, "acc_step": 0, "speed/wps": 12959.356075600896, "speed/FLOPS": 203544495188707.22, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6657255291938782, "optim/lr": 0.0029402013143217244, "optim/total_tokens": 2611216384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359599, "loss/out": 4.430622577667236, "created_at": "2025-01-15T03:02:12.792638+00:00"} {"global_step": 9962, "acc_step": 0, "speed/wps": 12963.059058954963, "speed/FLOPS": 203602655630715.2, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.5630014538764954, "optim/lr": 0.002940177361101115, "optim/total_tokens": 2611478528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443835, "loss/out": 4.279486656188965, "created_at": "2025-01-15T03:02:17.857737+00:00"} {"global_step": 9963, "acc_step": 0, "speed/wps": 12962.467870848739, "speed/FLOPS": 203593370209129.38, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0673056840896606, "optim/lr": 0.0029401534031816743, "optim/total_tokens": 2611740672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489424, "loss/out": 4.418583393096924, "created_at": "2025-01-15T03:02:22.924675+00:00"} {"global_step": 9964, "acc_step": 0, "speed/wps": 12961.33499358405, "speed/FLOPS": 203575576814950.94, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2019996643066406, "optim/lr": 0.0029401294405634795, "optim/total_tokens": 2612002816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431605, "loss/out": 4.4831743240356445, "created_at": "2025-01-15T03:02:27.988364+00:00"} {"global_step": 9965, "acc_step": 0, "speed/wps": 12962.88874152507, "speed/FLOPS": 203599980561438.06, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.6812880039215088, "optim/lr": 0.002940105473246609, "optim/total_tokens": 2612264960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 4.374149799346924, "created_at": "2025-01-15T03:02:33.047659+00:00"} {"global_step": 9966, "acc_step": 0, "speed/wps": 12960.538193487686, "speed/FLOPS": 203563061974519.47, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2053714990615845, "optim/lr": 0.0029400815012311427, "optim/total_tokens": 2612527104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445933, "loss/out": 4.4267802238464355, "created_at": "2025-01-15T03:02:38.108377+00:00"} {"global_step": 9967, "acc_step": 0, "speed/wps": 12957.748168543274, "speed/FLOPS": 203519240798871.44, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1378589868545532, "optim/lr": 0.002940057524517156, "optim/total_tokens": 2612789248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355618, "loss/out": 4.387555122375488, "created_at": "2025-01-15T03:02:43.170781+00:00"} {"global_step": 9968, "acc_step": 0, "speed/wps": 12959.986883512207, "speed/FLOPS": 203554402893775.28, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1205281019210815, "optim/lr": 0.0029400335431047286, "optim/total_tokens": 2613051392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445976, "loss/out": 4.408293724060059, "created_at": "2025-01-15T03:02:48.235957+00:00"} {"global_step": 9969, "acc_step": 0, "speed/wps": 12963.360709190847, "speed/FLOPS": 203607393462180.62, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1015021800994873, "optim/lr": 0.0029400095569939396, "optim/total_tokens": 2613313536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 4.2434186935424805, "created_at": "2025-01-15T03:02:53.299319+00:00"} {"global_step": 9970, "acc_step": 0, "speed/wps": 12968.20199292667, "speed/FLOPS": 203683432475872.72, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9071289896965027, "optim/lr": 0.002939985566184866, "optim/total_tokens": 2613575680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430791, "loss/out": 4.219222068786621, "created_at": "2025-01-15T03:02:58.361819+00:00"} {"global_step": 9971, "acc_step": 0, "speed/wps": 12961.903193371703, "speed/FLOPS": 203584501173404.34, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0572044849395752, "optim/lr": 0.0029399615706775864, "optim/total_tokens": 2613837824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 4.4327874183654785, "created_at": "2025-01-15T03:03:03.421728+00:00"} {"global_step": 9972, "acc_step": 0, "speed/wps": 12961.716776273657, "speed/FLOPS": 203581573236715.6, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6812001466751099, "optim/lr": 0.002939937570472179, "optim/total_tokens": 2614099968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 4.337757587432861, "created_at": "2025-01-15T03:03:08.482684+00:00"} {"global_step": 9973, "acc_step": 0, "speed/wps": 12961.924256627686, "speed/FLOPS": 203584832000783.5, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5491554141044617, "optim/lr": 0.0029399135655687222, "optim/total_tokens": 2614362112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436381, "loss/out": 4.320508003234863, "created_at": "2025-01-15T03:03:13.573792+00:00"} {"global_step": 9974, "acc_step": 0, "speed/wps": 12963.023554196567, "speed/FLOPS": 203602097979695.94, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0691745281219482, "optim/lr": 0.0029398895559672946, "optim/total_tokens": 2614624256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 4.370941162109375, "created_at": "2025-01-15T03:03:18.632959+00:00"} {"global_step": 9975, "acc_step": 0, "speed/wps": 12962.32823695632, "speed/FLOPS": 203591177066971.84, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7325437664985657, "optim/lr": 0.0029398655416679738, "optim/total_tokens": 2614886400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463795, "loss/out": 4.245721340179443, "created_at": "2025-01-15T03:03:23.692539+00:00"} {"global_step": 9976, "acc_step": 0, "speed/wps": 12963.1967026262, "speed/FLOPS": 203604817513714.2, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7377788424491882, "optim/lr": 0.0029398415226708392, "optim/total_tokens": 2615148544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 4.326504707336426, "created_at": "2025-01-15T03:03:28.752315+00:00"} {"global_step": 9977, "acc_step": 0, "speed/wps": 12964.013693581755, "speed/FLOPS": 203617649479334.44, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.132831335067749, "optim/lr": 0.002939817498975968, "optim/total_tokens": 2615410688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451026, "loss/out": 4.308183670043945, "created_at": "2025-01-15T03:03:33.813868+00:00"} {"global_step": 9978, "acc_step": 0, "speed/wps": 12962.93500567982, "speed/FLOPS": 203600707203562.0, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9129763841629028, "optim/lr": 0.00293979347058344, "optim/total_tokens": 2615672832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444067, "loss/out": 4.387517929077148, "created_at": "2025-01-15T03:03:38.875807+00:00"} {"global_step": 9979, "acc_step": 0, "speed/wps": 12957.940016842489, "speed/FLOPS": 203522254040037.78, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5899202823638916, "optim/lr": 0.002939769437493332, "optim/total_tokens": 2615934976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 4.371169090270996, "created_at": "2025-01-15T03:03:43.941617+00:00"} {"global_step": 9980, "acc_step": 0, "speed/wps": 12955.342833989193, "speed/FLOPS": 203481461714423.53, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.089414119720459, "optim/lr": 0.0029397453997057235, "optim/total_tokens": 2616197120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299938, "loss/out": 4.648571014404297, "created_at": "2025-01-15T03:03:49.009074+00:00"} {"global_step": 9981, "acc_step": 0, "speed/wps": 12964.014902090283, "speed/FLOPS": 203617668460621.5, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5516676902770996, "optim/lr": 0.0029397213572206924, "optim/total_tokens": 2616459264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413426, "loss/out": 4.282477378845215, "created_at": "2025-01-15T03:03:54.075029+00:00"} {"global_step": 9982, "acc_step": 0, "speed/wps": 12964.680408938491, "speed/FLOPS": 203628121160174.88, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7680626511573792, "optim/lr": 0.0029396973100383173, "optim/total_tokens": 2616721408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459528, "loss/out": 4.311849594116211, "created_at": "2025-01-15T03:03:59.133915+00:00"} {"global_step": 9983, "acc_step": 0, "speed/wps": 12960.41834666414, "speed/FLOPS": 203561179615468.88, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4223411083221436, "optim/lr": 0.002939673258158677, "optim/total_tokens": 2616983552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 4.318887233734131, "created_at": "2025-01-15T03:04:04.198300+00:00"} {"global_step": 9984, "acc_step": 0, "speed/wps": 12962.370337646587, "speed/FLOPS": 203591838316164.9, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8359857201576233, "optim/lr": 0.0029396492015818496, "optim/total_tokens": 2617245696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 4.440650939941406, "created_at": "2025-01-15T03:04:09.261324+00:00"} {"global_step": 9985, "acc_step": 0, "speed/wps": 12960.955369954658, "speed/FLOPS": 203569614304193.78, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0310322046279907, "optim/lr": 0.0029396251403079137, "optim/total_tokens": 2617507840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 4.427243709564209, "created_at": "2025-01-15T03:04:14.323206+00:00"} {"global_step": 9986, "acc_step": 0, "speed/wps": 12956.348928448335, "speed/FLOPS": 203497263810423.03, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6684876680374146, "optim/lr": 0.0029396010743369476, "optim/total_tokens": 2617769984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323942, "loss/out": 4.309048652648926, "created_at": "2025-01-15T03:04:19.389440+00:00"} {"global_step": 9987, "acc_step": 0, "speed/wps": 12962.628289140936, "speed/FLOPS": 203595889798846.16, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.47189006209373474, "optim/lr": 0.0029395770036690304, "optim/total_tokens": 2618032128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296535, "loss/out": 4.412741661071777, "created_at": "2025-01-15T03:04:24.452038+00:00"} {"global_step": 9988, "acc_step": 0, "speed/wps": 12961.956436601644, "speed/FLOPS": 203585337431494.44, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5638740658760071, "optim/lr": 0.0029395529283042397, "optim/total_tokens": 2618294272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 280980, "loss/out": 4.256523132324219, "created_at": "2025-01-15T03:04:29.516469+00:00"} {"global_step": 9989, "acc_step": 0, "speed/wps": 12958.247912273177, "speed/FLOPS": 203527089960869.4, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6903361678123474, "optim/lr": 0.002939528848242655, "optim/total_tokens": 2618556416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356674, "loss/out": 4.4571332931518555, "created_at": "2025-01-15T03:04:34.581006+00:00"} {"global_step": 9990, "acc_step": 0, "speed/wps": 12957.502312383049, "speed/FLOPS": 203515379290033.34, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.001, "optim/grad_norm": 1.2438006401062012, "optim/lr": 0.002939504763484354, "optim/total_tokens": 2618818560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439820, "loss/out": 4.269911289215088, "created_at": "2025-01-15T03:04:39.642650+00:00"} {"global_step": 9991, "acc_step": 0, "speed/wps": 12963.776768165357, "speed/FLOPS": 203613928240100.25, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6379173994064331, "optim/lr": 0.0029394806740294164, "optim/total_tokens": 2619080704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 4.373415946960449, "created_at": "2025-01-15T03:04:44.704297+00:00"} {"global_step": 9992, "acc_step": 0, "speed/wps": 12958.970594472652, "speed/FLOPS": 203538440677881.56, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8688126802444458, "optim/lr": 0.0029394565798779196, "optim/total_tokens": 2619342848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432146, "loss/out": 4.2696146965026855, "created_at": "2025-01-15T03:04:49.774957+00:00"} {"global_step": 9993, "acc_step": 0, "speed/wps": 12955.816267934384, "speed/FLOPS": 203488897645099.2, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0068494081497192, "optim/lr": 0.002939432481029943, "optim/total_tokens": 2619604992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452325, "loss/out": 4.408670425415039, "created_at": "2025-01-15T03:04:54.839793+00:00"} {"global_step": 9994, "acc_step": 0, "speed/wps": 12956.047652149375, "speed/FLOPS": 203492531852150.84, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1197924613952637, "optim/lr": 0.0029394083774855646, "optim/total_tokens": 2619867136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290611, "loss/out": 4.416743755340576, "created_at": "2025-01-15T03:04:59.910203+00:00"} {"global_step": 9995, "acc_step": 0, "speed/wps": 12959.628633316112, "speed/FLOPS": 203548776082158.22, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.772914409637451, "optim/lr": 0.002939384269244864, "optim/total_tokens": 2620129280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438769, "loss/out": 4.35262393951416, "created_at": "2025-01-15T03:05:04.974998+00:00"} {"global_step": 9996, "acc_step": 0, "speed/wps": 12960.918554201036, "speed/FLOPS": 203569036062193.03, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3305277824401855, "optim/lr": 0.002939360156307919, "optim/total_tokens": 2620391424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480143, "loss/out": 4.319123268127441, "created_at": "2025-01-15T03:05:10.034997+00:00"} {"global_step": 9997, "acc_step": 0, "speed/wps": 12956.08078336426, "speed/FLOPS": 203493052223407.66, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1070176362991333, "optim/lr": 0.002939336038674809, "optim/total_tokens": 2620653568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434807, "loss/out": 4.451838493347168, "created_at": "2025-01-15T03:05:15.104876+00:00"} {"global_step": 9998, "acc_step": 0, "speed/wps": 12950.913915910804, "speed/FLOPS": 203411899470031.25, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1088216304779053, "optim/lr": 0.0029393119163456114, "optim/total_tokens": 2620915712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 4.487303733825684, "created_at": "2025-01-15T03:05:20.169226+00:00"} {"global_step": 9999, "acc_step": 0, "speed/wps": 12961.525424922755, "speed/FLOPS": 203578567800805.38, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.460983395576477, "optim/lr": 0.002939287789320406, "optim/total_tokens": 2621177856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432365, "loss/out": 4.454988479614258, "created_at": "2025-01-15T03:05:25.229217+00:00"} {"global_step": 10000, "acc_step": 0, "speed/wps": 12957.038114644298, "speed/FLOPS": 203508088426671.72, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.4569262266159058, "optim/lr": 0.002939263657599271, "optim/total_tokens": 2621440000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317879, "loss/out": 4.406734466552734, "created_at": "2025-01-15T03:05:30.290897+00:00"} {"global_step": 10001, "acc_step": 0, "speed/wps": 3132.4244438536534, "speed/FLOPS": 49199030292976.6, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.2735, "optim/grad_norm": 1.0723106861114502, "optim/lr": 0.002939239521182286, "optim/total_tokens": 2621702144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 4.322606086730957, "created_at": "2025-01-15T03:05:51.219719+00:00"} {"global_step": 10002, "acc_step": 0, "speed/wps": 13019.368051665195, "speed/FLOPS": 204487065737893.88, "speed/curr_iter_time": 1.2707, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.7801815867424011, "optim/lr": 0.002939215380069528, "optim/total_tokens": 2621964288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 4.33608865737915, "created_at": "2025-01-15T03:05:56.260757+00:00"} {"global_step": 10003, "acc_step": 0, "speed/wps": 12991.358361705657, "speed/FLOPS": 204047135067731.94, "speed/curr_iter_time": 1.2712, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.050897479057312, "optim/lr": 0.0029391912342610783, "optim/total_tokens": 2622226432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450990, "loss/out": 4.222307205200195, "created_at": "2025-01-15T03:06:01.313677+00:00"} {"global_step": 10004, "acc_step": 0, "speed/wps": 12979.875866291075, "speed/FLOPS": 203866786698643.75, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.7568531632423401, "optim/lr": 0.0029391670837570134, "optim/total_tokens": 2622488576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311430, "loss/out": 4.462263107299805, "created_at": "2025-01-15T03:06:06.367784+00:00"} {"global_step": 10005, "acc_step": 0, "speed/wps": 12962.17187559111, "speed/FLOPS": 203588721196868.2, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.098353385925293, "optim/lr": 0.0029391429285574133, "optim/total_tokens": 2622750720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 417160, "loss/out": 4.3393378257751465, "created_at": "2025-01-15T03:06:11.429225+00:00"} {"global_step": 10006, "acc_step": 0, "speed/wps": 12959.710229348044, "speed/FLOPS": 203550057659964.44, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7418365478515625, "optim/lr": 0.0029391187686623565, "optim/total_tokens": 2623012864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461622, "loss/out": 4.178177356719971, "created_at": "2025-01-15T03:06:16.491649+00:00"} {"global_step": 10007, "acc_step": 0, "speed/wps": 12968.504515917874, "speed/FLOPS": 203688184015159.72, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.001, "optim/grad_norm": 0.7191137075424194, "optim/lr": 0.0029390946040719216, "optim/total_tokens": 2623275008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464535, "loss/out": 4.413787841796875, "created_at": "2025-01-15T03:06:21.550862+00:00"} {"global_step": 10008, "acc_step": 0, "speed/wps": 12965.722159676263, "speed/FLOPS": 203644483287027.1, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6007426381111145, "optim/lr": 0.0029390704347861877, "optim/total_tokens": 2623537152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459765, "loss/out": 4.295257091522217, "created_at": "2025-01-15T03:06:26.612795+00:00"} {"global_step": 10009, "acc_step": 0, "speed/wps": 12972.326924095605, "speed/FLOPS": 203748220188130.25, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8509787321090698, "optim/lr": 0.002939046260805234, "optim/total_tokens": 2623799296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 4.467254638671875, "created_at": "2025-01-15T03:06:31.672911+00:00"} {"global_step": 10010, "acc_step": 0, "speed/wps": 12956.408388136955, "speed/FLOPS": 203498197706539.0, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5232000350952148, "optim/lr": 0.002939022082129139, "optim/total_tokens": 2624061440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 340756, "loss/out": 4.392594814300537, "created_at": "2025-01-15T03:06:36.737112+00:00"} {"global_step": 10011, "acc_step": 0, "speed/wps": 12958.746654074403, "speed/FLOPS": 203534923386200.72, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0249931812286377, "optim/lr": 0.002938997898757981, "optim/total_tokens": 2624323584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304748, "loss/out": 4.351925849914551, "created_at": "2025-01-15T03:06:41.801190+00:00"} {"global_step": 10012, "acc_step": 0, "speed/wps": 12956.580923323372, "speed/FLOPS": 203500907608729.88, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0015, "optim/grad_norm": 1.01448392868042, "optim/lr": 0.0029389737106918393, "optim/total_tokens": 2624585728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 373889, "loss/out": 4.510425090789795, "created_at": "2025-01-15T03:06:46.873026+00:00"} {"global_step": 10013, "acc_step": 0, "speed/wps": 12966.29778753968, "speed/FLOPS": 203653524313617.62, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.0359193086624146, "optim/lr": 0.002938949517930794, "optim/total_tokens": 2624847872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441442, "loss/out": 4.277225494384766, "created_at": "2025-01-15T03:06:51.935707+00:00"} {"global_step": 10014, "acc_step": 0, "speed/wps": 12964.075652623487, "speed/FLOPS": 203618622631225.47, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0510057210922241, "optim/lr": 0.002938925320474922, "optim/total_tokens": 2625110016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 4.462122917175293, "created_at": "2025-01-15T03:06:57.000999+00:00"} {"global_step": 10015, "acc_step": 0, "speed/wps": 12962.736347241802, "speed/FLOPS": 203597586999807.66, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7082402110099792, "optim/lr": 0.0029389011183243038, "optim/total_tokens": 2625372160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449657, "loss/out": 4.411045074462891, "created_at": "2025-01-15T03:07:02.068677+00:00"} {"global_step": 10016, "acc_step": 0, "speed/wps": 12967.58497638964, "speed/FLOPS": 203673741383289.1, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5990594625473022, "optim/lr": 0.0029388769114790173, "optim/total_tokens": 2625634304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 4.2966837882995605, "created_at": "2025-01-15T03:07:07.129590+00:00"} {"global_step": 10017, "acc_step": 0, "speed/wps": 12960.75807557549, "speed/FLOPS": 203566515524858.06, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.203973650932312, "optim/lr": 0.0029388526999391427, "optim/total_tokens": 2625896448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315714, "loss/out": 4.296670436859131, "created_at": "2025-01-15T03:07:12.190785+00:00"} {"global_step": 10018, "acc_step": 0, "speed/wps": 12957.818538268257, "speed/FLOPS": 203520346052100.44, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0600744485855103, "optim/lr": 0.0029388284837047578, "optim/total_tokens": 2626158592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372985, "loss/out": 4.318803310394287, "created_at": "2025-01-15T03:07:17.255620+00:00"} {"global_step": 10019, "acc_step": 0, "speed/wps": 12964.755172618914, "speed/FLOPS": 203629295426511.97, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0382113456726074, "optim/lr": 0.0029388042627759422, "optim/total_tokens": 2626420736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451523, "loss/out": 4.448241233825684, "created_at": "2025-01-15T03:07:22.320492+00:00"} {"global_step": 10020, "acc_step": 0, "speed/wps": 12961.153360270979, "speed/FLOPS": 203572724014178.8, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.057417869567871, "optim/lr": 0.002938780037152775, "optim/total_tokens": 2626682880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445404, "loss/out": 4.275397300720215, "created_at": "2025-01-15T03:07:27.390419+00:00"} {"global_step": 10021, "acc_step": 0, "speed/wps": 12970.147412398797, "speed/FLOPS": 203713987961968.97, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.017564058303833, "optim/lr": 0.002938755806835335, "optim/total_tokens": 2626945024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 4.302204132080078, "created_at": "2025-01-15T03:07:32.453655+00:00"} {"global_step": 10022, "acc_step": 0, "speed/wps": 12956.440671934904, "speed/FLOPS": 203498704767948.38, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1538140773773193, "optim/lr": 0.0029387315718237017, "optim/total_tokens": 2627207168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434515, "loss/out": 4.320802688598633, "created_at": "2025-01-15T03:07:37.523989+00:00"} {"global_step": 10023, "acc_step": 0, "speed/wps": 12960.530965425163, "speed/FLOPS": 203562948447865.16, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9139813184738159, "optim/lr": 0.0029387073321179534, "optim/total_tokens": 2627469312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335952, "loss/out": 4.533885955810547, "created_at": "2025-01-15T03:07:42.584850+00:00"} {"global_step": 10024, "acc_step": 0, "speed/wps": 12964.773108864614, "speed/FLOPS": 203629577139897.7, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.224597692489624, "optim/lr": 0.0029386830877181696, "optim/total_tokens": 2627731456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356429, "loss/out": 4.467328071594238, "created_at": "2025-01-15T03:07:47.649250+00:00"} {"global_step": 10025, "acc_step": 0, "speed/wps": 12958.896069318784, "speed/FLOPS": 203537270157931.78, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.550945520401001, "optim/lr": 0.0029386588386244297, "optim/total_tokens": 2627993600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 385346, "loss/out": 4.282108306884766, "created_at": "2025-01-15T03:07:52.710118+00:00"} {"global_step": 10026, "acc_step": 0, "speed/wps": 12959.59399702175, "speed/FLOPS": 203548232071560.03, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.1390979290008545, "optim/lr": 0.002938634584836812, "optim/total_tokens": 2628255744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435852, "loss/out": 4.340964317321777, "created_at": "2025-01-15T03:07:57.771676+00:00"} {"global_step": 10027, "acc_step": 0, "speed/wps": 12961.484585506016, "speed/FLOPS": 203577926361646.25, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0317260026931763, "optim/lr": 0.0029386103263553966, "optim/total_tokens": 2628517888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 4.27373743057251, "created_at": "2025-01-15T03:08:02.835410+00:00"} {"global_step": 10028, "acc_step": 0, "speed/wps": 12963.021899318504, "speed/FLOPS": 203602071987561.94, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9279959797859192, "optim/lr": 0.0029385860631802618, "optim/total_tokens": 2628780032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448361, "loss/out": 4.320349216461182, "created_at": "2025-01-15T03:08:07.899000+00:00"} {"global_step": 10029, "acc_step": 0, "speed/wps": 12966.28663238279, "speed/FLOPS": 203653349106549.88, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2150273323059082, "optim/lr": 0.002938561795311488, "optim/total_tokens": 2629042176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 4.393263339996338, "created_at": "2025-01-15T03:08:12.986938+00:00"} {"global_step": 10030, "acc_step": 0, "speed/wps": 12961.996264256168, "speed/FLOPS": 203585962979537.66, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.261553168296814, "optim/lr": 0.0029385375227491523, "optim/total_tokens": 2629304320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329951, "loss/out": 4.25100040435791, "created_at": "2025-01-15T03:08:18.046898+00:00"} {"global_step": 10031, "acc_step": 0, "speed/wps": 12958.340009573598, "speed/FLOPS": 203528536475527.47, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8294975161552429, "optim/lr": 0.002938513245493336, "optim/total_tokens": 2629566464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446218, "loss/out": 4.4402666091918945, "created_at": "2025-01-15T03:08:23.113035+00:00"} {"global_step": 10032, "acc_step": 0, "speed/wps": 12962.321243282615, "speed/FLOPS": 203591067221716.0, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.1583657264709473, "optim/lr": 0.002938488963544117, "optim/total_tokens": 2629828608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 4.372986316680908, "created_at": "2025-01-15T03:08:28.174263+00:00"} {"global_step": 10033, "acc_step": 0, "speed/wps": 12961.61022733969, "speed/FLOPS": 203579899739294.06, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5842294692993164, "optim/lr": 0.002938464676901575, "optim/total_tokens": 2630090752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 4.431230068206787, "created_at": "2025-01-15T03:08:33.258450+00:00"} {"global_step": 10034, "acc_step": 0, "speed/wps": 12960.78365640186, "speed/FLOPS": 203566917306887.94, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9808799624443054, "optim/lr": 0.00293844038556579, "optim/total_tokens": 2630352896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438769, "loss/out": 4.3569746017456055, "created_at": "2025-01-15T03:08:38.323026+00:00"} {"global_step": 10035, "acc_step": 0, "speed/wps": 12962.019407432883, "speed/FLOPS": 203586326474930.03, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1157870292663574, "optim/lr": 0.002938416089536839, "optim/total_tokens": 2630615040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 344264, "loss/out": 4.39735221862793, "created_at": "2025-01-15T03:08:43.387731+00:00"} {"global_step": 10036, "acc_step": 0, "speed/wps": 12966.061691270632, "speed/FLOPS": 203649816097281.97, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8270699381828308, "optim/lr": 0.0029383917888148037, "optim/total_tokens": 2630877184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433464, "loss/out": 4.3426899909973145, "created_at": "2025-01-15T03:08:48.449983+00:00"} {"global_step": 10037, "acc_step": 0, "speed/wps": 12967.848842238587, "speed/FLOPS": 203677885759036.47, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8291338682174683, "optim/lr": 0.0029383674833997624, "optim/total_tokens": 2631139328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442778, "loss/out": 4.344216823577881, "created_at": "2025-01-15T03:08:53.513337+00:00"} {"global_step": 10038, "acc_step": 0, "speed/wps": 12961.59101538861, "speed/FLOPS": 203579597989202.38, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1123545169830322, "optim/lr": 0.0029383431732917936, "optim/total_tokens": 2631401472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 4.354983329772949, "created_at": "2025-01-15T03:08:58.573823+00:00"} {"global_step": 10039, "acc_step": 0, "speed/wps": 12961.212477318591, "speed/FLOPS": 203573652528646.9, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0576826333999634, "optim/lr": 0.0029383188584909785, "optim/total_tokens": 2631663616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 4.3987603187561035, "created_at": "2025-01-15T03:09:03.637365+00:00"} {"global_step": 10040, "acc_step": 0, "speed/wps": 12958.891239102702, "speed/FLOPS": 203537194292750.8, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.803227424621582, "optim/lr": 0.0029382945389973946, "optim/total_tokens": 2631925760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441971, "loss/out": 4.351978302001953, "created_at": "2025-01-15T03:09:08.698887+00:00"} {"global_step": 10041, "acc_step": 0, "speed/wps": 12959.087641360215, "speed/FLOPS": 203540279060090.03, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7386037111282349, "optim/lr": 0.0029382702148111227, "optim/total_tokens": 2632187904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 4.313981533050537, "created_at": "2025-01-15T03:09:13.764469+00:00"} {"global_step": 10042, "acc_step": 0, "speed/wps": 12964.913027138724, "speed/FLOPS": 203631774748660.44, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.628389835357666, "optim/lr": 0.002938245885932241, "optim/total_tokens": 2632450048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334076, "loss/out": 4.329470157623291, "created_at": "2025-01-15T03:09:18.823866+00:00"} {"global_step": 10043, "acc_step": 0, "speed/wps": 12964.298162642206, "speed/FLOPS": 203622117456830.25, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.099427342414856, "optim/lr": 0.0029382215523608294, "optim/total_tokens": 2632712192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337297, "loss/out": 4.4992170333862305, "created_at": "2025-01-15T03:09:23.885622+00:00"} {"global_step": 10044, "acc_step": 0, "speed/wps": 12950.108501332204, "speed/FLOPS": 203399249327318.72, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6321650147438049, "optim/lr": 0.0029381972140969676, "optim/total_tokens": 2632974336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435567, "loss/out": 4.412468433380127, "created_at": "2025-01-15T03:09:28.953994+00:00"} {"global_step": 10045, "acc_step": 0, "speed/wps": 12957.871316063576, "speed/FLOPS": 203521174999900.22, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0421926975250244, "optim/lr": 0.002938172871140734, "optim/total_tokens": 2633236480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462938, "loss/out": 4.351146697998047, "created_at": "2025-01-15T03:09:34.020336+00:00"} {"global_step": 10046, "acc_step": 0, "speed/wps": 12967.616546431267, "speed/FLOPS": 203674237234174.62, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.743522047996521, "optim/lr": 0.002938148523492209, "optim/total_tokens": 2633498624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452081, "loss/out": 4.275976181030273, "created_at": "2025-01-15T03:09:39.082834+00:00"} {"global_step": 10047, "acc_step": 0, "speed/wps": 12959.822153174688, "speed/FLOPS": 203551815577461.3, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7817507386207581, "optim/lr": 0.0029381241711514718, "optim/total_tokens": 2633760768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460593, "loss/out": 4.397497653961182, "created_at": "2025-01-15T03:09:44.148898+00:00"} {"global_step": 10048, "acc_step": 0, "speed/wps": 12958.685244715962, "speed/FLOPS": 203533958867841.1, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.0772619247436523, "optim/lr": 0.0029380998141186015, "optim/total_tokens": 2634022912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314655, "loss/out": 4.2207489013671875, "created_at": "2025-01-15T03:09:49.209945+00:00"} {"global_step": 10049, "acc_step": 0, "speed/wps": 12963.962291575903, "speed/FLOPS": 203616842140198.75, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0644044876098633, "optim/lr": 0.0029380754523936777, "optim/total_tokens": 2634285056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 373317, "loss/out": 4.317370891571045, "created_at": "2025-01-15T03:09:54.269992+00:00"} {"global_step": 10050, "acc_step": 0, "speed/wps": 12966.124173997292, "speed/FLOPS": 203650797474364.44, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5695430040359497, "optim/lr": 0.0029380510859767805, "optim/total_tokens": 2634547200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446747, "loss/out": 4.311025619506836, "created_at": "2025-01-15T03:09:59.328181+00:00"} {"global_step": 10051, "acc_step": 0, "speed/wps": 12956.687864333726, "speed/FLOPS": 203502587264249.8, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3381068706512451, "optim/lr": 0.002938026714867989, "optim/total_tokens": 2634809344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448605, "loss/out": 4.333844184875488, "created_at": "2025-01-15T03:10:04.390394+00:00"} {"global_step": 10052, "acc_step": 0, "speed/wps": 12962.049458110434, "speed/FLOPS": 203586798462114.94, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.866506040096283, "optim/lr": 0.002938002339067382, "optim/total_tokens": 2635071488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438531, "loss/out": 4.327579021453857, "created_at": "2025-01-15T03:10:09.452404+00:00"} {"global_step": 10053, "acc_step": 0, "speed/wps": 12959.14076122632, "speed/FLOPS": 203541113380581.53, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2866758108139038, "optim/lr": 0.00293797795857504, "optim/total_tokens": 2635333632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492087, "loss/out": 4.307008743286133, "created_at": "2025-01-15T03:10:14.516932+00:00"} {"global_step": 10054, "acc_step": 0, "speed/wps": 12970.513411862466, "speed/FLOPS": 203719736486481.88, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0187575817108154, "optim/lr": 0.002937953573391042, "optim/total_tokens": 2635595776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445695, "loss/out": 4.271815776824951, "created_at": "2025-01-15T03:10:19.575860+00:00"} {"global_step": 10055, "acc_step": 0, "speed/wps": 12958.944767834828, "speed/FLOPS": 203538035035045.03, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1095592975616455, "optim/lr": 0.002937929183515468, "optim/total_tokens": 2635857920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 4.328036785125732, "created_at": "2025-01-15T03:10:24.636694+00:00"} {"global_step": 10056, "acc_step": 0, "speed/wps": 12963.273913697512, "speed/FLOPS": 203606030219688.28, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.685326099395752, "optim/lr": 0.002937904788948398, "optim/total_tokens": 2636120064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445170, "loss/out": 4.285770893096924, "created_at": "2025-01-15T03:10:29.695895+00:00"} {"global_step": 10057, "acc_step": 0, "speed/wps": 12956.705573097572, "speed/FLOPS": 203502865404720.4, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0723797082901, "optim/lr": 0.0029378803896899096, "optim/total_tokens": 2636382208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294654, "loss/out": 4.295340061187744, "created_at": "2025-01-15T03:10:34.762383+00:00"} {"global_step": 10058, "acc_step": 0, "speed/wps": 12966.498626652845, "speed/FLOPS": 203656678767871.75, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.559617519378662, "optim/lr": 0.0029378559857400847, "optim/total_tokens": 2636644352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 403566, "loss/out": 4.234514236450195, "created_at": "2025-01-15T03:10:39.824795+00:00"} {"global_step": 10059, "acc_step": 0, "speed/wps": 12959.810111717386, "speed/FLOPS": 203551626449827.72, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8263300657272339, "optim/lr": 0.002937831577099001, "optim/total_tokens": 2636906496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466683, "loss/out": 4.3344926834106445, "created_at": "2025-01-15T03:10:44.886562+00:00"} {"global_step": 10060, "acc_step": 0, "speed/wps": 12963.789485711604, "speed/FLOPS": 203614127986639.8, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.591542363166809, "optim/lr": 0.0029378071637667404, "optim/total_tokens": 2637168640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450990, "loss/out": 4.42681360244751, "created_at": "2025-01-15T03:10:49.950582+00:00"} {"global_step": 10061, "acc_step": 0, "speed/wps": 12960.585815831673, "speed/FLOPS": 203563809948871.75, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0961270332336426, "optim/lr": 0.0029377827457433806, "optim/total_tokens": 2637430784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471703, "loss/out": 4.298699855804443, "created_at": "2025-01-15T03:10:55.017330+00:00"} {"global_step": 10062, "acc_step": 0, "speed/wps": 12960.964985723484, "speed/FLOPS": 203569765333056.66, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2863035202026367, "optim/lr": 0.0029377583230290015, "optim/total_tokens": 2637692928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439054, "loss/out": 4.304701805114746, "created_at": "2025-01-15T03:11:00.082285+00:00"} {"global_step": 10063, "acc_step": 0, "speed/wps": 12956.918332542562, "speed/FLOPS": 203506207084166.1, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1063135862350464, "optim/lr": 0.002937733895623684, "optim/total_tokens": 2637955072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315714, "loss/out": 4.351216793060303, "created_at": "2025-01-15T03:11:05.148857+00:00"} {"global_step": 10064, "acc_step": 0, "speed/wps": 12972.52822983296, "speed/FLOPS": 203751381971358.5, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5738739967346191, "optim/lr": 0.0029377094635275065, "optim/total_tokens": 2638217216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326870, "loss/out": 4.349711894989014, "created_at": "2025-01-15T03:11:10.205179+00:00"} {"global_step": 10065, "acc_step": 0, "speed/wps": 12956.50244171607, "speed/FLOPS": 203499674947242.12, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.8199411034584045, "optim/lr": 0.002937685026740549, "optim/total_tokens": 2638479360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352401, "loss/out": 4.281279563903809, "created_at": "2025-01-15T03:11:15.273303+00:00"} {"global_step": 10066, "acc_step": 0, "speed/wps": 12964.48179970703, "speed/FLOPS": 203625001729277.06, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8256007432937622, "optim/lr": 0.002937660585262892, "optim/total_tokens": 2638741504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439820, "loss/out": 4.400400161743164, "created_at": "2025-01-15T03:11:20.337681+00:00"} {"global_step": 10067, "acc_step": 0, "speed/wps": 12961.873592235119, "speed/FLOPS": 203584036246878.06, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2688193321228027, "optim/lr": 0.0029376361390946143, "optim/total_tokens": 2639003648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 4.208064079284668, "created_at": "2025-01-15T03:11:25.400941+00:00"} {"global_step": 10068, "acc_step": 0, "speed/wps": 12961.650294002737, "speed/FLOPS": 203580529041294.6, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.7434273958206177, "optim/lr": 0.0029376116882357966, "optim/total_tokens": 2639265792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 4.33836030960083, "created_at": "2025-01-15T03:11:30.461523+00:00"} {"global_step": 10069, "acc_step": 0, "speed/wps": 12957.235813516645, "speed/FLOPS": 203511193559128.62, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2243165969848633, "optim/lr": 0.0029375872326865174, "optim/total_tokens": 2639527936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 4.306366920471191, "created_at": "2025-01-15T03:11:35.528680+00:00"} {"global_step": 10070, "acc_step": 0, "speed/wps": 12958.592085350925, "speed/FLOPS": 203532495671998.8, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.044338345527649, "optim/lr": 0.0029375627724468577, "optim/total_tokens": 2639790080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319377, "loss/out": 4.3996357917785645, "created_at": "2025-01-15T03:11:40.591734+00:00"} {"global_step": 10071, "acc_step": 0, "speed/wps": 12960.511448741798, "speed/FLOPS": 203562641911534.16, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4033280611038208, "optim/lr": 0.002937538307516897, "optim/total_tokens": 2640052224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361859, "loss/out": 4.32885217666626, "created_at": "2025-01-15T03:11:45.652785+00:00"} {"global_step": 10072, "acc_step": 0, "speed/wps": 12957.065162368242, "speed/FLOPS": 203508513248343.72, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1071834564208984, "optim/lr": 0.0029375138378967144, "optim/total_tokens": 2640314368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450756, "loss/out": 4.421672821044922, "created_at": "2025-01-15T03:11:50.714726+00:00"} {"global_step": 10073, "acc_step": 0, "speed/wps": 12972.816227209769, "speed/FLOPS": 203755905365911.16, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0519704818725586, "optim/lr": 0.0029374893635863908, "optim/total_tokens": 2640576512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 4.3554792404174805, "created_at": "2025-01-15T03:11:55.772828+00:00"} {"global_step": 10074, "acc_step": 0, "speed/wps": 12959.7592180113, "speed/FLOPS": 203550827094238.34, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6304931640625, "optim/lr": 0.002937464884586005, "optim/total_tokens": 2640838656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442784, "loss/out": 4.38625955581665, "created_at": "2025-01-15T03:12:00.841283+00:00"} {"global_step": 10075, "acc_step": 0, "speed/wps": 12961.328356699796, "speed/FLOPS": 203575472573563.7, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9862273931503296, "optim/lr": 0.0029374404008956374, "optim/total_tokens": 2641100800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 4.417466640472412, "created_at": "2025-01-15T03:12:05.905338+00:00"} {"global_step": 10076, "acc_step": 0, "speed/wps": 12963.983464223396, "speed/FLOPS": 203617174685721.9, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1557292938232422, "optim/lr": 0.0029374159125153685, "optim/total_tokens": 2641362944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444357, "loss/out": 4.486809730529785, "created_at": "2025-01-15T03:12:10.967727+00:00"} {"global_step": 10077, "acc_step": 0, "speed/wps": 12963.41305212775, "speed/FLOPS": 203608215579931.4, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8583469986915588, "optim/lr": 0.0029373914194452773, "optim/total_tokens": 2641625088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292300, "loss/out": 4.321930408477783, "created_at": "2025-01-15T03:12:16.033067+00:00"} {"global_step": 10078, "acc_step": 0, "speed/wps": 12958.162841254478, "speed/FLOPS": 203525753803620.9, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.812386155128479, "optim/lr": 0.002937366921685444, "optim/total_tokens": 2641887232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326870, "loss/out": 4.318764686584473, "created_at": "2025-01-15T03:12:21.095264+00:00"} {"global_step": 10079, "acc_step": 0, "speed/wps": 12966.81118444495, "speed/FLOPS": 203661587917496.2, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6778464913368225, "optim/lr": 0.002937342419235948, "optim/total_tokens": 2642149376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290365, "loss/out": 4.329194068908691, "created_at": "2025-01-15T03:12:26.159193+00:00"} {"global_step": 10080, "acc_step": 0, "speed/wps": 12944.624494286534, "speed/FLOPS": 203313115460851.1, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7318022847175598, "optim/lr": 0.0029373179120968703, "optim/total_tokens": 2642411520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430506, "loss/out": 4.282284259796143, "created_at": "2025-01-15T03:12:31.225918+00:00"} {"global_step": 10081, "acc_step": 0, "speed/wps": 12962.55429737583, "speed/FLOPS": 203594727656499.8, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5346885919570923, "optim/lr": 0.0029372934002682904, "optim/total_tokens": 2642673664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484902, "loss/out": 4.2281904220581055, "created_at": "2025-01-15T03:12:36.289489+00:00"} {"global_step": 10082, "acc_step": 0, "speed/wps": 12962.444869337281, "speed/FLOPS": 203593008938784.53, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6574047207832336, "optim/lr": 0.0029372688837502876, "optim/total_tokens": 2642935808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441442, "loss/out": 4.287919998168945, "created_at": "2025-01-15T03:12:41.353178+00:00"} {"global_step": 10083, "acc_step": 0, "speed/wps": 12960.925644237184, "speed/FLOPS": 203569147420953.47, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9349343776702881, "optim/lr": 0.002937244362542943, "optim/total_tokens": 2643197952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 4.35870885848999, "created_at": "2025-01-15T03:12:46.414364+00:00"} {"global_step": 10084, "acc_step": 0, "speed/wps": 12965.354894524895, "speed/FLOPS": 203638714883149.66, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7542359828948975, "optim/lr": 0.002937219836646336, "optim/total_tokens": 2643460096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445404, "loss/out": 4.3719162940979, "created_at": "2025-01-15T03:12:51.473174+00:00"} {"global_step": 10085, "acc_step": 0, "speed/wps": 12962.083104720083, "speed/FLOPS": 203587326928354.75, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.041027545928955, "optim/lr": 0.002937195306060547, "optim/total_tokens": 2643722240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354978, "loss/out": 4.32006311416626, "created_at": "2025-01-15T03:12:56.533960+00:00"} {"global_step": 10086, "acc_step": 0, "speed/wps": 12963.92555427089, "speed/FLOPS": 203616265130341.2, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0593467950820923, "optim/lr": 0.0029371707707856556, "optim/total_tokens": 2643984384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445976, "loss/out": 4.3304972648620605, "created_at": "2025-01-15T03:13:01.597850+00:00"} {"global_step": 10087, "acc_step": 0, "speed/wps": 12961.524522470569, "speed/FLOPS": 203578553626553.7, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1664584875106812, "optim/lr": 0.0029371462308217418, "optim/total_tokens": 2644246528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481758, "loss/out": 4.357937335968018, "created_at": "2025-01-15T03:13:06.658041+00:00"} {"global_step": 10088, "acc_step": 0, "speed/wps": 12961.087511633792, "speed/FLOPS": 203571689770846.28, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0497554540634155, "optim/lr": 0.002937121686168886, "optim/total_tokens": 2644508672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449657, "loss/out": 4.3554582595825195, "created_at": "2025-01-15T03:13:11.720879+00:00"} {"global_step": 10089, "acc_step": 0, "speed/wps": 12961.18355972112, "speed/FLOPS": 203573198338041.66, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.0718472003936768, "optim/lr": 0.002937097136827168, "optim/total_tokens": 2644770816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 4.357515811920166, "created_at": "2025-01-15T03:13:16.782033+00:00"} {"global_step": 10090, "acc_step": 0, "speed/wps": 12960.527110302644, "speed/FLOPS": 203562887897868.5, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0319032669067383, "optim/lr": 0.0029370725827966683, "optim/total_tokens": 2645032960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 4.353896141052246, "created_at": "2025-01-15T03:13:21.843711+00:00"} {"global_step": 10091, "acc_step": 0, "speed/wps": 12967.631993368588, "speed/FLOPS": 203674479849551.56, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.061524510383606, "optim/lr": 0.0029370480240774664, "optim/total_tokens": 2645295104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481190, "loss/out": 4.340767860412598, "created_at": "2025-01-15T03:13:26.904398+00:00"} {"global_step": 10092, "acc_step": 0, "speed/wps": 12964.570885286443, "speed/FLOPS": 203626400940718.2, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9327085614204407, "optim/lr": 0.002937023460669643, "optim/total_tokens": 2645557248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460287, "loss/out": 4.357100486755371, "created_at": "2025-01-15T03:13:31.964539+00:00"} {"global_step": 10093, "acc_step": 0, "speed/wps": 12967.674360319817, "speed/FLOPS": 203675145280738.9, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.810025930404663, "optim/lr": 0.0029369988925732776, "optim/total_tokens": 2645819392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 4.407759666442871, "created_at": "2025-01-15T03:13:37.027279+00:00"} {"global_step": 10094, "acc_step": 0, "speed/wps": 12963.143000255888, "speed/FLOPS": 203603974044193.66, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1039913892745972, "optim/lr": 0.0029369743197884515, "optim/total_tokens": 2646081536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 337829, "loss/out": 4.398820877075195, "created_at": "2025-01-15T03:13:42.090930+00:00"} {"global_step": 10095, "acc_step": 0, "speed/wps": 12965.73624754255, "speed/FLOPS": 203644704556325.9, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2818788290023804, "optim/lr": 0.0029369497423152435, "optim/total_tokens": 2646343680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449420, "loss/out": 4.388866901397705, "created_at": "2025-01-15T03:13:47.149762+00:00"} {"global_step": 10096, "acc_step": 0, "speed/wps": 12961.979157029691, "speed/FLOPS": 203585694287038.06, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9564456343650818, "optim/lr": 0.0029369251601537353, "optim/total_tokens": 2646605824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 4.410742282867432, "created_at": "2025-01-15T03:13:52.210899+00:00"} {"global_step": 10097, "acc_step": 0, "speed/wps": 12960.593339955329, "speed/FLOPS": 203563928125572.97, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9209898710250854, "optim/lr": 0.0029369005733040055, "optim/total_tokens": 2646867968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454695, "loss/out": 4.201892852783203, "created_at": "2025-01-15T03:13:57.276116+00:00"} {"global_step": 10098, "acc_step": 0, "speed/wps": 12962.232444101006, "speed/FLOPS": 203589672508545.97, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.305720329284668, "optim/lr": 0.0029368759817661352, "optim/total_tokens": 2647130112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491858, "loss/out": 4.275996208190918, "created_at": "2025-01-15T03:14:02.337321+00:00"} {"global_step": 10099, "acc_step": 0, "speed/wps": 12968.12166410487, "speed/FLOPS": 203682170801343.0, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1237425804138184, "optim/lr": 0.0029368513855402047, "optim/total_tokens": 2647392256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462150, "loss/out": 4.196530342102051, "created_at": "2025-01-15T03:14:07.394683+00:00"} {"global_step": 10100, "acc_step": 0, "speed/wps": 12964.849863371544, "speed/FLOPS": 203630782674900.75, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9826892018318176, "optim/lr": 0.002936826784626294, "optim/total_tokens": 2647654400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319464, "loss/out": 4.367114543914795, "created_at": "2025-01-15T03:14:12.453780+00:00"} {"global_step": 10101, "acc_step": 0, "speed/wps": 12958.999993817248, "speed/FLOPS": 203538902434987.38, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7184916734695435, "optim/lr": 0.0029368021790244832, "optim/total_tokens": 2647916544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338354, "loss/out": 4.3022780418396, "created_at": "2025-01-15T03:14:17.519815+00:00"} {"global_step": 10102, "acc_step": 0, "speed/wps": 12961.90153877968, "speed/FLOPS": 203584475185763.0, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.890554666519165, "optim/lr": 0.002936777568734853, "optim/total_tokens": 2648178688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451275, "loss/out": 4.366377353668213, "created_at": "2025-01-15T03:14:22.579862+00:00"} {"global_step": 10103, "acc_step": 0, "speed/wps": 12964.048133774695, "speed/FLOPS": 203618190409890.3, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2640249729156494, "optim/lr": 0.0029367529537574835, "optim/total_tokens": 2648440832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 4.345850467681885, "created_at": "2025-01-15T03:14:27.643426+00:00"} {"global_step": 10104, "acc_step": 0, "speed/wps": 12962.612866061252, "speed/FLOPS": 203595647558186.2, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5404930710792542, "optim/lr": 0.002936728334092455, "optim/total_tokens": 2648702976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 4.390421390533447, "created_at": "2025-01-15T03:14:32.704945+00:00"} {"global_step": 10105, "acc_step": 0, "speed/wps": 12962.138278661578, "speed/FLOPS": 203588193510922.88, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9137521982192993, "optim/lr": 0.0029367037097398478, "optim/total_tokens": 2648965120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 4.274990081787109, "created_at": "2025-01-15T03:14:37.770195+00:00"} {"global_step": 10106, "acc_step": 0, "speed/wps": 12960.566501885469, "speed/FLOPS": 203563506596806.47, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.432684063911438, "optim/lr": 0.002936679080699742, "optim/total_tokens": 2649227264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432656, "loss/out": 4.521302223205566, "created_at": "2025-01-15T03:14:42.836301+00:00"} {"global_step": 10107, "acc_step": 0, "speed/wps": 12963.967302310146, "speed/FLOPS": 203616920840665.5, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9658620953559875, "optim/lr": 0.0029366544469722187, "optim/total_tokens": 2649489408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479331, "loss/out": 4.362753868103027, "created_at": "2025-01-15T03:14:47.901265+00:00"} {"global_step": 10108, "acc_step": 0, "speed/wps": 12963.221610093246, "speed/FLOPS": 203605208719710.6, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5088741183280945, "optim/lr": 0.002936629808557357, "optim/total_tokens": 2649751552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443545, "loss/out": 4.336928367614746, "created_at": "2025-01-15T03:14:52.960745+00:00"} {"global_step": 10109, "acc_step": 0, "speed/wps": 12967.989106517834, "speed/FLOPS": 203680088802285.38, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.9752512574195862, "optim/lr": 0.0029366051654552395, "optim/total_tokens": 2650013696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 285549, "loss/out": 4.3849005699157715, "created_at": "2025-01-15T03:14:58.018072+00:00"} {"global_step": 10110, "acc_step": 0, "speed/wps": 12959.645194933812, "speed/FLOPS": 203549036205122.25, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6548582315444946, "optim/lr": 0.002936580517665944, "optim/total_tokens": 2650275840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 403566, "loss/out": 4.302133083343506, "created_at": "2025-01-15T03:15:03.080345+00:00"} {"global_step": 10111, "acc_step": 0, "speed/wps": 12965.415363786036, "speed/FLOPS": 203639664635989.34, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5484331250190735, "optim/lr": 0.0029365558651895524, "optim/total_tokens": 2650537984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441157, "loss/out": 4.242352485656738, "created_at": "2025-01-15T03:15:08.143124+00:00"} {"global_step": 10112, "acc_step": 0, "speed/wps": 12964.20799885271, "speed/FLOPS": 203620701310618.2, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8778344988822937, "optim/lr": 0.002936531208026145, "optim/total_tokens": 2650800128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 4.396150588989258, "created_at": "2025-01-15T03:15:13.206716+00:00"} {"global_step": 10113, "acc_step": 0, "speed/wps": 12959.346717589411, "speed/FLOPS": 203544348208277.6, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.578162431716919, "optim/lr": 0.0029365065461758023, "optim/total_tokens": 2651062272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450746, "loss/out": 4.371307373046875, "created_at": "2025-01-15T03:15:18.272338+00:00"} {"global_step": 10114, "acc_step": 0, "speed/wps": 12965.847164578687, "speed/FLOPS": 203646446660794.4, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.416262686252594, "optim/lr": 0.002936481879638604, "optim/total_tokens": 2651324416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 4.300249099731445, "created_at": "2025-01-15T03:15:23.333234+00:00"} {"global_step": 10115, "acc_step": 0, "speed/wps": 12963.004070601266, "speed/FLOPS": 203601791963058.47, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6500447988510132, "optim/lr": 0.0029364572084146314, "optim/total_tokens": 2651586560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329272, "loss/out": 4.341161251068115, "created_at": "2025-01-15T03:15:28.397428+00:00"} {"global_step": 10116, "acc_step": 0, "speed/wps": 12958.739518693734, "speed/FLOPS": 203534811315242.25, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7364940047264099, "optim/lr": 0.002936432532503964, "optim/total_tokens": 2651848704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367993, "loss/out": 4.4101667404174805, "created_at": "2025-01-15T03:15:33.463879+00:00"} {"global_step": 10117, "acc_step": 0, "speed/wps": 12964.144099161109, "speed/FLOPS": 203619697678154.22, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8406534194946289, "optim/lr": 0.0029364078519066843, "optim/total_tokens": 2652110848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447555, "loss/out": 4.290002822875977, "created_at": "2025-01-15T03:15:38.523052+00:00"} {"global_step": 10118, "acc_step": 0, "speed/wps": 12963.355763404586, "speed/FLOPS": 203607315781811.2, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8802404403686523, "optim/lr": 0.0029363831666228708, "optim/total_tokens": 2652372992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444644, "loss/out": 4.37142276763916, "created_at": "2025-01-15T03:15:43.587245+00:00"} {"global_step": 10119, "acc_step": 0, "speed/wps": 12969.203756268535, "speed/FLOPS": 203699166545723.78, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.0589340925216675, "optim/lr": 0.002936358476652605, "optim/total_tokens": 2652635136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453660, "loss/out": 4.462995529174805, "created_at": "2025-01-15T03:15:48.647479+00:00"} {"global_step": 10120, "acc_step": 0, "speed/wps": 12960.28606824926, "speed/FLOPS": 203559102001193.6, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8133130073547363, "optim/lr": 0.002936333781995967, "optim/total_tokens": 2652897280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 4.359721660614014, "created_at": "2025-01-15T03:15:53.713143+00:00"} {"global_step": 10121, "acc_step": 0, "speed/wps": 12965.136358974982, "speed/FLOPS": 203635282482038.97, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.327810525894165, "optim/lr": 0.0029363090826530383, "optim/total_tokens": 2653159424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425730, "loss/out": 4.392882347106934, "created_at": "2025-01-15T03:15:58.774032+00:00"} {"global_step": 10122, "acc_step": 0, "speed/wps": 12966.450332473045, "speed/FLOPS": 203655920241414.84, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7891309261322021, "optim/lr": 0.0029362843786238982, "optim/total_tokens": 2653421568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 4.472312927246094, "created_at": "2025-01-15T03:16:03.832269+00:00"} {"global_step": 10123, "acc_step": 0, "speed/wps": 12963.337680629578, "speed/FLOPS": 203607031766981.34, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8368648886680603, "optim/lr": 0.0029362596699086282, "optim/total_tokens": 2653683712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448077, "loss/out": 4.354251384735107, "created_at": "2025-01-15T03:16:08.894898+00:00"} {"global_step": 10124, "acc_step": 0, "speed/wps": 12967.22053288418, "speed/FLOPS": 203668017297237.97, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9189376831054688, "optim/lr": 0.0029362349565073085, "optim/total_tokens": 2653945856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294654, "loss/out": 4.344005107879639, "created_at": "2025-01-15T03:16:13.953060+00:00"} {"global_step": 10125, "acc_step": 0, "speed/wps": 12959.778014003552, "speed/FLOPS": 203551122311125.06, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9466427564620972, "optim/lr": 0.0029362102384200197, "optim/total_tokens": 2654208000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444068, "loss/out": 4.203106880187988, "created_at": "2025-01-15T03:16:19.016452+00:00"} {"global_step": 10126, "acc_step": 0, "speed/wps": 12960.167355110783, "speed/FLOPS": 203557237448224.88, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8124598264694214, "optim/lr": 0.002936185515646843, "optim/total_tokens": 2654470144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 4.251174449920654, "created_at": "2025-01-15T03:16:24.082137+00:00"} {"global_step": 10127, "acc_step": 0, "speed/wps": 12953.73454907321, "speed/FLOPS": 203456201389802.97, "speed/curr_iter_time": 1.2774, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0996288061141968, "optim/lr": 0.002936160788187858, "optim/total_tokens": 2654732288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437432, "loss/out": 4.374044895172119, "created_at": "2025-01-15T03:16:29.150492+00:00"} {"global_step": 10128, "acc_step": 0, "speed/wps": 12962.760056604493, "speed/FLOPS": 203597959387928.78, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.7720900177955627, "optim/lr": 0.0029361360560431465, "optim/total_tokens": 2654994432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 4.315341949462891, "created_at": "2025-01-15T03:16:34.212153+00:00"} {"global_step": 10129, "acc_step": 0, "speed/wps": 12961.750199116583, "speed/FLOPS": 203582098188391.53, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5626440644264221, "optim/lr": 0.0029361113192127894, "optim/total_tokens": 2655256576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443835, "loss/out": 4.326692581176758, "created_at": "2025-01-15T03:16:39.273208+00:00"} {"global_step": 10130, "acc_step": 0, "speed/wps": 12960.789512124078, "speed/FLOPS": 203567009279052.0, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9214035868644714, "optim/lr": 0.0029360865776968657, "optim/total_tokens": 2655518720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313309, "loss/out": 4.262044906616211, "created_at": "2025-01-15T03:16:44.337276+00:00"} {"global_step": 10131, "acc_step": 0, "speed/wps": 12965.64153906393, "speed/FLOPS": 203643217029526.1, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9899094700813293, "optim/lr": 0.0029360618314954578, "optim/total_tokens": 2655780864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302050, "loss/out": 4.350533485412598, "created_at": "2025-01-15T03:16:49.396925+00:00"} {"global_step": 10132, "acc_step": 0, "speed/wps": 12965.193618050132, "speed/FLOPS": 203636181814496.4, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4402846693992615, "optim/lr": 0.0029360370806086455, "optim/total_tokens": 2656043008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357905, "loss/out": 4.406211853027344, "created_at": "2025-01-15T03:16:54.457329+00:00"} {"global_step": 10133, "acc_step": 0, "speed/wps": 12968.204191635059, "speed/FLOPS": 203683467009609.3, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2902770042419434, "optim/lr": 0.00293601232503651, "optim/total_tokens": 2656305152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451275, "loss/out": 4.262792110443115, "created_at": "2025-01-15T03:16:59.515818+00:00"} {"global_step": 10134, "acc_step": 0, "speed/wps": 12966.693739948701, "speed/FLOPS": 203659743290296.94, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8282506465911865, "optim/lr": 0.0029359875647791316, "optim/total_tokens": 2656567296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 4.342883110046387, "created_at": "2025-01-15T03:17:04.578705+00:00"} {"global_step": 10135, "acc_step": 0, "speed/wps": 12962.644872753397, "speed/FLOPS": 203596150267268.12, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0373528003692627, "optim/lr": 0.002935962799836592, "optim/total_tokens": 2656829440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 4.321547508239746, "created_at": "2025-01-15T03:17:09.638322+00:00"} {"global_step": 10136, "acc_step": 0, "speed/wps": 12964.424733222617, "speed/FLOPS": 203624105421721.8, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7017218470573425, "optim/lr": 0.002935938030208971, "optim/total_tokens": 2657091584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429454, "loss/out": 4.345034599304199, "created_at": "2025-01-15T03:17:14.697638+00:00"} {"global_step": 10137, "acc_step": 0, "speed/wps": 12961.56309644503, "speed/FLOPS": 203579159483828.8, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5721731185913086, "optim/lr": 0.00293591325589635, "optim/total_tokens": 2657353728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 4.175889015197754, "created_at": "2025-01-15T03:17:19.763421+00:00"} {"global_step": 10138, "acc_step": 0, "speed/wps": 12971.486453990348, "speed/FLOPS": 203735019450200.25, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7953441739082336, "optim/lr": 0.0029358884768988097, "optim/total_tokens": 2657615872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448605, "loss/out": 4.384923934936523, "created_at": "2025-01-15T03:17:24.833658+00:00"} {"global_step": 10139, "acc_step": 0, "speed/wps": 12956.102082943822, "speed/FLOPS": 203493386762573.4, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.552741050720215, "optim/lr": 0.002935863693216431, "optim/total_tokens": 2657878016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361049, "loss/out": 4.332118034362793, "created_at": "2025-01-15T03:17:29.897088+00:00"} {"global_step": 10140, "acc_step": 0, "speed/wps": 12958.692921748836, "speed/FLOPS": 203534079446191.72, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0535920858383179, "optim/lr": 0.002935838904849294, "optim/total_tokens": 2658140160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436096, "loss/out": 4.292900085449219, "created_at": "2025-01-15T03:17:34.963625+00:00"} {"global_step": 10141, "acc_step": 0, "speed/wps": 12961.37548349996, "speed/FLOPS": 203576212764719.28, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.466749370098114, "optim/lr": 0.002935814111797481, "optim/total_tokens": 2658402304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489475, "loss/out": 4.304272174835205, "created_at": "2025-01-15T03:17:40.025805+00:00"} {"global_step": 10142, "acc_step": 0, "speed/wps": 12963.621253102006, "speed/FLOPS": 203611485662330.22, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6150230169296265, "optim/lr": 0.0029357893140610717, "optim/total_tokens": 2658664448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429977, "loss/out": 4.385842323303223, "created_at": "2025-01-15T03:17:45.091869+00:00"} {"global_step": 10143, "acc_step": 0, "speed/wps": 12957.920498417076, "speed/FLOPS": 203521947476345.56, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6228241920471191, "optim/lr": 0.002935764511640148, "optim/total_tokens": 2658926592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 4.282167434692383, "created_at": "2025-01-15T03:17:50.154634+00:00"} {"global_step": 10144, "acc_step": 0, "speed/wps": 12957.286556182196, "speed/FLOPS": 203511990542418.84, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0207998752593994, "optim/lr": 0.0029357397045347896, "optim/total_tokens": 2659188736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 4.328726768493652, "created_at": "2025-01-15T03:17:55.216451+00:00"} {"global_step": 10145, "acc_step": 0, "speed/wps": 12963.52334431602, "speed/FLOPS": 203609947870305.53, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9122794270515442, "optim/lr": 0.0029357148927450787, "optim/total_tokens": 2659450880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466155, "loss/out": 4.334813117980957, "created_at": "2025-01-15T03:18:00.275775+00:00"} {"global_step": 10146, "acc_step": 0, "speed/wps": 12960.376428130876, "speed/FLOPS": 203560521227301.88, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.6133323907852173, "optim/lr": 0.0029356900762710953, "optim/total_tokens": 2659713024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435567, "loss/out": 4.3466477394104, "created_at": "2025-01-15T03:18:05.336271+00:00"} {"global_step": 10147, "acc_step": 0, "speed/wps": 12965.28288779655, "speed/FLOPS": 203637583918533.38, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.4175127744674683, "optim/lr": 0.002935665255112921, "optim/total_tokens": 2659975168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294127, "loss/out": 4.46760368347168, "created_at": "2025-01-15T03:18:10.394664+00:00"} {"global_step": 10148, "acc_step": 0, "speed/wps": 12960.542210938076, "speed/FLOPS": 203563125074098.4, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1265554428100586, "optim/lr": 0.002935640429270636, "optim/total_tokens": 2660237312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 370682, "loss/out": 4.250065803527832, "created_at": "2025-01-15T03:18:15.455345+00:00"} {"global_step": 10149, "acc_step": 0, "speed/wps": 12961.191189580752, "speed/FLOPS": 203573318175471.72, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.488976240158081, "optim/lr": 0.0029356155987443227, "optim/total_tokens": 2660499456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428688, "loss/out": 4.454583168029785, "created_at": "2025-01-15T03:18:20.516508+00:00"} {"global_step": 10150, "acc_step": 0, "speed/wps": 12964.724748713024, "speed/FLOPS": 203628817577263.16, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.8155560493469238, "optim/lr": 0.0029355907635340603, "optim/total_tokens": 2660761600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 4.380012512207031, "created_at": "2025-01-15T03:18:25.581147+00:00"} {"global_step": 10151, "acc_step": 0, "speed/wps": 12957.266641284836, "speed/FLOPS": 203511677751589.88, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7313136458396912, "optim/lr": 0.0029355659236399313, "optim/total_tokens": 2661023744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 4.391119003295898, "created_at": "2025-01-15T03:18:30.642761+00:00"} {"global_step": 10152, "acc_step": 0, "speed/wps": 12961.007101431213, "speed/FLOPS": 203570426818119.6, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8333636522293091, "optim/lr": 0.002935541079062016, "optim/total_tokens": 2661285888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503221, "loss/out": 4.371689796447754, "created_at": "2025-01-15T03:18:35.707972+00:00"} {"global_step": 10153, "acc_step": 0, "speed/wps": 12966.280002372527, "speed/FLOPS": 203653244973128.12, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0212467908859253, "optim/lr": 0.0029355162298003957, "optim/total_tokens": 2661548032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440105, "loss/out": 4.457179546356201, "created_at": "2025-01-15T03:18:40.771349+00:00"} {"global_step": 10154, "acc_step": 0, "speed/wps": 12957.947552211212, "speed/FLOPS": 203522372393358.25, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6615980863571167, "optim/lr": 0.0029354913758551514, "optim/total_tokens": 2661810176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338244, "loss/out": 4.275848388671875, "created_at": "2025-01-15T03:18:45.837078+00:00"} {"global_step": 10155, "acc_step": 0, "speed/wps": 12961.158703056733, "speed/FLOPS": 203572807929970.84, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.020297884941101, "optim/lr": 0.002935466517226364, "optim/total_tokens": 2662072320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455769, "loss/out": 4.273437023162842, "created_at": "2025-01-15T03:18:50.898089+00:00"} {"global_step": 10156, "acc_step": 0, "speed/wps": 12958.286855705683, "speed/FLOPS": 203527701620990.78, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0317130088806152, "optim/lr": 0.0029354416539141156, "optim/total_tokens": 2662334464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485475, "loss/out": 4.33307409286499, "created_at": "2025-01-15T03:18:55.960404+00:00"} {"global_step": 10157, "acc_step": 0, "speed/wps": 12964.690393303434, "speed/FLOPS": 203628277978346.25, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0394037961959839, "optim/lr": 0.002935416785918486, "optim/total_tokens": 2662596608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 4.263189792633057, "created_at": "2025-01-15T03:19:01.020790+00:00"} {"global_step": 10158, "acc_step": 0, "speed/wps": 12963.964300168143, "speed/FLOPS": 203616873687899.97, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0252991914749146, "optim/lr": 0.0029353919132395568, "optim/total_tokens": 2662858752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430791, "loss/out": 4.3014068603515625, "created_at": "2025-01-15T03:19:06.080058+00:00"} {"global_step": 10159, "acc_step": 0, "speed/wps": 12961.234872848674, "speed/FLOPS": 203574004281221.53, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6739327311515808, "optim/lr": 0.0029353670358774095, "optim/total_tokens": 2663120896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449411, "loss/out": 4.230170249938965, "created_at": "2025-01-15T03:19:11.140292+00:00"} {"global_step": 10160, "acc_step": 0, "speed/wps": 12962.169614460472, "speed/FLOPS": 203588685682704.4, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9613316655158997, "optim/lr": 0.0029353421538321246, "optim/total_tokens": 2663383040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449135, "loss/out": 4.332205772399902, "created_at": "2025-01-15T03:19:16.199801+00:00"} {"global_step": 10161, "acc_step": 0, "speed/wps": 12959.914616559425, "speed/FLOPS": 203553267841976.62, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8158600926399231, "optim/lr": 0.002935317267103784, "optim/total_tokens": 2663645184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 4.250384330749512, "created_at": "2025-01-15T03:19:21.264031+00:00"} {"global_step": 10162, "acc_step": 0, "speed/wps": 12960.173649473407, "speed/FLOPS": 203557336309839.25, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0790504217147827, "optim/lr": 0.0029352923756924684, "optim/total_tokens": 2663907328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 4.337786674499512, "created_at": "2025-01-15T03:19:26.324893+00:00"} {"global_step": 10163, "acc_step": 0, "speed/wps": 12957.456497380057, "speed/FLOPS": 203514659702454.94, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9523510932922363, "optim/lr": 0.0029352674795982596, "optim/total_tokens": 2664169472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332063, "loss/out": 4.330789566040039, "created_at": "2025-01-15T03:19:31.386485+00:00"} {"global_step": 10164, "acc_step": 0, "speed/wps": 12967.581657135242, "speed/FLOPS": 203673689249837.66, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5740816593170166, "optim/lr": 0.0029352425788212386, "optim/total_tokens": 2664431616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415820, "loss/out": 4.305936813354492, "created_at": "2025-01-15T03:19:36.444248+00:00"} {"global_step": 10165, "acc_step": 0, "speed/wps": 12957.598717732184, "speed/FLOPS": 203516893468516.34, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7394521236419678, "optim/lr": 0.002935217673361486, "optim/total_tokens": 2664693760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 4.386394500732422, "created_at": "2025-01-15T03:19:41.515978+00:00"} {"global_step": 10166, "acc_step": 0, "speed/wps": 12964.416793826875, "speed/FLOPS": 203623980722601.8, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8250665664672852, "optim/lr": 0.0029351927632190835, "optim/total_tokens": 2664955904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444119, "loss/out": 4.187407493591309, "created_at": "2025-01-15T03:19:46.575349+00:00"} {"global_step": 10167, "acc_step": 0, "speed/wps": 12953.976380046432, "speed/FLOPS": 203459999677548.0, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6443864703178406, "optim/lr": 0.0029351678483941126, "optim/total_tokens": 2665218048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463248, "loss/out": 4.3126115798950195, "created_at": "2025-01-15T03:19:51.642776+00:00"} {"global_step": 10168, "acc_step": 0, "speed/wps": 12959.483077157847, "speed/FLOPS": 203546489922677.56, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6281278729438782, "optim/lr": 0.002935142928886654, "optim/total_tokens": 2665480192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432656, "loss/out": 4.317807197570801, "created_at": "2025-01-15T03:19:56.708342+00:00"} {"global_step": 10169, "acc_step": 0, "speed/wps": 12962.124150859707, "speed/FLOPS": 203587971614380.84, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.5899292230606079, "optim/lr": 0.00293511800469679, "optim/total_tokens": 2665742336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298942, "loss/out": 4.322073936462402, "created_at": "2025-01-15T03:20:01.771049+00:00"} {"global_step": 10170, "acc_step": 0, "speed/wps": 12965.402924936072, "speed/FLOPS": 203639469266757.9, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3476351499557495, "optim/lr": 0.0029350930758246007, "optim/total_tokens": 2666004480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 389084, "loss/out": 4.307716369628906, "created_at": "2025-01-15T03:20:06.830624+00:00"} {"global_step": 10171, "acc_step": 0, "speed/wps": 12966.224690432231, "speed/FLOPS": 203652376223099.2, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9851723909378052, "optim/lr": 0.002935068142270168, "optim/total_tokens": 2666266624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 396331, "loss/out": 4.292149066925049, "created_at": "2025-01-15T03:20:11.889839+00:00"} {"global_step": 10172, "acc_step": 0, "speed/wps": 12960.245365538398, "speed/FLOPS": 203558462709188.12, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0009, "optim/grad_norm": 2.0832815170288086, "optim/lr": 0.0029350432040335744, "optim/total_tokens": 2666528768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450462, "loss/out": 4.189216136932373, "created_at": "2025-01-15T03:20:16.951251+00:00"} {"global_step": 10173, "acc_step": 0, "speed/wps": 12962.63955262083, "speed/FLOPS": 203596066707275.53, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.9277622699737549, "optim/lr": 0.0029350182611148994, "optim/total_tokens": 2666790912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466683, "loss/out": 4.351563930511475, "created_at": "2025-01-15T03:20:22.015780+00:00"} {"global_step": 10174, "acc_step": 0, "speed/wps": 12965.104676792918, "speed/FLOPS": 203634784869833.78, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4485820531845093, "optim/lr": 0.0029349933135142253, "optim/total_tokens": 2667053056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440390, "loss/out": 4.3418073654174805, "created_at": "2025-01-15T03:20:27.078191+00:00"} {"global_step": 10175, "acc_step": 0, "speed/wps": 12965.904477713477, "speed/FLOPS": 203647346842332.78, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0682709217071533, "optim/lr": 0.0029349683612316328, "optim/total_tokens": 2667315200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486572, "loss/out": 4.495689868927002, "created_at": "2025-01-15T03:20:32.136568+00:00"} {"global_step": 10176, "acc_step": 0, "speed/wps": 12963.235327151444, "speed/FLOPS": 203605424164958.8, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1122963428497314, "optim/lr": 0.002934943404267204, "optim/total_tokens": 2667577344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454995, "loss/out": 4.302058696746826, "created_at": "2025-01-15T03:20:37.201628+00:00"} {"global_step": 10177, "acc_step": 0, "speed/wps": 12960.227881019367, "speed/FLOPS": 203558188090790.5, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.001, "optim/grad_norm": 1.246031403541565, "optim/lr": 0.0029349184426210205, "optim/total_tokens": 2667839488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311717, "loss/out": 4.192605018615723, "created_at": "2025-01-15T03:20:42.264538+00:00"} {"global_step": 10178, "acc_step": 0, "speed/wps": 12957.506549845793, "speed/FLOPS": 203515445845208.62, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.938427209854126, "optim/lr": 0.002934893476293164, "optim/total_tokens": 2668101632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418737, "loss/out": 4.260239601135254, "created_at": "2025-01-15T03:20:47.325952+00:00"} {"global_step": 10179, "acc_step": 0, "speed/wps": 12961.179911906547, "speed/FLOPS": 203573141044101.2, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1154682636260986, "optim/lr": 0.002934868505283715, "optim/total_tokens": 2668363776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451808, "loss/out": 4.236522674560547, "created_at": "2025-01-15T03:20:52.386125+00:00"} {"global_step": 10180, "acc_step": 0, "speed/wps": 12960.579165392086, "speed/FLOPS": 203563705494579.34, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8258405923843384, "optim/lr": 0.002934843529592754, "optim/total_tokens": 2668625920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479906, "loss/out": 4.3397040367126465, "created_at": "2025-01-15T03:20:57.451978+00:00"} {"global_step": 10181, "acc_step": 0, "speed/wps": 12959.895614902736, "speed/FLOPS": 203552969394847.3, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6284888982772827, "optim/lr": 0.0029348185492203654, "optim/total_tokens": 2668888064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442784, "loss/out": 4.303882598876953, "created_at": "2025-01-15T03:21:02.513548+00:00"} {"global_step": 10182, "acc_step": 0, "speed/wps": 12958.443264147138, "speed/FLOPS": 203530158230491.56, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9632183909416199, "optim/lr": 0.002934793564166629, "optim/total_tokens": 2669150208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 4.330302715301514, "created_at": "2025-01-15T03:21:07.574771+00:00"} {"global_step": 10183, "acc_step": 0, "speed/wps": 12963.106673557932, "speed/FLOPS": 203603403483484.16, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.2711496353149414, "optim/lr": 0.002934768574431626, "optim/total_tokens": 2669412352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 4.215856552124023, "created_at": "2025-01-15T03:21:12.657919+00:00"} {"global_step": 10184, "acc_step": 0, "speed/wps": 12966.365719968568, "speed/FLOPS": 203654591285762.25, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8330334424972534, "optim/lr": 0.0029347435800154385, "optim/total_tokens": 2669674496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295180, "loss/out": 4.310878276824951, "created_at": "2025-01-15T03:21:17.716110+00:00"} {"global_step": 10185, "acc_step": 0, "speed/wps": 12955.401964829012, "speed/FLOPS": 203482390445517.2, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4334683418273926, "optim/lr": 0.002934718580918148, "optim/total_tokens": 2669936640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350526, "loss/out": 4.177706718444824, "created_at": "2025-01-15T03:21:22.786033+00:00"} {"global_step": 10186, "acc_step": 0, "speed/wps": 12963.77982753899, "speed/FLOPS": 203613976291767.2, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.6399718523025513, "optim/lr": 0.0029346935771398363, "optim/total_tokens": 2670198784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 4.229537010192871, "created_at": "2025-01-15T03:21:27.851643+00:00"} {"global_step": 10187, "acc_step": 0, "speed/wps": 12957.150963353342, "speed/FLOPS": 203509860870717.62, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0008, "optim/grad_norm": 7.560225009918213, "optim/lr": 0.002934668568680584, "optim/total_tokens": 2670460928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458972, "loss/out": 4.255539417266846, "created_at": "2025-01-15T03:21:32.919845+00:00"} {"global_step": 10188, "acc_step": 0, "speed/wps": 12942.833521060189, "speed/FLOPS": 203284985765276.22, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.5435221195220947, "optim/lr": 0.002934643555540474, "optim/total_tokens": 2670723072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 4.325778961181641, "created_at": "2025-01-15T03:21:37.992651+00:00"} {"global_step": 10189, "acc_step": 0, "speed/wps": 12968.465895836698, "speed/FLOPS": 203687577433715.3, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1387065649032593, "optim/lr": 0.0029346185377195867, "optim/total_tokens": 2670985216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432637, "loss/out": 4.300217151641846, "created_at": "2025-01-15T03:21:43.050431+00:00"} {"global_step": 10190, "acc_step": 0, "speed/wps": 12960.712674392938, "speed/FLOPS": 203565802436898.34, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.349759578704834, "optim/lr": 0.002934593515218005, "optim/total_tokens": 2671247360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303689, "loss/out": 4.228483200073242, "created_at": "2025-01-15T03:21:48.115869+00:00"} {"global_step": 10191, "acc_step": 0, "speed/wps": 12960.398610713668, "speed/FLOPS": 203560869635246.7, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.630183219909668, "optim/lr": 0.00293456848803581, "optim/total_tokens": 2671509504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429438, "loss/out": 4.3651123046875, "created_at": "2025-01-15T03:21:53.177368+00:00"} {"global_step": 10192, "acc_step": 0, "speed/wps": 12964.378933743446, "speed/FLOPS": 203623386077966.34, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3603200912475586, "optim/lr": 0.0029345434561730823, "optim/total_tokens": 2671771648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449411, "loss/out": 4.361673355102539, "created_at": "2025-01-15T03:21:58.236355+00:00"} {"global_step": 10193, "acc_step": 0, "speed/wps": 12956.912392045435, "speed/FLOPS": 203506113780495.62, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0981050729751587, "optim/lr": 0.002934518419629905, "optim/total_tokens": 2672033792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 4.393991947174072, "created_at": "2025-01-15T03:22:03.305503+00:00"} {"global_step": 10194, "acc_step": 0, "speed/wps": 12962.039635396277, "speed/FLOPS": 203586644182891.28, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0011, "optim/grad_norm": 1.3060388565063477, "optim/lr": 0.002934493378406359, "optim/total_tokens": 2672295936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 4.419195652008057, "created_at": "2025-01-15T03:22:08.370308+00:00"} {"global_step": 10195, "acc_step": 0, "speed/wps": 12963.436809340683, "speed/FLOPS": 203608588719606.34, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1191118955612183, "optim/lr": 0.002934468332502527, "optim/total_tokens": 2672558080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450746, "loss/out": 4.294739723205566, "created_at": "2025-01-15T03:22:13.429558+00:00"} {"global_step": 10196, "acc_step": 0, "speed/wps": 12962.545140097003, "speed/FLOPS": 203594583828852.44, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5991261005401611, "optim/lr": 0.002934443281918489, "optim/total_tokens": 2672820224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335181, "loss/out": 4.380706787109375, "created_at": "2025-01-15T03:22:18.489562+00:00"} {"global_step": 10197, "acc_step": 0, "speed/wps": 12968.401839445807, "speed/FLOPS": 203686571340073.66, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1442152261734009, "optim/lr": 0.0029344182266543284, "optim/total_tokens": 2673082368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322311, "loss/out": 4.266077995300293, "created_at": "2025-01-15T03:22:23.547915+00:00"} {"global_step": 10198, "acc_step": 0, "speed/wps": 12967.501395739362, "speed/FLOPS": 203672428634324.62, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9243684411048889, "optim/lr": 0.002934393166710126, "optim/total_tokens": 2673344512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318405, "loss/out": 4.367006301879883, "created_at": "2025-01-15T03:22:28.605812+00:00"} {"global_step": 10199, "acc_step": 0, "speed/wps": 12967.576794155808, "speed/FLOPS": 203673612870063.2, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7258962988853455, "optim/lr": 0.002934368102085964, "optim/total_tokens": 2673606656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308039, "loss/out": 4.267621994018555, "created_at": "2025-01-15T03:22:33.663438+00:00"} {"global_step": 10200, "acc_step": 0, "speed/wps": 12963.163386949107, "speed/FLOPS": 203604294245224.9, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0904382467269897, "optim/lr": 0.0029343430327819237, "optim/total_tokens": 2673868800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438471, "loss/out": 4.132068157196045, "created_at": "2025-01-15T03:22:38.725867+00:00"} {"global_step": 10201, "acc_step": 0, "speed/wps": 12962.363869232146, "speed/FLOPS": 203591736720827.56, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0546646118164062, "optim/lr": 0.0029343179587980875, "optim/total_tokens": 2674130944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 4.187431335449219, "created_at": "2025-01-15T03:22:43.790797+00:00"} {"global_step": 10202, "acc_step": 0, "speed/wps": 12961.74359761991, "speed/FLOPS": 203581994502814.9, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7953580021858215, "optim/lr": 0.0029342928801345367, "optim/total_tokens": 2674393088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425492, "loss/out": 4.301522731781006, "created_at": "2025-01-15T03:22:48.856947+00:00"} {"global_step": 10203, "acc_step": 0, "speed/wps": 12955.374399310042, "speed/FLOPS": 203481957491162.78, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.2378665208816528, "optim/lr": 0.0029342677967913538, "optim/total_tokens": 2674655232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 4.370398044586182, "created_at": "2025-01-15T03:22:53.920066+00:00"} {"global_step": 10204, "acc_step": 0, "speed/wps": 12963.534972354806, "speed/FLOPS": 203610130504633.28, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8688748478889465, "optim/lr": 0.00293424270876862, "optim/total_tokens": 2674917376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440390, "loss/out": 4.294313430786133, "created_at": "2025-01-15T03:22:58.979400+00:00"} {"global_step": 10205, "acc_step": 0, "speed/wps": 12963.29206793734, "speed/FLOPS": 203606315356971.34, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.001, "optim/grad_norm": 0.7054471969604492, "optim/lr": 0.002934217616066417, "optim/total_tokens": 2675179520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 286083, "loss/out": 4.274646759033203, "created_at": "2025-01-15T03:23:04.043382+00:00"} {"global_step": 10206, "acc_step": 0, "speed/wps": 12962.841553573631, "speed/FLOPS": 203599239409817.7, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1647610664367676, "optim/lr": 0.002934192518684827, "optim/total_tokens": 2675441664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444591, "loss/out": 4.3526716232299805, "created_at": "2025-01-15T03:23:09.103251+00:00"} {"global_step": 10207, "acc_step": 0, "speed/wps": 12967.431111975702, "speed/FLOPS": 203671324731236.34, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7070885300636292, "optim/lr": 0.002934167416623932, "optim/total_tokens": 2675703808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439583, "loss/out": 4.326043128967285, "created_at": "2025-01-15T03:23:14.163574+00:00"} {"global_step": 10208, "acc_step": 0, "speed/wps": 12961.32411196228, "speed/FLOPS": 203575405904128.12, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5623273849487305, "optim/lr": 0.0029341423098838137, "optim/total_tokens": 2675965952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 4.232810020446777, "created_at": "2025-01-15T03:23:19.232393+00:00"} {"global_step": 10209, "acc_step": 0, "speed/wps": 12962.677512387596, "speed/FLOPS": 203596662917574.78, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.001, "optim/grad_norm": 0.6139003038406372, "optim/lr": 0.0029341171984645542, "optim/total_tokens": 2676228096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471217, "loss/out": 4.382628917694092, "created_at": "2025-01-15T03:23:24.292789+00:00"} {"global_step": 10210, "acc_step": 0, "speed/wps": 12961.364740278277, "speed/FLOPS": 203576044027660.0, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.001, "optim/grad_norm": 1.1448853015899658, "optim/lr": 0.002934092082366235, "optim/total_tokens": 2676490240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 4.291368007659912, "created_at": "2025-01-15T03:23:29.353737+00:00"} {"global_step": 10211, "acc_step": 0, "speed/wps": 12965.725726433478, "speed/FLOPS": 203644539307850.4, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5320470929145813, "optim/lr": 0.002934066961588939, "optim/total_tokens": 2676752384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 4.3307647705078125, "created_at": "2025-01-15T03:23:34.412282+00:00"} {"global_step": 10212, "acc_step": 0, "speed/wps": 12961.349098179347, "speed/FLOPS": 203575798347001.8, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4522593021392822, "optim/lr": 0.0029340418361327473, "optim/total_tokens": 2677014528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428948, "loss/out": 4.469870567321777, "created_at": "2025-01-15T03:23:39.476085+00:00"} {"global_step": 10213, "acc_step": 0, "speed/wps": 12962.045444500833, "speed/FLOPS": 203586735422860.88, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0300121307373047, "optim/lr": 0.0029340167059977423, "optim/total_tokens": 2677276672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330091, "loss/out": 4.242181777954102, "created_at": "2025-01-15T03:23:44.535680+00:00"} {"global_step": 10214, "acc_step": 0, "speed/wps": 12967.002854379596, "speed/FLOPS": 203664598357201.88, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.675873279571533, "optim/lr": 0.002933991571184005, "optim/total_tokens": 2677538816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444359, "loss/out": 4.3274078369140625, "created_at": "2025-01-15T03:23:49.609523+00:00"} {"global_step": 10215, "acc_step": 0, "speed/wps": 12963.508229568502, "speed/FLOPS": 203609710472425.94, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2726479768753052, "optim/lr": 0.0029339664316916187, "optim/total_tokens": 2677800960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458694, "loss/out": 4.257708549499512, "created_at": "2025-01-15T03:23:54.668841+00:00"} {"global_step": 10216, "acc_step": 0, "speed/wps": 12967.169963282646, "speed/FLOPS": 203667223032155.97, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0854465961456299, "optim/lr": 0.002933941287520665, "optim/total_tokens": 2678063104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432875, "loss/out": 4.404513359069824, "created_at": "2025-01-15T03:23:59.731880+00:00"} {"global_step": 10217, "acc_step": 0, "speed/wps": 12967.576072475633, "speed/FLOPS": 203673601535084.34, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5588387250900269, "optim/lr": 0.002933916138671226, "optim/total_tokens": 2678325248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 4.314970970153809, "created_at": "2025-01-15T03:24:04.797948+00:00"} {"global_step": 10218, "acc_step": 0, "speed/wps": 12968.520342593487, "speed/FLOPS": 203688432594848.38, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9776596426963806, "optim/lr": 0.0029338909851433833, "optim/total_tokens": 2678587392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430506, "loss/out": 4.293074607849121, "created_at": "2025-01-15T03:24:09.859897+00:00"} {"global_step": 10219, "acc_step": 0, "speed/wps": 12962.779362761554, "speed/FLOPS": 203598262617654.88, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0592361688613892, "optim/lr": 0.0029338658269372196, "optim/total_tokens": 2678849536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301104, "loss/out": 4.2462968826293945, "created_at": "2025-01-15T03:24:14.926209+00:00"} {"global_step": 10220, "acc_step": 0, "speed/wps": 12969.708811147337, "speed/FLOPS": 203707099126613.5, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0885241031646729, "optim/lr": 0.0029338406640528163, "optim/total_tokens": 2679111680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455517, "loss/out": 4.252084732055664, "created_at": "2025-01-15T03:24:19.983606+00:00"} {"global_step": 10221, "acc_step": 0, "speed/wps": 12961.253613438068, "speed/FLOPS": 203574298627930.25, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5216124057769775, "optim/lr": 0.0029338154964902567, "optim/total_tokens": 2679373824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 4.288083076477051, "created_at": "2025-01-15T03:24:25.043831+00:00"} {"global_step": 10222, "acc_step": 0, "speed/wps": 12960.787110632793, "speed/FLOPS": 203566971560331.38, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9652538299560547, "optim/lr": 0.0029337903242496214, "optim/total_tokens": 2679635968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446742, "loss/out": 4.278669834136963, "created_at": "2025-01-15T03:24:30.108436+00:00"} {"global_step": 10223, "acc_step": 0, "speed/wps": 12962.878597555102, "speed/FLOPS": 203599821236450.44, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8022195100784302, "optim/lr": 0.0029337651473309934, "optim/total_tokens": 2679898112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468540, "loss/out": 4.241422653198242, "created_at": "2025-01-15T03:24:35.168034+00:00"} {"global_step": 10224, "acc_step": 0, "speed/wps": 12963.062043961672, "speed/FLOPS": 203602702514347.4, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.001, "optim/grad_norm": 0.8158519864082336, "optim/lr": 0.002933739965734454, "optim/total_tokens": 2680160256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 4.13138484954834, "created_at": "2025-01-15T03:24:40.227230+00:00"} {"global_step": 10225, "acc_step": 0, "speed/wps": 12966.523757066912, "speed/FLOPS": 203657073475557.5, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0992881059646606, "optim/lr": 0.0029337147794600868, "optim/total_tokens": 2680422400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 4.417540073394775, "created_at": "2025-01-15T03:24:45.289900+00:00"} {"global_step": 10226, "acc_step": 0, "speed/wps": 12948.056981902386, "speed/FLOPS": 203367027395590.3, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1221671104431152, "optim/lr": 0.002933689588507973, "optim/total_tokens": 2680684544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318319, "loss/out": 4.3217644691467285, "created_at": "2025-01-15T03:24:50.358842+00:00"} {"global_step": 10227, "acc_step": 0, "speed/wps": 12961.014955484756, "speed/FLOPS": 203570550176823.1, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3317065238952637, "optim/lr": 0.002933664392878195, "optim/total_tokens": 2680946688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377871, "loss/out": 4.312381744384766, "created_at": "2025-01-15T03:24:55.419172+00:00"} {"global_step": 10228, "acc_step": 0, "speed/wps": 12968.126049518556, "speed/FLOPS": 203682239680291.12, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.600443422794342, "optim/lr": 0.002933639192570835, "optim/total_tokens": 2681208832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444359, "loss/out": 4.299604892730713, "created_at": "2025-01-15T03:25:00.476529+00:00"} {"global_step": 10229, "acc_step": 0, "speed/wps": 12962.940020395898, "speed/FLOPS": 203600785966568.97, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.5633225440979004, "optim/lr": 0.002933613987585975, "optim/total_tokens": 2681470976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 4.3340559005737305, "created_at": "2025-01-15T03:25:05.541770+00:00"} {"global_step": 10230, "acc_step": 0, "speed/wps": 12961.92099518502, "speed/FLOPS": 203584780775344.75, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2567857503890991, "optim/lr": 0.0029335887779236977, "optim/total_tokens": 2681733120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442208, "loss/out": 4.266676902770996, "created_at": "2025-01-15T03:25:10.602489+00:00"} {"global_step": 10231, "acc_step": 0, "speed/wps": 12961.745803688937, "speed/FLOPS": 203582029152160.4, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0468919277191162, "optim/lr": 0.002933563563584085, "optim/total_tokens": 2681995264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454995, "loss/out": 4.284815788269043, "created_at": "2025-01-15T03:25:15.664689+00:00"} {"global_step": 10232, "acc_step": 0, "speed/wps": 12965.94605454099, "speed/FLOPS": 203647999863541.4, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.1283977031707764, "optim/lr": 0.0029335383445672192, "optim/total_tokens": 2682257408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468224, "loss/out": 4.362945556640625, "created_at": "2025-01-15T03:25:20.722872+00:00"} {"global_step": 10233, "acc_step": 0, "speed/wps": 12968.212410509213, "speed/FLOPS": 203683596098321.8, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3410669565200806, "optim/lr": 0.002933513120873182, "optim/total_tokens": 2682519552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436666, "loss/out": 4.3369245529174805, "created_at": "2025-01-15T03:25:25.785217+00:00"} {"global_step": 10234, "acc_step": 0, "speed/wps": 12960.409469298236, "speed/FLOPS": 203561040184238.6, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7713319063186646, "optim/lr": 0.0029334878925020573, "optim/total_tokens": 2682781696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307800, "loss/out": 4.29512882232666, "created_at": "2025-01-15T03:25:30.848114+00:00"} {"global_step": 10235, "acc_step": 0, "speed/wps": 12960.31871987565, "speed/FLOPS": 203559614839854.1, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5807892680168152, "optim/lr": 0.0029334626594539268, "optim/total_tokens": 2683043840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444641, "loss/out": 4.320827960968018, "created_at": "2025-01-15T03:25:35.908350+00:00"} {"global_step": 10236, "acc_step": 0, "speed/wps": 12963.252017392497, "speed/FLOPS": 203605686308129.34, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0658320188522339, "optim/lr": 0.002933437421728871, "optim/total_tokens": 2683305984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444881, "loss/out": 4.186872482299805, "created_at": "2025-01-15T03:25:40.972199+00:00"} {"global_step": 10237, "acc_step": 0, "speed/wps": 12964.233889465493, "speed/FLOPS": 203621107958269.38, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1027276515960693, "optim/lr": 0.002933412179326975, "optim/total_tokens": 2683568128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 4.318111419677734, "created_at": "2025-01-15T03:25:46.032118+00:00"} {"global_step": 10238, "acc_step": 0, "speed/wps": 12960.926119293756, "speed/FLOPS": 203569154882369.7, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2054252624511719, "optim/lr": 0.0029333869322483194, "optim/total_tokens": 2683830272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 4.376238822937012, "created_at": "2025-01-15T03:25:51.093271+00:00"} {"global_step": 10239, "acc_step": 0, "speed/wps": 12960.663719102267, "speed/FLOPS": 203565033526786.72, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.108130931854248, "optim/lr": 0.0029333616804929873, "optim/total_tokens": 2684092416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 4.343841552734375, "created_at": "2025-01-15T03:25:56.153760+00:00"} {"global_step": 10240, "acc_step": 0, "speed/wps": 12963.998384234352, "speed/FLOPS": 203617409024996.2, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1435763835906982, "optim/lr": 0.0029333364240610605, "optim/total_tokens": 2684354560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444359, "loss/out": 4.307412147521973, "created_at": "2025-01-15T03:26:01.213258+00:00"} {"global_step": 10241, "acc_step": 0, "speed/wps": 12956.614560633812, "speed/FLOPS": 203501435928912.84, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5129926204681396, "optim/lr": 0.0029333111629526215, "optim/total_tokens": 2684616704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464535, "loss/out": 4.350205421447754, "created_at": "2025-01-15T03:26:06.278115+00:00"} {"global_step": 10242, "acc_step": 0, "speed/wps": 12962.418171640451, "speed/FLOPS": 203592589614769.34, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.221571922302246, "optim/lr": 0.0029332858971677536, "optim/total_tokens": 2684878848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320967, "loss/out": 4.247300624847412, "created_at": "2025-01-15T03:26:11.339204+00:00"} {"global_step": 10243, "acc_step": 0, "speed/wps": 12950.107262049707, "speed/FLOPS": 203399229862684.16, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.917063295841217, "optim/lr": 0.0029332606267065383, "optim/total_tokens": 2685140992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457106, "loss/out": 4.35208797454834, "created_at": "2025-01-15T03:26:16.412104+00:00"} {"global_step": 10244, "acc_step": 0, "speed/wps": 12962.602218689522, "speed/FLOPS": 203595480326582.16, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.297432780265808, "optim/lr": 0.0029332353515690583, "optim/total_tokens": 2685403136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 4.486332893371582, "created_at": "2025-01-15T03:26:21.474585+00:00"} {"global_step": 10245, "acc_step": 0, "speed/wps": 12967.887370222663, "speed/FLOPS": 203678490893969.38, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0448195934295654, "optim/lr": 0.0029332100717553967, "optim/total_tokens": 2685665280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469590, "loss/out": 4.270265102386475, "created_at": "2025-01-15T03:26:26.536524+00:00"} {"global_step": 10246, "acc_step": 0, "speed/wps": 12969.453618318248, "speed/FLOPS": 203703090972560.97, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0650581121444702, "optim/lr": 0.0029331847872656343, "optim/total_tokens": 2685927424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453375, "loss/out": 4.265508651733398, "created_at": "2025-01-15T03:26:31.593597+00:00"} {"global_step": 10247, "acc_step": 0, "speed/wps": 12958.44803198531, "speed/FLOPS": 203530233115941.75, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0702632665634155, "optim/lr": 0.0029331594980998555, "optim/total_tokens": 2686189568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 4.271862030029297, "created_at": "2025-01-15T03:26:36.654604+00:00"} {"global_step": 10248, "acc_step": 0, "speed/wps": 12964.657803246559, "speed/FLOPS": 203627766106719.6, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0588537454605103, "optim/lr": 0.0029331342042581417, "optim/total_tokens": 2686451712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432146, "loss/out": 4.300111770629883, "created_at": "2025-01-15T03:26:41.717987+00:00"} {"global_step": 10249, "acc_step": 0, "speed/wps": 12964.010841884663, "speed/FLOPS": 203617604689512.94, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6468380689620972, "optim/lr": 0.0029331089057405757, "optim/total_tokens": 2686713856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296001, "loss/out": 4.329507827758789, "created_at": "2025-01-15T03:26:46.782361+00:00"} {"global_step": 10250, "acc_step": 0, "speed/wps": 12964.362078183472, "speed/FLOPS": 203623121338235.0, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1334872245788574, "optim/lr": 0.00293308360254724, "optim/total_tokens": 2686976000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 382950, "loss/out": 4.307528495788574, "created_at": "2025-01-15T03:26:51.842139+00:00"} {"global_step": 10251, "acc_step": 0, "speed/wps": 12962.928228655397, "speed/FLOPS": 203600600761080.84, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.001, "optim/grad_norm": 1.3281396627426147, "optim/lr": 0.0029330582946782173, "optim/total_tokens": 2687238144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 399017, "loss/out": 4.313941955566406, "created_at": "2025-01-15T03:26:56.902459+00:00"} {"global_step": 10252, "acc_step": 0, "speed/wps": 12965.437171844665, "speed/FLOPS": 203640007161517.53, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.284380555152893, "optim/lr": 0.00293303298213359, "optim/total_tokens": 2687500288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440108, "loss/out": 4.399177074432373, "created_at": "2025-01-15T03:27:01.965856+00:00"} {"global_step": 10253, "acc_step": 0, "speed/wps": 12957.906422757782, "speed/FLOPS": 203521726398774.34, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3004343509674072, "optim/lr": 0.002933007664913441, "optim/total_tokens": 2687762432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498247, "loss/out": 4.302217483520508, "created_at": "2025-01-15T03:27:07.031469+00:00"} {"global_step": 10254, "acc_step": 0, "speed/wps": 12963.589363386238, "speed/FLOPS": 203610984790523.03, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8205982446670532, "optim/lr": 0.0029329823430178526, "optim/total_tokens": 2688024576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437432, "loss/out": 4.303989410400391, "created_at": "2025-01-15T03:27:12.091492+00:00"} {"global_step": 10255, "acc_step": 0, "speed/wps": 12955.344324716089, "speed/FLOPS": 203481485128337.9, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9842212796211243, "optim/lr": 0.002932957016446907, "optim/total_tokens": 2688286720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460840, "loss/out": 4.272387504577637, "created_at": "2025-01-15T03:27:17.158266+00:00"} {"global_step": 10256, "acc_step": 0, "speed/wps": 12958.189070431537, "speed/FLOPS": 203526165768889.28, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.944637656211853, "optim/lr": 0.0029329316852006885, "optim/total_tokens": 2688548864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328746, "loss/out": 4.269327163696289, "created_at": "2025-01-15T03:27:22.219510+00:00"} {"global_step": 10257, "acc_step": 0, "speed/wps": 12961.100211967814, "speed/FLOPS": 203571889247044.2, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1141350269317627, "optim/lr": 0.0029329063492792777, "optim/total_tokens": 2688811008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350820, "loss/out": 4.364925384521484, "created_at": "2025-01-15T03:27:27.279621+00:00"} {"global_step": 10258, "acc_step": 0, "speed/wps": 12969.059753705636, "speed/FLOPS": 203696904787593.22, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5618168115615845, "optim/lr": 0.002932881008682758, "optim/total_tokens": 2689073152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443543, "loss/out": 4.497549057006836, "created_at": "2025-01-15T03:27:32.341464+00:00"} {"global_step": 10259, "acc_step": 0, "speed/wps": 12960.529416214662, "speed/FLOPS": 203562924115385.38, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0423673391342163, "optim/lr": 0.002932855663411213, "optim/total_tokens": 2689335296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 4.188509941101074, "created_at": "2025-01-15T03:27:37.405089+00:00"} {"global_step": 10260, "acc_step": 0, "speed/wps": 12954.536381953953, "speed/FLOPS": 203468795277028.88, "speed/curr_iter_time": 1.2771, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6046339273452759, "optim/lr": 0.0029328303134647238, "optim/total_tokens": 2689597440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443543, "loss/out": 4.1211652755737305, "created_at": "2025-01-15T03:27:42.470658+00:00"} {"global_step": 10261, "acc_step": 0, "speed/wps": 12962.792336244856, "speed/FLOPS": 203598466384037.56, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.972252607345581, "optim/lr": 0.0029328049588433743, "optim/total_tokens": 2689859584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465057, "loss/out": 4.3629913330078125, "created_at": "2025-01-15T03:27:47.533425+00:00"} {"global_step": 10262, "acc_step": 0, "speed/wps": 12966.422308968322, "speed/FLOPS": 203655480093764.84, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1136597394943237, "optim/lr": 0.002932779599547247, "optim/total_tokens": 2690121728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 4.255404472351074, "created_at": "2025-01-15T03:27:52.595325+00:00"} {"global_step": 10263, "acc_step": 0, "speed/wps": 12957.023885826116, "speed/FLOPS": 203507864943529.72, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9942702651023865, "optim/lr": 0.002932754235576424, "optim/total_tokens": 2690383872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437201, "loss/out": 4.3087615966796875, "created_at": "2025-01-15T03:27:57.656850+00:00"} {"global_step": 10264, "acc_step": 0, "speed/wps": 12965.411527257076, "speed/FLOPS": 203639604378030.1, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8392755389213562, "optim/lr": 0.002932728866930989, "optim/total_tokens": 2690646016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442784, "loss/out": 4.198443412780762, "created_at": "2025-01-15T03:28:02.715061+00:00"} {"global_step": 10265, "acc_step": 0, "speed/wps": 12961.47330254639, "speed/FLOPS": 203577749147260.9, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1121569871902466, "optim/lr": 0.0029327034936110245, "optim/total_tokens": 2690908160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429739, "loss/out": 4.4281158447265625, "created_at": "2025-01-15T03:28:07.775290+00:00"} {"global_step": 10266, "acc_step": 0, "speed/wps": 12960.213621325054, "speed/FLOPS": 203557964122696.44, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5730512142181396, "optim/lr": 0.0029326781156166124, "optim/total_tokens": 2691170304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363341, "loss/out": 4.341694355010986, "created_at": "2025-01-15T03:28:12.843482+00:00"} {"global_step": 10267, "acc_step": 0, "speed/wps": 12963.29047746915, "speed/FLOPS": 203606290376482.88, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.415924072265625, "optim/lr": 0.0029326527329478365, "optim/total_tokens": 2691432448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439820, "loss/out": 4.448216438293457, "created_at": "2025-01-15T03:28:17.906683+00:00"} {"global_step": 10268, "acc_step": 0, "speed/wps": 12965.605296307964, "speed/FLOPS": 203642647787240.88, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.081535816192627, "optim/lr": 0.0029326273456047793, "optim/total_tokens": 2691694592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492851, "loss/out": 4.1813201904296875, "created_at": "2025-01-15T03:28:22.967232+00:00"} {"global_step": 10269, "acc_step": 0, "speed/wps": 12955.443777248785, "speed/FLOPS": 203483047167026.03, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7997121810913086, "optim/lr": 0.002932601953587524, "optim/total_tokens": 2691956736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427351, "loss/out": 4.392189979553223, "created_at": "2025-01-15T03:28:28.030712+00:00"} {"global_step": 10270, "acc_step": 0, "speed/wps": 12958.66706280836, "speed/FLOPS": 203533673295997.62, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.6463756561279297, "optim/lr": 0.002932576556896152, "optim/total_tokens": 2692218880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481758, "loss/out": 4.3524580001831055, "created_at": "2025-01-15T03:28:33.092799+00:00"} {"global_step": 10271, "acc_step": 0, "speed/wps": 12962.570056999411, "speed/FLOPS": 203594975183044.16, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7299138307571411, "optim/lr": 0.0029325511555307483, "optim/total_tokens": 2692481024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429217, "loss/out": 4.196104049682617, "created_at": "2025-01-15T03:28:38.159437+00:00"} {"global_step": 10272, "acc_step": 0, "speed/wps": 12957.620373045165, "speed/FLOPS": 203517233594964.22, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8272972106933594, "optim/lr": 0.002932525749491395, "optim/total_tokens": 2692743168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351583, "loss/out": 4.2677459716796875, "created_at": "2025-01-15T03:28:43.221717+00:00"} {"global_step": 10273, "acc_step": 0, "speed/wps": 12954.971820621693, "speed/FLOPS": 203475634439661.16, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0011, "optim/grad_norm": 1.2982949018478394, "optim/lr": 0.002932500338778174, "optim/total_tokens": 2693005312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445695, "loss/out": 4.188389778137207, "created_at": "2025-01-15T03:28:48.286230+00:00"} {"global_step": 10274, "acc_step": 0, "speed/wps": 12962.988109332126, "speed/FLOPS": 203601541269393.53, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4089269638061523, "optim/lr": 0.0029324749233911687, "optim/total_tokens": 2693267456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488376, "loss/out": 4.292513847351074, "created_at": "2025-01-15T03:28:53.346638+00:00"} {"global_step": 10275, "acc_step": 0, "speed/wps": 12966.1502753414, "speed/FLOPS": 203651207431840.66, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.684059977531433, "optim/lr": 0.0029324495033304625, "optim/total_tokens": 2693529600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448077, "loss/out": 4.238766670227051, "created_at": "2025-01-15T03:28:58.404869+00:00"} {"global_step": 10276, "acc_step": 0, "speed/wps": 12962.409929068137, "speed/FLOPS": 203592460153844.66, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.088158130645752, "optim/lr": 0.0029324240785961386, "optim/total_tokens": 2693791744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445695, "loss/out": 4.217933654785156, "created_at": "2025-01-15T03:29:03.465052+00:00"} {"global_step": 10277, "acc_step": 0, "speed/wps": 12963.295048275178, "speed/FLOPS": 203606362167272.5, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.027870535850525, "optim/lr": 0.002932398649188279, "optim/total_tokens": 2694053888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452081, "loss/out": 4.268811225891113, "created_at": "2025-01-15T03:29:08.525431+00:00"} {"global_step": 10278, "acc_step": 0, "speed/wps": 12959.333638830383, "speed/FLOPS": 203544142788394.97, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2185924053192139, "optim/lr": 0.002932373215106967, "optim/total_tokens": 2694316032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 4.2761969566345215, "created_at": "2025-01-15T03:29:13.591029+00:00"} {"global_step": 10279, "acc_step": 0, "speed/wps": 12955.470567869763, "speed/FLOPS": 203483467950542.62, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.2924753427505493, "optim/lr": 0.002932347776352286, "optim/total_tokens": 2694578176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427589, "loss/out": 4.413736820220947, "created_at": "2025-01-15T03:29:18.656893+00:00"} {"global_step": 10280, "acc_step": 0, "speed/wps": 12954.632588155644, "speed/FLOPS": 203470306327627.6, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.100999355316162, "optim/lr": 0.002932322332924318, "optim/total_tokens": 2694840320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351470, "loss/out": 4.230815410614014, "created_at": "2025-01-15T03:29:23.722417+00:00"} {"global_step": 10281, "acc_step": 0, "speed/wps": 12961.707453052628, "speed/FLOPS": 203581426802717.88, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6109488010406494, "optim/lr": 0.0029322968848231475, "optim/total_tokens": 2695102464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 4.389534950256348, "created_at": "2025-01-15T03:29:28.788426+00:00"} {"global_step": 10282, "acc_step": 0, "speed/wps": 12962.971436486438, "speed/FLOPS": 203601279399441.12, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3423212766647339, "optim/lr": 0.0029322714320488565, "optim/total_tokens": 2695364608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495864, "loss/out": 4.387470245361328, "created_at": "2025-01-15T03:29:33.849168+00:00"} {"global_step": 10283, "acc_step": 0, "speed/wps": 12960.94083412862, "speed/FLOPS": 203569385999071.56, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1575181484222412, "optim/lr": 0.002932245974601528, "optim/total_tokens": 2695626752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432953, "loss/out": 4.309547424316406, "created_at": "2025-01-15T03:29:38.912194+00:00"} {"global_step": 10284, "acc_step": 0, "speed/wps": 12959.361703302146, "speed/FLOPS": 203544583579488.62, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8035679459571838, "optim/lr": 0.0029322205124812455, "optim/total_tokens": 2695888896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 4.426431655883789, "created_at": "2025-01-15T03:29:43.973054+00:00"} {"global_step": 10285, "acc_step": 0, "speed/wps": 12964.199125848229, "speed/FLOPS": 203620561947890.03, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1107491254806519, "optim/lr": 0.002932195045688092, "optim/total_tokens": 2696151040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441687, "loss/out": 4.315956115722656, "created_at": "2025-01-15T03:29:49.034173+00:00"} {"global_step": 10286, "acc_step": 0, "speed/wps": 12964.079342670679, "speed/FLOPS": 203618680588487.25, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0439995527267456, "optim/lr": 0.00293216957422215, "optim/total_tokens": 2696413184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291084, "loss/out": 4.3087921142578125, "created_at": "2025-01-15T03:29:54.097253+00:00"} {"global_step": 10287, "acc_step": 0, "speed/wps": 12963.896536198345, "speed/FLOPS": 203615809361636.84, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.111413598060608, "optim/lr": 0.0029321440980835042, "optim/total_tokens": 2696675328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437948, "loss/out": 4.194425106048584, "created_at": "2025-01-15T03:29:59.163594+00:00"} {"global_step": 10288, "acc_step": 0, "speed/wps": 12963.284748443579, "speed/FLOPS": 203606200394263.78, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1151689291000366, "optim/lr": 0.0029321186172722355, "optim/total_tokens": 2696937472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451275, "loss/out": 4.257571220397949, "created_at": "2025-01-15T03:30:04.223393+00:00"} {"global_step": 10289, "acc_step": 0, "speed/wps": 12959.674277513783, "speed/FLOPS": 203549492987004.34, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.70177161693573, "optim/lr": 0.0029320931317884285, "optim/total_tokens": 2697199616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 4.3019795417785645, "created_at": "2025-01-15T03:30:09.283970+00:00"} {"global_step": 10290, "acc_step": 0, "speed/wps": 12957.166122925883, "speed/FLOPS": 203510098972635.8, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.0965337753295898, "optim/lr": 0.0029320676416321662, "optim/total_tokens": 2697461760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 4.302884101867676, "created_at": "2025-01-15T03:30:14.346998+00:00"} {"global_step": 10291, "acc_step": 0, "speed/wps": 12958.865584606303, "speed/FLOPS": 203536791353631.88, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1062490940093994, "optim/lr": 0.0029320421468035313, "optim/total_tokens": 2697723904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 4.193925857543945, "created_at": "2025-01-15T03:30:19.410442+00:00"} {"global_step": 10292, "acc_step": 0, "speed/wps": 12962.619301294404, "speed/FLOPS": 203595748632365.78, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.6934505701065063, "optim/lr": 0.0029320166473026073, "optim/total_tokens": 2697986048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459222, "loss/out": 4.2763214111328125, "created_at": "2025-01-15T03:30:24.473474+00:00"} {"global_step": 10293, "acc_step": 0, "speed/wps": 12957.284547272799, "speed/FLOPS": 203511958989736.22, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9258005619049072, "optim/lr": 0.0029319911431294776, "optim/total_tokens": 2698248192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466155, "loss/out": 4.296084403991699, "created_at": "2025-01-15T03:30:29.534899+00:00"} {"global_step": 10294, "acc_step": 0, "speed/wps": 12956.8321648531, "speed/FLOPS": 203504853702196.28, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.19798743724823, "optim/lr": 0.002931965634284225, "optim/total_tokens": 2698510336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452081, "loss/out": 4.3936381340026855, "created_at": "2025-01-15T03:30:34.597104+00:00"} {"global_step": 10295, "acc_step": 0, "speed/wps": 12955.107677905846, "speed/FLOPS": 203477768264996.25, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.100454568862915, "optim/lr": 0.002931940120766933, "optim/total_tokens": 2698772480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312537, "loss/out": 4.3457536697387695, "created_at": "2025-01-15T03:30:39.664731+00:00"} {"global_step": 10296, "acc_step": 0, "speed/wps": 12962.111684896849, "speed/FLOPS": 203587775819304.12, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5017468929290771, "optim/lr": 0.0029319146025776847, "optim/total_tokens": 2699034624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 408877, "loss/out": 4.3099236488342285, "created_at": "2025-01-15T03:30:44.727567+00:00"} {"global_step": 10297, "acc_step": 0, "speed/wps": 12961.053473906442, "speed/FLOPS": 203571155161565.53, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.073315143585205, "optim/lr": 0.0029318890797165632, "optim/total_tokens": 2699296768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447555, "loss/out": 4.3890767097473145, "created_at": "2025-01-15T03:30:49.788838+00:00"} {"global_step": 10298, "acc_step": 0, "speed/wps": 12961.459302727053, "speed/FLOPS": 203577529260860.47, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1193956136703491, "optim/lr": 0.0029318635521836524, "optim/total_tokens": 2699558912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 4.418408393859863, "created_at": "2025-01-15T03:30:54.849804+00:00"} {"global_step": 10299, "acc_step": 0, "speed/wps": 12965.953792738801, "speed/FLOPS": 203648121402571.44, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6232742071151733, "optim/lr": 0.0029318380199790345, "optim/total_tokens": 2699821056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438240, "loss/out": 4.3138251304626465, "created_at": "2025-01-15T03:30:59.913162+00:00"} {"global_step": 10300, "acc_step": 0, "speed/wps": 12969.740185950759, "speed/FLOPS": 203707591911014.75, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1764492988586426, "optim/lr": 0.002931812483102794, "optim/total_tokens": 2700083200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 499747, "loss/out": 4.265681743621826, "created_at": "2025-01-15T03:31:04.970948+00:00"} {"global_step": 10301, "acc_step": 0, "speed/wps": 12964.852708271772, "speed/FLOPS": 203630827357968.12, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4518635272979736, "optim/lr": 0.0029317869415550132, "optim/total_tokens": 2700345344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449696, "loss/out": 4.284963607788086, "created_at": "2025-01-15T03:31:10.032675+00:00"} {"global_step": 10302, "acc_step": 0, "speed/wps": 12959.604410325901, "speed/FLOPS": 203548395626811.1, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1103365421295166, "optim/lr": 0.0029317613953357764, "optim/total_tokens": 2700607488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292143, "loss/out": 4.34734582901001, "created_at": "2025-01-15T03:31:15.093477+00:00"} {"global_step": 10303, "acc_step": 0, "speed/wps": 12961.088967873835, "speed/FLOPS": 203571712643097.25, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4767584800720215, "optim/lr": 0.002931735844445166, "optim/total_tokens": 2700869632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 344917, "loss/out": 4.238995552062988, "created_at": "2025-01-15T03:31:20.158958+00:00"} {"global_step": 10304, "acc_step": 0, "speed/wps": 12965.192863192118, "speed/FLOPS": 203636169958414.0, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0373955965042114, "optim/lr": 0.0029317102888832664, "optim/total_tokens": 2701131776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352401, "loss/out": 4.237286567687988, "created_at": "2025-01-15T03:31:25.220283+00:00"} {"global_step": 10305, "acc_step": 0, "speed/wps": 12959.592701028385, "speed/FLOPS": 203548211716203.34, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3603154420852661, "optim/lr": 0.0029316847286501597, "optim/total_tokens": 2701393920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452093, "loss/out": 4.138192176818848, "created_at": "2025-01-15T03:31:30.281383+00:00"} {"global_step": 10306, "acc_step": 0, "speed/wps": 12966.705091732649, "speed/FLOPS": 203659921585662.8, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2871625423431396, "optim/lr": 0.0029316591637459305, "optim/total_tokens": 2701656064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444881, "loss/out": 4.291752815246582, "created_at": "2025-01-15T03:31:35.339476+00:00"} {"global_step": 10307, "acc_step": 0, "speed/wps": 12968.376567913392, "speed/FLOPS": 203686174415930.38, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.161885380744934, "optim/lr": 0.0029316335941706614, "optim/total_tokens": 2701918208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 4.280634880065918, "created_at": "2025-01-15T03:31:40.400861+00:00"} {"global_step": 10308, "acc_step": 0, "speed/wps": 12960.815668384499, "speed/FLOPS": 203567420099064.38, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0435044765472412, "optim/lr": 0.0029316080199244363, "optim/total_tokens": 2702180352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424156, "loss/out": 4.319155216217041, "created_at": "2025-01-15T03:31:45.466346+00:00"} {"global_step": 10309, "acc_step": 0, "speed/wps": 12965.623672083768, "speed/FLOPS": 203642936404050.8, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.7104672193527222, "optim/lr": 0.0029315824410073386, "optim/total_tokens": 2702442496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 4.313943862915039, "created_at": "2025-01-15T03:31:50.527967+00:00"} {"global_step": 10310, "acc_step": 0, "speed/wps": 12966.18656892734, "speed/FLOPS": 203651777472480.44, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4710838794708252, "optim/lr": 0.0029315568574194515, "optim/total_tokens": 2702704640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438531, "loss/out": 4.324807167053223, "created_at": "2025-01-15T03:31:55.587172+00:00"} {"global_step": 10311, "acc_step": 0, "speed/wps": 12967.987608099893, "speed/FLOPS": 203680065267572.56, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0851496458053589, "optim/lr": 0.0029315312691608583, "optim/total_tokens": 2702966784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293783, "loss/out": 4.211212635040283, "created_at": "2025-01-15T03:32:00.644492+00:00"} {"global_step": 10312, "acc_step": 0, "speed/wps": 12964.976177189097, "speed/FLOPS": 203632766606978.8, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7476547956466675, "optim/lr": 0.0029315056762316428, "optim/total_tokens": 2703228928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360703, "loss/out": 4.284204006195068, "created_at": "2025-01-15T03:32:05.708090+00:00"} {"global_step": 10313, "acc_step": 0, "speed/wps": 12956.891912879422, "speed/FLOPS": 203505792127052.06, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1713827848434448, "optim/lr": 0.0029314800786318894, "optim/total_tokens": 2703491072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 4.284039497375488, "created_at": "2025-01-15T03:32:10.770080+00:00"} {"global_step": 10314, "acc_step": 0, "speed/wps": 12960.24961911726, "speed/FLOPS": 203558529517489.2, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.0720986127853394, "optim/lr": 0.00293145447636168, "optim/total_tokens": 2703753216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465083, "loss/out": 4.316899299621582, "created_at": "2025-01-15T03:32:15.836191+00:00"} {"global_step": 10315, "acc_step": 0, "speed/wps": 12959.993972529084, "speed/FLOPS": 203554514236526.66, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0998514890670776, "optim/lr": 0.0029314288694210987, "optim/total_tokens": 2704015360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 4.275601863861084, "created_at": "2025-01-15T03:32:20.898032+00:00"} {"global_step": 10316, "acc_step": 0, "speed/wps": 12960.668137667297, "speed/FLOPS": 203565102926422.25, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3660645484924316, "optim/lr": 0.0029314032578102297, "optim/total_tokens": 2704277504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 4.2925896644592285, "created_at": "2025-01-15T03:32:25.960735+00:00"} {"global_step": 10317, "acc_step": 0, "speed/wps": 12959.040648135173, "speed/FLOPS": 203539540966915.28, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9403488636016846, "optim/lr": 0.0029313776415291555, "optim/total_tokens": 2704539648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449165, "loss/out": 4.255865097045898, "created_at": "2025-01-15T03:32:31.021548+00:00"} {"global_step": 10318, "acc_step": 0, "speed/wps": 12968.65262662342, "speed/FLOPS": 203690510297315.75, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0269414186477661, "optim/lr": 0.0029313520205779603, "optim/total_tokens": 2704801792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448077, "loss/out": 4.249721527099609, "created_at": "2025-01-15T03:32:36.084301+00:00"} {"global_step": 10319, "acc_step": 0, "speed/wps": 12962.263718274684, "speed/FLOPS": 203590163712418.78, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9083487391471863, "optim/lr": 0.002931326394956728, "optim/total_tokens": 2705063936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315474, "loss/out": 4.119500160217285, "created_at": "2025-01-15T03:32:41.144859+00:00"} {"global_step": 10320, "acc_step": 0, "speed/wps": 12966.147346251691, "speed/FLOPS": 203651161426461.78, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2203177213668823, "optim/lr": 0.0029313007646655414, "optim/total_tokens": 2705326080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 369627, "loss/out": 4.3455681800842285, "created_at": "2025-01-15T03:32:46.207748+00:00"} {"global_step": 10321, "acc_step": 0, "speed/wps": 12964.912812156814, "speed/FLOPS": 203631771372074.1, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0904145240783691, "optim/lr": 0.0029312751297044847, "optim/total_tokens": 2705588224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441157, "loss/out": 4.341713905334473, "created_at": "2025-01-15T03:32:51.268268+00:00"} {"global_step": 10322, "acc_step": 0, "speed/wps": 12961.127360208893, "speed/FLOPS": 203572315647475.94, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9203609824180603, "optim/lr": 0.0029312494900736413, "optim/total_tokens": 2705850368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 4.212281703948975, "created_at": "2025-01-15T03:32:56.328430+00:00"} {"global_step": 10323, "acc_step": 0, "speed/wps": 12961.555282313817, "speed/FLOPS": 203579036752160.34, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9757306575775146, "optim/lr": 0.0029312238457730954, "optim/total_tokens": 2706112512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444881, "loss/out": 4.325852394104004, "created_at": "2025-01-15T03:33:01.388740+00:00"} {"global_step": 10324, "acc_step": 0, "speed/wps": 12962.843994033117, "speed/FLOPS": 203599277740587.44, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1925017833709717, "optim/lr": 0.0029311981968029297, "optim/total_tokens": 2706374656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494467, "loss/out": 4.207037448883057, "created_at": "2025-01-15T03:33:06.451249+00:00"} {"global_step": 10325, "acc_step": 0, "speed/wps": 12959.165268871779, "speed/FLOPS": 203541498306830.5, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2011229991912842, "optim/lr": 0.0029311725431632285, "optim/total_tokens": 2706636800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429217, "loss/out": 4.331298828125, "created_at": "2025-01-15T03:33:11.517520+00:00"} {"global_step": 10326, "acc_step": 0, "speed/wps": 12958.326226513043, "speed/FLOPS": 203528319993621.56, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1566120386123657, "optim/lr": 0.002931146884854076, "optim/total_tokens": 2706898944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361859, "loss/out": 4.319771766662598, "created_at": "2025-01-15T03:33:16.579896+00:00"} {"global_step": 10327, "acc_step": 0, "speed/wps": 12961.46911739941, "speed/FLOPS": 203577683413776.6, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0518832206726074, "optim/lr": 0.002931121221875554, "optim/total_tokens": 2707161088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429454, "loss/out": 4.224575042724609, "created_at": "2025-01-15T03:33:21.645001+00:00"} {"global_step": 10328, "acc_step": 0, "speed/wps": 12967.688619708499, "speed/FLOPS": 203675369244032.62, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0538628101348877, "optim/lr": 0.0029310955542277486, "optim/total_tokens": 2707423232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461605, "loss/out": 4.266040802001953, "created_at": "2025-01-15T03:33:26.706892+00:00"} {"global_step": 10329, "acc_step": 0, "speed/wps": 12963.549466236424, "speed/FLOPS": 203610358150960.75, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1718330383300781, "optim/lr": 0.0029310698819107416, "optim/total_tokens": 2707685376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441442, "loss/out": 4.369928359985352, "created_at": "2025-01-15T03:33:31.768932+00:00"} {"global_step": 10330, "acc_step": 0, "speed/wps": 12965.533826693394, "speed/FLOPS": 203641525258734.5, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6635447144508362, "optim/lr": 0.0029310442049246188, "optim/total_tokens": 2707947520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445695, "loss/out": 4.254685401916504, "created_at": "2025-01-15T03:33:36.827595+00:00"} {"global_step": 10331, "acc_step": 0, "speed/wps": 12965.272061654228, "speed/FLOPS": 203637413879091.53, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0880759954452515, "optim/lr": 0.002931018523269462, "optim/total_tokens": 2708209664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441971, "loss/out": 4.273837089538574, "created_at": "2025-01-15T03:33:41.892310+00:00"} {"global_step": 10332, "acc_step": 0, "speed/wps": 12967.119038039893, "speed/FLOPS": 203666423181239.94, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.670462965965271, "optim/lr": 0.002930992836945356, "optim/total_tokens": 2708471808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457095, "loss/out": 4.349498271942139, "created_at": "2025-01-15T03:33:46.951092+00:00"} {"global_step": 10333, "acc_step": 0, "speed/wps": 12969.19739815542, "speed/FLOPS": 203699066682820.38, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.8422167301177979, "optim/lr": 0.0029309671459523843, "optim/total_tokens": 2708733952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315181, "loss/out": 4.3465776443481445, "created_at": "2025-01-15T03:33:52.009235+00:00"} {"global_step": 10334, "acc_step": 0, "speed/wps": 12954.678540595805, "speed/FLOPS": 203471028073845.62, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9207860231399536, "optim/lr": 0.002930941450290631, "optim/total_tokens": 2708996096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328606, "loss/out": 4.3212480545043945, "created_at": "2025-01-15T03:33:57.075475+00:00"} {"global_step": 10335, "acc_step": 0, "speed/wps": 12961.410019497247, "speed/FLOPS": 203576755200014.1, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.256076455116272, "optim/lr": 0.00293091574996018, "optim/total_tokens": 2709258240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444878, "loss/out": 4.365680694580078, "created_at": "2025-01-15T03:34:02.137718+00:00"} {"global_step": 10336, "acc_step": 0, "speed/wps": 12956.879724180055, "speed/FLOPS": 203505600686779.47, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0973509550094604, "optim/lr": 0.0029308900449611145, "optim/total_tokens": 2709520384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 4.338373184204102, "created_at": "2025-01-15T03:34:07.204270+00:00"} {"global_step": 10337, "acc_step": 0, "speed/wps": 12960.389926723541, "speed/FLOPS": 203560733241248.7, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2146241664886475, "optim/lr": 0.002930864335293519, "optim/total_tokens": 2709782528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451797, "loss/out": 4.391249656677246, "created_at": "2025-01-15T03:34:12.270078+00:00"} {"global_step": 10338, "acc_step": 0, "speed/wps": 12964.797203084472, "speed/FLOPS": 203629955572729.47, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4096285104751587, "optim/lr": 0.002930838620957477, "optim/total_tokens": 2710044672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462963, "loss/out": 4.2645792961120605, "created_at": "2025-01-15T03:34:17.329802+00:00"} {"global_step": 10339, "acc_step": 0, "speed/wps": 12962.034373063558, "speed/FLOPS": 203586561530724.72, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2705429792404175, "optim/lr": 0.002930812901953073, "optim/total_tokens": 2710306816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319663, "loss/out": 4.219034671783447, "created_at": "2025-01-15T03:34:22.395702+00:00"} {"global_step": 10340, "acc_step": 0, "speed/wps": 12959.297848988996, "speed/FLOPS": 203543580659755.56, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5343432426452637, "optim/lr": 0.0029307871782803903, "optim/total_tokens": 2710568960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432875, "loss/out": 4.424581527709961, "created_at": "2025-01-15T03:34:27.460637+00:00"} {"global_step": 10341, "acc_step": 0, "speed/wps": 12965.11761909966, "speed/FLOPS": 203634988146545.72, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.2452869415283203, "optim/lr": 0.0029307614499395123, "optim/total_tokens": 2710831104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462963, "loss/out": 4.380382537841797, "created_at": "2025-01-15T03:34:32.519551+00:00"} {"global_step": 10342, "acc_step": 0, "speed/wps": 12958.460202158714, "speed/FLOPS": 203530424265238.66, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4628784656524658, "optim/lr": 0.0029307357169305243, "optim/total_tokens": 2711093248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 4.325778007507324, "created_at": "2025-01-15T03:34:37.581743+00:00"} {"global_step": 10343, "acc_step": 0, "speed/wps": 12960.696565989454, "speed/FLOPS": 203565549432286.34, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.55616295337677, "optim/lr": 0.00293070997925351, "optim/total_tokens": 2711355392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 4.3162736892700195, "created_at": "2025-01-15T03:34:42.646342+00:00"} {"global_step": 10344, "acc_step": 0, "speed/wps": 12958.79644922729, "speed/FLOPS": 203535705487502.5, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 6.1540350914001465, "optim/lr": 0.0029306842369085525, "optim/total_tokens": 2711617536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 4.218506813049316, "created_at": "2025-01-15T03:34:47.708152+00:00"} {"global_step": 10345, "acc_step": 0, "speed/wps": 12955.621703680887, "speed/FLOPS": 203485841746138.72, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0774166584014893, "optim/lr": 0.0029306584898957364, "optim/total_tokens": 2711879680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489719, "loss/out": 4.345833778381348, "created_at": "2025-01-15T03:34:52.776344+00:00"} {"global_step": 10346, "acc_step": 0, "speed/wps": 12965.392996806386, "speed/FLOPS": 203639313331838.5, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9905303716659546, "optim/lr": 0.0029306327382151456, "optim/total_tokens": 2712141824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439583, "loss/out": 4.289444923400879, "created_at": "2025-01-15T03:34:57.838330+00:00"} {"global_step": 10347, "acc_step": 0, "speed/wps": 12961.231960303341, "speed/FLOPS": 203573958535694.84, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.653939962387085, "optim/lr": 0.002930606981866864, "optim/total_tokens": 2712403968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307679, "loss/out": 4.286716461181641, "created_at": "2025-01-15T03:35:02.899455+00:00"} {"global_step": 10348, "acc_step": 0, "speed/wps": 12959.451129037825, "speed/FLOPS": 203545988133551.3, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1869490146636963, "optim/lr": 0.0029305812208509758, "optim/total_tokens": 2712666112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463770, "loss/out": 4.426665306091309, "created_at": "2025-01-15T03:35:07.965782+00:00"} {"global_step": 10349, "acc_step": 0, "speed/wps": 12962.287349501432, "speed/FLOPS": 203590534873307.6, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1144243478775024, "optim/lr": 0.002930555455167565, "optim/total_tokens": 2712928256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 4.346908092498779, "created_at": "2025-01-15T03:35:13.030307+00:00"} {"global_step": 10350, "acc_step": 0, "speed/wps": 12966.294077115906, "speed/FLOPS": 203653466036313.62, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.041845440864563, "optim/lr": 0.0029305296848167158, "optim/total_tokens": 2713190400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441157, "loss/out": 4.426712512969971, "created_at": "2025-01-15T03:35:18.088731+00:00"} {"global_step": 10351, "acc_step": 0, "speed/wps": 12962.092805730032, "speed/FLOPS": 203587479296046.6, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.1609599590301514, "optim/lr": 0.002930503909798512, "optim/total_tokens": 2713452544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 4.250199794769287, "created_at": "2025-01-15T03:35:23.148691+00:00"} {"global_step": 10352, "acc_step": 0, "speed/wps": 12964.61861848233, "speed/FLOPS": 203627150656151.84, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0630528926849365, "optim/lr": 0.0029304781301130378, "optim/total_tokens": 2713714688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467987, "loss/out": 4.212344169616699, "created_at": "2025-01-15T03:35:28.215385+00:00"} {"global_step": 10353, "acc_step": 0, "speed/wps": 12962.929668593397, "speed/FLOPS": 203600623377285.8, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7962992787361145, "optim/lr": 0.002930452345760377, "optim/total_tokens": 2713976832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312903, "loss/out": 4.309988498687744, "created_at": "2025-01-15T03:35:33.275018+00:00"} {"global_step": 10354, "acc_step": 0, "speed/wps": 12970.513842197846, "speed/FLOPS": 203719743245490.34, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6372621059417725, "optim/lr": 0.0029304265567406144, "optim/total_tokens": 2714238976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372699, "loss/out": 4.314347743988037, "created_at": "2025-01-15T03:35:38.336635+00:00"} {"global_step": 10355, "acc_step": 0, "speed/wps": 12961.85763613732, "speed/FLOPS": 203583785634436.16, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.1606100797653198, "optim/lr": 0.0029304007630538338, "optim/total_tokens": 2714501120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448890, "loss/out": 4.314417839050293, "created_at": "2025-01-15T03:35:43.399863+00:00"} {"global_step": 10356, "acc_step": 0, "speed/wps": 12954.43345434777, "speed/FLOPS": 203467178657536.44, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4044902324676514, "optim/lr": 0.002930374964700119, "optim/total_tokens": 2714763264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433196, "loss/out": 4.201385498046875, "created_at": "2025-01-15T03:35:48.470442+00:00"} {"global_step": 10357, "acc_step": 0, "speed/wps": 12963.11126095051, "speed/FLOPS": 203603475534788.34, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.424196481704712, "optim/lr": 0.0029303491616795552, "optim/total_tokens": 2715025408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436381, "loss/out": 4.291827201843262, "created_at": "2025-01-15T03:35:53.534765+00:00"} {"global_step": 10358, "acc_step": 0, "speed/wps": 12971.160910040662, "speed/FLOPS": 203729906335126.16, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2364979982376099, "optim/lr": 0.002930323353992226, "optim/total_tokens": 2715287552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 4.307250499725342, "created_at": "2025-01-15T03:35:58.591377+00:00"} {"global_step": 10359, "acc_step": 0, "speed/wps": 12962.793876444764, "speed/FLOPS": 203598490574993.6, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1110901832580566, "optim/lr": 0.0029302975416382144, "optim/total_tokens": 2715549696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 4.191982269287109, "created_at": "2025-01-15T03:36:03.650744+00:00"} {"global_step": 10360, "acc_step": 0, "speed/wps": 12955.926700441403, "speed/FLOPS": 203490632139372.1, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.001, "optim/grad_norm": 1.0365160703659058, "optim/lr": 0.002930271724617607, "optim/total_tokens": 2715811840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443306, "loss/out": 4.451680660247803, "created_at": "2025-01-15T03:36:08.746053+00:00"} {"global_step": 10361, "acc_step": 0, "speed/wps": 12965.316209591221, "speed/FLOPS": 203638107283107.66, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1045337915420532, "optim/lr": 0.002930245902930486, "optim/total_tokens": 2716073984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450746, "loss/out": 4.2045674324035645, "created_at": "2025-01-15T03:36:13.808817+00:00"} {"global_step": 10362, "acc_step": 0, "speed/wps": 12959.576273201274, "speed/FLOPS": 203547953694604.38, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.450658917427063, "optim/lr": 0.0029302200765769365, "optim/total_tokens": 2716336128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 4.386466026306152, "created_at": "2025-01-15T03:36:18.870948+00:00"} {"global_step": 10363, "acc_step": 0, "speed/wps": 12957.558387409985, "speed/FLOPS": 203516260025386.53, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0516281127929688, "optim/lr": 0.002930194245557043, "optim/total_tokens": 2716598272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470660, "loss/out": 4.272424697875977, "created_at": "2025-01-15T03:36:23.938594+00:00"} {"global_step": 10364, "acc_step": 0, "speed/wps": 12959.357648397317, "speed/FLOPS": 203544519891636.28, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.175308346748352, "optim/lr": 0.0029301684098708893, "optim/total_tokens": 2716860416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451275, "loss/out": 4.212950229644775, "created_at": "2025-01-15T03:36:29.005797+00:00"} {"global_step": 10365, "acc_step": 0, "speed/wps": 12964.36415855417, "speed/FLOPS": 203623154013315.53, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.046589732170105, "optim/lr": 0.00293014256951856, "optim/total_tokens": 2717122560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309025, "loss/out": 4.3150835037231445, "created_at": "2025-01-15T03:36:34.068257+00:00"} {"global_step": 10366, "acc_step": 0, "speed/wps": 12955.435004489394, "speed/FLOPS": 203482909378784.75, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.318615198135376, "optim/lr": 0.0029301167245001387, "optim/total_tokens": 2717384704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440624, "loss/out": 4.241951942443848, "created_at": "2025-01-15T03:36:39.136221+00:00"} {"global_step": 10367, "acc_step": 0, "speed/wps": 12964.596003451592, "speed/FLOPS": 203626795456017.8, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1557146310806274, "optim/lr": 0.002930090874815711, "optim/total_tokens": 2717646848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 4.234642028808594, "created_at": "2025-01-15T03:36:44.198873+00:00"} {"global_step": 10368, "acc_step": 0, "speed/wps": 12957.641961622265, "speed/FLOPS": 203517572673233.38, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1972243785858154, "optim/lr": 0.0029300650204653602, "optim/total_tokens": 2717908992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434807, "loss/out": 4.383098602294922, "created_at": "2025-01-15T03:36:49.260257+00:00"} {"global_step": 10369, "acc_step": 0, "speed/wps": 12958.524172306823, "speed/FLOPS": 203531429004318.88, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6035391092300415, "optim/lr": 0.0029300391614491715, "optim/total_tokens": 2718171136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 4.31971549987793, "created_at": "2025-01-15T03:36:54.326692+00:00"} {"global_step": 10370, "acc_step": 0, "speed/wps": 12961.039811597791, "speed/FLOPS": 203570940576234.2, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9120213985443115, "optim/lr": 0.002930013297767228, "optim/total_tokens": 2718433280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 4.267412185668945, "created_at": "2025-01-15T03:36:59.393023+00:00"} {"global_step": 10371, "acc_step": 0, "speed/wps": 12964.584234972404, "speed/FLOPS": 203626610615880.6, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0884521007537842, "optim/lr": 0.002929987429419615, "optim/total_tokens": 2718695424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447026, "loss/out": 4.166573524475098, "created_at": "2025-01-15T03:37:04.451657+00:00"} {"global_step": 10372, "acc_step": 0, "speed/wps": 12961.826867696338, "speed/FLOPS": 203583302373790.28, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0734238624572754, "optim/lr": 0.002929961556406417, "optim/total_tokens": 2718957568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322837, "loss/out": 4.336346626281738, "created_at": "2025-01-15T03:37:09.512932+00:00"} {"global_step": 10373, "acc_step": 0, "speed/wps": 12960.785339354503, "speed/FLOPS": 203566943739971.84, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2779127359390259, "optim/lr": 0.0029299356787277175, "optim/total_tokens": 2719219712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 344789, "loss/out": 4.172115325927734, "created_at": "2025-01-15T03:37:14.574270+00:00"} {"global_step": 10374, "acc_step": 0, "speed/wps": 12961.07584262703, "speed/FLOPS": 203571506493060.2, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.001, "optim/grad_norm": 1.3733278512954712, "optim/lr": 0.0029299097963836024, "optim/total_tokens": 2719481856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465369, "loss/out": 4.21713399887085, "created_at": "2025-01-15T03:37:19.635105+00:00"} {"global_step": 10375, "acc_step": 0, "speed/wps": 12963.704382516746, "speed/FLOPS": 203612791324020.5, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1527687311172485, "optim/lr": 0.0029298839093741553, "optim/total_tokens": 2719744000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 4.456594944000244, "created_at": "2025-01-15T03:37:24.694332+00:00"} {"global_step": 10376, "acc_step": 0, "speed/wps": 12956.920014491534, "speed/FLOPS": 203506233501485.94, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0797284841537476, "optim/lr": 0.00292985801769946, "optim/total_tokens": 2720006144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 4.361408233642578, "created_at": "2025-01-15T03:37:29.756342+00:00"} {"global_step": 10377, "acc_step": 0, "speed/wps": 12960.759118763242, "speed/FLOPS": 203566531909555.22, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.120924472808838, "optim/lr": 0.0029298321213596023, "optim/total_tokens": 2720268288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 4.207454681396484, "created_at": "2025-01-15T03:37:34.822768+00:00"} {"global_step": 10378, "acc_step": 0, "speed/wps": 12958.620444907205, "speed/FLOPS": 203532941097800.28, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0529030561447144, "optim/lr": 0.002929806220354666, "optim/total_tokens": 2720530432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466369, "loss/out": 4.2739105224609375, "created_at": "2025-01-15T03:37:39.890591+00:00"} {"global_step": 10379, "acc_step": 0, "speed/wps": 12961.904919591534, "speed/FLOPS": 203584528286058.88, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.44266939163208, "optim/lr": 0.0029297803146847356, "optim/total_tokens": 2720792576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322156, "loss/out": 4.370246887207031, "created_at": "2025-01-15T03:37:44.951458+00:00"} {"global_step": 10380, "acc_step": 0, "speed/wps": 12960.909251225978, "speed/FLOPS": 203568889946186.1, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2888579368591309, "optim/lr": 0.0029297544043498957, "optim/total_tokens": 2721054720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 378926, "loss/out": 4.349587440490723, "created_at": "2025-01-15T03:37:50.012091+00:00"} {"global_step": 10381, "acc_step": 0, "speed/wps": 12960.348402230133, "speed/FLOPS": 203560081042019.06, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.5656208992004395, "optim/lr": 0.0029297284893502304, "optim/total_tokens": 2721316864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422528, "loss/out": 4.358445167541504, "created_at": "2025-01-15T03:37:55.075602+00:00"} {"global_step": 10382, "acc_step": 0, "speed/wps": 12962.189202993246, "speed/FLOPS": 203588993347529.1, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0725750923156738, "optim/lr": 0.0029297025696858247, "optim/total_tokens": 2721579008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 4.254847526550293, "created_at": "2025-01-15T03:38:00.135184+00:00"} {"global_step": 10383, "acc_step": 0, "speed/wps": 12958.259044057659, "speed/FLOPS": 203527264800841.38, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3077152967453003, "optim/lr": 0.0029296766453567637, "optim/total_tokens": 2721841152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463222, "loss/out": 4.407020568847656, "created_at": "2025-01-15T03:38:05.197425+00:00"} {"global_step": 10384, "acc_step": 0, "speed/wps": 12964.677432740817, "speed/FLOPS": 203628074414900.66, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3488467931747437, "optim/lr": 0.0029296507163631314, "optim/total_tokens": 2722103296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 4.202415943145752, "created_at": "2025-01-15T03:38:10.261391+00:00"} {"global_step": 10385, "acc_step": 0, "speed/wps": 12961.412098920626, "speed/FLOPS": 203576787860215.7, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3508673906326294, "optim/lr": 0.002929624782705012, "optim/total_tokens": 2722365440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437201, "loss/out": 4.23502779006958, "created_at": "2025-01-15T03:38:15.327011+00:00"} {"global_step": 10386, "acc_step": 0, "speed/wps": 12959.505588433385, "speed/FLOPS": 203546843493193.62, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.001, "optim/grad_norm": 6.501448631286621, "optim/lr": 0.002929598844382491, "optim/total_tokens": 2722627584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 4.300542831420898, "created_at": "2025-01-15T03:38:20.387617+00:00"} {"global_step": 10387, "acc_step": 0, "speed/wps": 12960.303108945218, "speed/FLOPS": 203559369648740.0, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.322370171546936, "optim/lr": 0.002929572901395652, "optim/total_tokens": 2722889728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449135, "loss/out": 4.263308525085449, "created_at": "2025-01-15T03:38:25.447871+00:00"} {"global_step": 10388, "acc_step": 0, "speed/wps": 12962.236672719575, "speed/FLOPS": 203589738924811.3, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5201022624969482, "optim/lr": 0.0029295469537445806, "optim/total_tokens": 2723151872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315917, "loss/out": 4.329315662384033, "created_at": "2025-01-15T03:38:30.512821+00:00"} {"global_step": 10389, "acc_step": 0, "speed/wps": 12969.18533927004, "speed/FLOPS": 203698877281454.8, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2888139486312866, "optim/lr": 0.0029295210014293606, "optim/total_tokens": 2723414016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348659, "loss/out": 4.337396144866943, "created_at": "2025-01-15T03:38:35.570266+00:00"} {"global_step": 10390, "acc_step": 0, "speed/wps": 12964.70284511791, "speed/FLOPS": 203628473551203.2, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0795202255249023, "optim/lr": 0.0029294950444500776, "optim/total_tokens": 2723676160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438240, "loss/out": 4.17243766784668, "created_at": "2025-01-15T03:38:40.632937+00:00"} {"global_step": 10391, "acc_step": 0, "speed/wps": 12961.401310277757, "speed/FLOPS": 203576618409754.28, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0301549434661865, "optim/lr": 0.002929469082806816, "optim/total_tokens": 2723938304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 4.2855000495910645, "created_at": "2025-01-15T03:38:45.693168+00:00"} {"global_step": 10392, "acc_step": 0, "speed/wps": 12961.25957224065, "speed/FLOPS": 203574392219113.03, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9598389267921448, "optim/lr": 0.00292944311649966, "optim/total_tokens": 2724200448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440108, "loss/out": 4.214143753051758, "created_at": "2025-01-15T03:38:50.753901+00:00"} {"global_step": 10393, "acc_step": 0, "speed/wps": 12962.966839640936, "speed/FLOPS": 203601207199665.78, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1125295162200928, "optim/lr": 0.002929417145528695, "optim/total_tokens": 2724462592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 4.301091194152832, "created_at": "2025-01-15T03:38:55.816759+00:00"} {"global_step": 10394, "acc_step": 0, "speed/wps": 12964.814831310385, "speed/FLOPS": 203630232448241.16, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3612420558929443, "optim/lr": 0.0029293911698940046, "optim/total_tokens": 2724724736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460815, "loss/out": 4.171992301940918, "created_at": "2025-01-15T03:39:00.877030+00:00"} {"global_step": 10395, "acc_step": 0, "speed/wps": 12962.513470577693, "speed/FLOPS": 203594086415533.2, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9271960854530334, "optim/lr": 0.0029293651895956753, "optim/total_tokens": 2724986880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324561, "loss/out": 4.305853366851807, "created_at": "2025-01-15T03:39:05.937824+00:00"} {"global_step": 10396, "acc_step": 0, "speed/wps": 12948.20225291453, "speed/FLOPS": 203369309076459.34, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0012, "optim/grad_norm": 3.1046504974365234, "optim/lr": 0.0029293392046337904, "optim/total_tokens": 2725249024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319751, "loss/out": 4.246466636657715, "created_at": "2025-01-15T03:39:11.007464+00:00"} {"global_step": 10397, "acc_step": 0, "speed/wps": 12967.205282868774, "speed/FLOPS": 203667777774790.3, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8569473028182983, "optim/lr": 0.0029293132150084353, "optim/total_tokens": 2725511168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331967, "loss/out": 4.313855171203613, "created_at": "2025-01-15T03:39:16.065720+00:00"} {"global_step": 10398, "acc_step": 0, "speed/wps": 12963.379522795663, "speed/FLOPS": 203607688955696.88, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1627777814865112, "optim/lr": 0.0029292872207196947, "optim/total_tokens": 2725773312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352687, "loss/out": 4.201820373535156, "created_at": "2025-01-15T03:39:21.127907+00:00"} {"global_step": 10399, "acc_step": 0, "speed/wps": 12963.762321540744, "speed/FLOPS": 203613701336008.97, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0898933410644531, "optim/lr": 0.002929261221767653, "optim/total_tokens": 2726035456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434515, "loss/out": 4.292214393615723, "created_at": "2025-01-15T03:39:26.188088+00:00"} {"global_step": 10400, "acc_step": 0, "speed/wps": 12962.120290014022, "speed/FLOPS": 203587910974494.0, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1660442352294922, "optim/lr": 0.0029292352181523964, "optim/total_tokens": 2726297600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 497658, "loss/out": 4.252933025360107, "created_at": "2025-01-15T03:39:31.247747+00:00"} {"global_step": 10401, "acc_step": 0, "speed/wps": 12965.141007523505, "speed/FLOPS": 203635355493881.28, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.238676905632019, "optim/lr": 0.002929209209874008, "optim/total_tokens": 2726559744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446220, "loss/out": 4.208907127380371, "created_at": "2025-01-15T03:39:36.307830+00:00"} {"global_step": 10402, "acc_step": 0, "speed/wps": 12963.605242548843, "speed/FLOPS": 203611234194592.0, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1252318620681763, "optim/lr": 0.0029291831969325732, "optim/total_tokens": 2726821888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453375, "loss/out": 4.3009419441223145, "created_at": "2025-01-15T03:39:41.369044+00:00"} {"global_step": 10403, "acc_step": 0, "speed/wps": 12959.778188239672, "speed/FLOPS": 203551125047742.75, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0431361198425293, "optim/lr": 0.0029291571793281777, "optim/total_tokens": 2727084032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 4.224686622619629, "created_at": "2025-01-15T03:39:46.431841+00:00"} {"global_step": 10404, "acc_step": 0, "speed/wps": 12952.051430252941, "speed/FLOPS": 203429765695883.0, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7192413806915283, "optim/lr": 0.0029291311570609053, "optim/total_tokens": 2727346176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310371, "loss/out": 4.332112789154053, "created_at": "2025-01-15T03:39:51.495429+00:00"} {"global_step": 10405, "acc_step": 0, "speed/wps": 12964.329731012951, "speed/FLOPS": 203622613281472.28, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1668758392333984, "optim/lr": 0.002929105130130842, "optim/total_tokens": 2727608320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 367376, "loss/out": 4.342695236206055, "created_at": "2025-01-15T03:39:56.554436+00:00"} {"global_step": 10406, "acc_step": 0, "speed/wps": 12970.03592692812, "speed/FLOPS": 203712236929454.9, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.804930686950684, "optim/lr": 0.002929079098538071, "optim/total_tokens": 2727870464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449696, "loss/out": 4.376660346984863, "created_at": "2025-01-15T03:40:01.612075+00:00"} {"global_step": 10407, "acc_step": 0, "speed/wps": 12964.942520335675, "speed/FLOPS": 203632237979846.44, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2037265300750732, "optim/lr": 0.0029290530622826795, "optim/total_tokens": 2728132608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 4.423089981079102, "created_at": "2025-01-15T03:40:06.671753+00:00"} {"global_step": 10408, "acc_step": 0, "speed/wps": 12966.985891639011, "speed/FLOPS": 203664331934051.75, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0861198902130127, "optim/lr": 0.0029290270213647505, "optim/total_tokens": 2728394752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456045, "loss/out": 4.333394527435303, "created_at": "2025-01-15T03:40:11.734637+00:00"} {"global_step": 10409, "acc_step": 0, "speed/wps": 12954.023958950293, "speed/FLOPS": 203460746969613.4, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.108905553817749, "optim/lr": 0.00292900097578437, "optim/total_tokens": 2728656896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469875, "loss/out": 4.304701805114746, "created_at": "2025-01-15T03:40:16.797392+00:00"} {"global_step": 10410, "acc_step": 0, "speed/wps": 12969.744123051096, "speed/FLOPS": 203707653748585.7, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.87202262878418, "optim/lr": 0.0029289749255416225, "optim/total_tokens": 2728919040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 4.271413326263428, "created_at": "2025-01-15T03:40:21.854851+00:00"} {"global_step": 10411, "acc_step": 0, "speed/wps": 12964.061845417002, "speed/FLOPS": 203618405770074.53, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.920325756072998, "optim/lr": 0.002928948870636594, "optim/total_tokens": 2729181184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481523, "loss/out": 4.367523193359375, "created_at": "2025-01-15T03:40:26.916961+00:00"} {"global_step": 10412, "acc_step": 0, "speed/wps": 12967.79819126434, "speed/FLOPS": 203677090215884.22, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7755330801010132, "optim/lr": 0.002928922811069368, "optim/total_tokens": 2729443328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438014, "loss/out": 4.334882736206055, "created_at": "2025-01-15T03:40:31.978590+00:00"} {"global_step": 10413, "acc_step": 0, "speed/wps": 12961.573417415548, "speed/FLOPS": 203579321588853.28, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.2766213417053223, "optim/lr": 0.0029288967468400306, "optim/total_tokens": 2729705472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334369, "loss/out": 4.404706954956055, "created_at": "2025-01-15T03:40:37.051262+00:00"} {"global_step": 10414, "acc_step": 0, "speed/wps": 12967.870199646959, "speed/FLOPS": 203678221206483.12, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1597753763198853, "optim/lr": 0.002928870677948666, "optim/total_tokens": 2729967616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441679, "loss/out": 4.303844928741455, "created_at": "2025-01-15T03:40:42.110917+00:00"} {"global_step": 10415, "acc_step": 0, "speed/wps": 12960.788526223472, "speed/FLOPS": 203566993794128.22, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1460391283035278, "optim/lr": 0.0029288446043953605, "optim/total_tokens": 2730229760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 4.319622993469238, "created_at": "2025-01-15T03:40:47.174702+00:00"} {"global_step": 10416, "acc_step": 0, "speed/wps": 12966.715146118191, "speed/FLOPS": 203660079503603.94, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4467257261276245, "optim/lr": 0.0029288185261801976, "optim/total_tokens": 2730491904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446505, "loss/out": 4.39654541015625, "created_at": "2025-01-15T03:40:52.234774+00:00"} {"global_step": 10417, "acc_step": 0, "speed/wps": 12966.807836914035, "speed/FLOPS": 203661535339923.16, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.131718635559082, "optim/lr": 0.0029287924433032634, "optim/total_tokens": 2730754048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460815, "loss/out": 4.288592338562012, "created_at": "2025-01-15T03:40:57.293487+00:00"} {"global_step": 10418, "acc_step": 0, "speed/wps": 12959.714694998425, "speed/FLOPS": 203550127799140.12, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1631280183792114, "optim/lr": 0.0029287663557646436, "optim/total_tokens": 2731016192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452847, "loss/out": 4.318717956542969, "created_at": "2025-01-15T03:41:02.355080+00:00"} {"global_step": 10419, "acc_step": 0, "speed/wps": 12967.845570644482, "speed/FLOPS": 203677834374155.4, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.2033841609954834, "optim/lr": 0.0029287402635644216, "optim/total_tokens": 2731278336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323123, "loss/out": 4.320255279541016, "created_at": "2025-01-15T03:41:07.418195+00:00"} {"global_step": 10420, "acc_step": 0, "speed/wps": 12970.430436562488, "speed/FLOPS": 203718433245375.7, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.991520643234253, "optim/lr": 0.0029287141667026836, "optim/total_tokens": 2731540480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355325, "loss/out": 4.277487754821777, "created_at": "2025-01-15T03:41:12.477305+00:00"} {"global_step": 10421, "acc_step": 0, "speed/wps": 12961.027567348097, "speed/FLOPS": 203570748263467.38, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1499062776565552, "optim/lr": 0.0029286880651795146, "optim/total_tokens": 2731802624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453131, "loss/out": 4.355574131011963, "created_at": "2025-01-15T03:41:17.546794+00:00"} {"global_step": 10422, "acc_step": 0, "speed/wps": 12956.03199187098, "speed/FLOPS": 203492285885959.16, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2959784269332886, "optim/lr": 0.002928661958995, "optim/total_tokens": 2732064768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 4.282773494720459, "created_at": "2025-01-15T03:41:22.610152+00:00"} {"global_step": 10423, "acc_step": 0, "speed/wps": 12964.565737937684, "speed/FLOPS": 203626320094532.72, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2253128290176392, "optim/lr": 0.002928635848149224, "optim/total_tokens": 2732326912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439871, "loss/out": 4.3773274421691895, "created_at": "2025-01-15T03:41:27.672749+00:00"} {"global_step": 10424, "acc_step": 0, "speed/wps": 12966.25200588504, "speed/FLOPS": 203652805249820.94, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.40946364402771, "optim/lr": 0.0029286097326422732, "optim/total_tokens": 2732589056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440637, "loss/out": 4.391054153442383, "created_at": "2025-01-15T03:41:32.731597+00:00"} {"global_step": 10425, "acc_step": 0, "speed/wps": 12965.039139144641, "speed/FLOPS": 203633755511009.4, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3850367069244385, "optim/lr": 0.0029285836124742326, "optim/total_tokens": 2732851200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456330, "loss/out": 4.3069329261779785, "created_at": "2025-01-15T03:41:37.790028+00:00"} {"global_step": 10426, "acc_step": 0, "speed/wps": 12965.876174661078, "speed/FLOPS": 203646902304002.38, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5101573467254639, "optim/lr": 0.002928557487645186, "optim/total_tokens": 2733113344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 4.387265205383301, "created_at": "2025-01-15T03:41:42.851918+00:00"} {"global_step": 10427, "acc_step": 0, "speed/wps": 12965.749348783986, "speed/FLOPS": 203644910329325.62, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.154672622680664, "optim/lr": 0.00292853135815522, "optim/total_tokens": 2733375488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326870, "loss/out": 4.291727066040039, "created_at": "2025-01-15T03:41:47.915497+00:00"} {"global_step": 10428, "acc_step": 0, "speed/wps": 12960.249750400602, "speed/FLOPS": 203558531579474.47, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4565386772155762, "optim/lr": 0.002928505224004419, "optim/total_tokens": 2733637632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365501, "loss/out": 4.26696252822876, "created_at": "2025-01-15T03:41:52.976524+00:00"} {"global_step": 10429, "acc_step": 0, "speed/wps": 12966.371291639593, "speed/FLOPS": 203654678796511.97, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.040635585784912, "optim/lr": 0.0029284790851928696, "optim/total_tokens": 2733899776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429193, "loss/out": 4.313010215759277, "created_at": "2025-01-15T03:41:58.034547+00:00"} {"global_step": 10430, "acc_step": 0, "speed/wps": 12965.628225423683, "speed/FLOPS": 203643007920511.12, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 10.29488754272461, "optim/lr": 0.0029284529417206552, "optim/total_tokens": 2734161920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446747, "loss/out": 4.376986980438232, "created_at": "2025-01-15T03:42:03.092711+00:00"} {"global_step": 10431, "acc_step": 0, "speed/wps": 12965.047750537871, "speed/FLOPS": 203633890764773.38, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.531484603881836, "optim/lr": 0.0029284267935878626, "optim/total_tokens": 2734424064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467205, "loss/out": 4.320840835571289, "created_at": "2025-01-15T03:42:08.157050+00:00"} {"global_step": 10432, "acc_step": 0, "speed/wps": 12957.455748194594, "speed/FLOPS": 203514647935467.75, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2325290441513062, "optim/lr": 0.0029284006407945765, "optim/total_tokens": 2734686208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436916, "loss/out": 4.576159954071045, "created_at": "2025-01-15T03:42:13.223907+00:00"} {"global_step": 10433, "acc_step": 0, "speed/wps": 12958.065983684151, "speed/FLOPS": 203524232522382.7, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.652287483215332, "optim/lr": 0.002928374483340882, "optim/total_tokens": 2734948352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 4.467865943908691, "created_at": "2025-01-15T03:42:18.285156+00:00"} {"global_step": 10434, "acc_step": 0, "speed/wps": 12956.500359096543, "speed/FLOPS": 203499642236840.66, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1109758615493774, "optim/lr": 0.0029283483212268657, "optim/total_tokens": 2735210496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435329, "loss/out": 4.294935703277588, "created_at": "2025-01-15T03:42:23.351716+00:00"} {"global_step": 10435, "acc_step": 0, "speed/wps": 12962.393102505346, "speed/FLOPS": 203592195869553.9, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.4933465719223022, "optim/lr": 0.0029283221544526114, "optim/total_tokens": 2735472640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301350, "loss/out": 4.156833171844482, "created_at": "2025-01-15T03:42:28.415595+00:00"} {"global_step": 10436, "acc_step": 0, "speed/wps": 12970.258510108786, "speed/FLOPS": 203715732904168.6, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.9788968563079834, "optim/lr": 0.0029282959830182046, "optim/total_tokens": 2735734784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 382133, "loss/out": 4.235747814178467, "created_at": "2025-01-15T03:42:33.471910+00:00"} {"global_step": 10437, "acc_step": 0, "speed/wps": 12955.17241635187, "speed/FLOPS": 203478785071251.03, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1871585845947266, "optim/lr": 0.0029282698069237317, "optim/total_tokens": 2735996928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 407538, "loss/out": 4.307026386260986, "created_at": "2025-01-15T03:42:38.534344+00:00"} {"global_step": 10438, "acc_step": 0, "speed/wps": 12960.87674224326, "speed/FLOPS": 203568379347940.47, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4259893894195557, "optim/lr": 0.002928243626169278, "optim/total_tokens": 2736259072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 4.217400074005127, "created_at": "2025-01-15T03:42:43.594424+00:00"} {"global_step": 10439, "acc_step": 0, "speed/wps": 12965.91449502161, "speed/FLOPS": 203647504177922.25, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.713762044906616, "optim/lr": 0.002928217440754928, "optim/total_tokens": 2736521216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 4.303187370300293, "created_at": "2025-01-15T03:42:48.657183+00:00"} {"global_step": 10440, "acc_step": 0, "speed/wps": 12965.61161984324, "speed/FLOPS": 203642747107051.8, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.2491743564605713, "optim/lr": 0.002928191250680768, "optim/total_tokens": 2736783360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457920, "loss/out": 4.333614349365234, "created_at": "2025-01-15T03:42:53.717667+00:00"} {"global_step": 10441, "acc_step": 0, "speed/wps": 12965.58921157452, "speed/FLOPS": 203642395154399.38, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.6244052648544312, "optim/lr": 0.002928165055946882, "optim/total_tokens": 2737045504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 4.224908351898193, "created_at": "2025-01-15T03:42:58.775992+00:00"} {"global_step": 10442, "acc_step": 0, "speed/wps": 12962.710445894956, "speed/FLOPS": 203597180183563.25, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.8678903579711914, "optim/lr": 0.002928138856553358, "optim/total_tokens": 2737307648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 4.4006781578063965, "created_at": "2025-01-15T03:43:03.835471+00:00"} {"global_step": 10443, "acc_step": 0, "speed/wps": 12970.58989486173, "speed/FLOPS": 203720937757083.97, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0775947570800781, "optim/lr": 0.002928112652500279, "optim/total_tokens": 2737569792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 4.223951816558838, "created_at": "2025-01-15T03:43:08.895976+00:00"} {"global_step": 10444, "acc_step": 0, "speed/wps": 12968.679911675585, "speed/FLOPS": 203690938846553.6, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.045528531074524, "optim/lr": 0.002928086443787732, "optim/total_tokens": 2737831936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436096, "loss/out": 4.315746307373047, "created_at": "2025-01-15T03:43:13.953332+00:00"} {"global_step": 10445, "acc_step": 0, "speed/wps": 12965.846158796227, "speed/FLOPS": 203646430863598.78, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.146697998046875, "optim/lr": 0.002928060230415802, "optim/total_tokens": 2738094080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 285748, "loss/out": 4.289954662322998, "created_at": "2025-01-15T03:43:19.012330+00:00"} {"global_step": 10446, "acc_step": 0, "speed/wps": 12967.127604406207, "speed/FLOPS": 203666557727794.3, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.7451329231262207, "optim/lr": 0.0029280340123845746, "optim/total_tokens": 2738356224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431521, "loss/out": 4.234560012817383, "created_at": "2025-01-15T03:43:24.074710+00:00"} {"global_step": 10447, "acc_step": 0, "speed/wps": 12963.75343242849, "speed/FLOPS": 203613561720286.1, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1792720556259155, "optim/lr": 0.002928007789694135, "optim/total_tokens": 2738618368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418566, "loss/out": 4.264387130737305, "created_at": "2025-01-15T03:43:29.133827+00:00"} {"global_step": 10448, "acc_step": 0, "speed/wps": 12963.961362512639, "speed/FLOPS": 203616827547983.53, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.066929817199707, "optim/lr": 0.0029279815623445694, "optim/total_tokens": 2738880512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 4.34881067276001, "created_at": "2025-01-15T03:43:34.194064+00:00"} {"global_step": 10449, "acc_step": 0, "speed/wps": 12965.534139641488, "speed/FLOPS": 203641530174014.34, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.16844642162323, "optim/lr": 0.0029279553303359633, "optim/total_tokens": 2739142656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459528, "loss/out": 4.185360908508301, "created_at": "2025-01-15T03:43:39.253274+00:00"} {"global_step": 10450, "acc_step": 0, "speed/wps": 12965.364039052029, "speed/FLOPS": 203638858510514.2, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.4271668195724487, "optim/lr": 0.0029279290936684017, "optim/total_tokens": 2739404800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453666, "loss/out": 4.1665496826171875, "created_at": "2025-01-15T03:43:44.323360+00:00"} {"global_step": 10451, "acc_step": 0, "speed/wps": 12967.964395810066, "speed/FLOPS": 203679700686663.8, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3755195140838623, "optim/lr": 0.0029279028523419705, "optim/total_tokens": 2739666944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447026, "loss/out": 4.274202823638916, "created_at": "2025-01-15T03:43:49.380935+00:00"} {"global_step": 10452, "acc_step": 0, "speed/wps": 12963.218367102185, "speed/FLOPS": 203605157784079.66, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1582051515579224, "optim/lr": 0.0029278766063567554, "optim/total_tokens": 2739929088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 4.268634796142578, "created_at": "2025-01-15T03:43:54.445834+00:00"} {"global_step": 10453, "acc_step": 0, "speed/wps": 12963.243234055248, "speed/FLOPS": 203605548353748.28, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1806151866912842, "optim/lr": 0.002927850355712842, "optim/total_tokens": 2740191232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433993, "loss/out": 4.267189025878906, "created_at": "2025-01-15T03:43:59.505502+00:00"} {"global_step": 10454, "acc_step": 0, "speed/wps": 12965.371595002023, "speed/FLOPS": 203638977187091.8, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2624456882476807, "optim/lr": 0.002927824100410316, "optim/total_tokens": 2740453376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 299756, "loss/out": 4.34714412689209, "created_at": "2025-01-15T03:44:04.568906+00:00"} {"global_step": 10455, "acc_step": 0, "speed/wps": 12963.999208215468, "speed/FLOPS": 203617421966751.9, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1377774477005005, "optim/lr": 0.002927797840449263, "optim/total_tokens": 2740715520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455247, "loss/out": 4.231276512145996, "created_at": "2025-01-15T03:44:09.632476+00:00"} {"global_step": 10456, "acc_step": 0, "speed/wps": 12963.588646926217, "speed/FLOPS": 203610973537533.88, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.001, "optim/grad_norm": 1.0747548341751099, "optim/lr": 0.002927771575829768, "optim/total_tokens": 2740977664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 4.285303115844727, "created_at": "2025-01-15T03:44:14.697140+00:00"} {"global_step": 10457, "acc_step": 0, "speed/wps": 12971.810691930648, "speed/FLOPS": 203740112052602.28, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.155017375946045, "optim/lr": 0.002927745306551918, "optim/total_tokens": 2741239808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430283, "loss/out": 4.319972515106201, "created_at": "2025-01-15T03:44:19.753003+00:00"} {"global_step": 10458, "acc_step": 0, "speed/wps": 12964.40455513841, "speed/FLOPS": 203623788497181.75, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1883403062820435, "optim/lr": 0.002927719032615798, "optim/total_tokens": 2741501952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 4.24859619140625, "created_at": "2025-01-15T03:44:24.812113+00:00"} {"global_step": 10459, "acc_step": 0, "speed/wps": 12960.261971697966, "speed/FLOPS": 203558723531743.4, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3417898416519165, "optim/lr": 0.0029276927540214936, "optim/total_tokens": 2741764096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448361, "loss/out": 4.3993940353393555, "created_at": "2025-01-15T03:44:29.876532+00:00"} {"global_step": 10460, "acc_step": 0, "speed/wps": 12960.994050354777, "speed/FLOPS": 203570221833030.1, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.440969944000244, "optim/lr": 0.0029276664707690906, "optim/total_tokens": 2742026240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312011, "loss/out": 4.188444137573242, "created_at": "2025-01-15T03:44:34.944200+00:00"} {"global_step": 10461, "acc_step": 0, "speed/wps": 12961.778579660067, "speed/FLOPS": 203582543943825.94, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.179743528366089, "optim/lr": 0.0029276401828586753, "optim/total_tokens": 2742288384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 398310, "loss/out": 4.37673282623291, "created_at": "2025-01-15T03:44:40.008463+00:00"} {"global_step": 10462, "acc_step": 0, "speed/wps": 12965.02194983469, "speed/FLOPS": 203633485529276.06, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.120697259902954, "optim/lr": 0.002927613890290333, "optim/total_tokens": 2742550528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429739, "loss/out": 4.283148288726807, "created_at": "2025-01-15T03:44:45.067323+00:00"} {"global_step": 10463, "acc_step": 0, "speed/wps": 12957.637110866966, "speed/FLOPS": 203517496485455.78, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.242830514907837, "optim/lr": 0.002927587593064149, "optim/total_tokens": 2742812672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471738, "loss/out": 4.141866683959961, "created_at": "2025-01-15T03:44:50.128668+00:00"} {"global_step": 10464, "acc_step": 0, "speed/wps": 12964.307532582918, "speed/FLOPS": 203622264624624.78, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.065224528312683, "optim/lr": 0.00292756129118021, "optim/total_tokens": 2743074816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447555, "loss/out": 4.263330459594727, "created_at": "2025-01-15T03:44:55.195396+00:00"} {"global_step": 10465, "acc_step": 0, "speed/wps": 12966.9484633739, "speed/FLOPS": 203663744071716.34, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3964500427246094, "optim/lr": 0.0029275349846386023, "optim/total_tokens": 2743336960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461913, "loss/out": 4.27424430847168, "created_at": "2025-01-15T03:45:00.255122+00:00"} {"global_step": 10466, "acc_step": 0, "speed/wps": 12967.120328370496, "speed/FLOPS": 203666443447655.16, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.374206781387329, "optim/lr": 0.00292750867343941, "optim/total_tokens": 2743599104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441971, "loss/out": 4.183032989501953, "created_at": "2025-01-15T03:45:05.313674+00:00"} {"global_step": 10467, "acc_step": 0, "speed/wps": 12964.741856021667, "speed/FLOPS": 203629086271053.28, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5157705545425415, "optim/lr": 0.0029274823575827194, "optim/total_tokens": 2743861248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 4.276822090148926, "created_at": "2025-01-15T03:45:10.372340+00:00"} {"global_step": 10468, "acc_step": 0, "speed/wps": 12965.872782221122, "speed/FLOPS": 203646849021071.2, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.19084894657135, "optim/lr": 0.0029274560370686174, "optim/total_tokens": 2744123392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439871, "loss/out": 4.169302940368652, "created_at": "2025-01-15T03:45:15.432090+00:00"} {"global_step": 10469, "acc_step": 0, "speed/wps": 12964.747665159919, "speed/FLOPS": 203629177511552.12, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.2580666542053223, "optim/lr": 0.002927429711897189, "optim/total_tokens": 2744385536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290898, "loss/out": 4.356834411621094, "created_at": "2025-01-15T03:45:20.494711+00:00"} {"global_step": 10470, "acc_step": 0, "speed/wps": 12963.286269655793, "speed/FLOPS": 203606224286991.97, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1829794645309448, "optim/lr": 0.0029274033820685208, "optim/total_tokens": 2744647680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 408396, "loss/out": 4.321095943450928, "created_at": "2025-01-15T03:45:25.554051+00:00"} {"global_step": 10471, "acc_step": 0, "speed/wps": 12964.276262876934, "speed/FLOPS": 203621773490923.22, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1199469566345215, "optim/lr": 0.0029273770475826974, "optim/total_tokens": 2744909824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444359, "loss/out": 4.148148059844971, "created_at": "2025-01-15T03:45:30.616518+00:00"} {"global_step": 10472, "acc_step": 0, "speed/wps": 12961.435882223777, "speed/FLOPS": 203577161409673.38, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0009, "optim/grad_norm": 3.153865098953247, "optim/lr": 0.0029273507084398064, "optim/total_tokens": 2745171968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 4.376486301422119, "created_at": "2025-01-15T03:45:35.681472+00:00"} {"global_step": 10473, "acc_step": 0, "speed/wps": 12970.735263101109, "speed/FLOPS": 203723220965040.28, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0358872413635254, "optim/lr": 0.0029273243646399326, "optim/total_tokens": 2745434112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434230, "loss/out": 4.362823486328125, "created_at": "2025-01-15T03:45:40.737937+00:00"} {"global_step": 10474, "acc_step": 0, "speed/wps": 12960.574049857592, "speed/FLOPS": 203563625148080.66, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0543073415756226, "optim/lr": 0.002927298016183162, "optim/total_tokens": 2745696256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 4.210151672363281, "created_at": "2025-01-15T03:45:45.798030+00:00"} {"global_step": 10475, "acc_step": 0, "speed/wps": 12965.2326199634, "speed/FLOPS": 203636794393138.53, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.1791110038757324, "optim/lr": 0.0029272716630695814, "optim/total_tokens": 2745958400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451275, "loss/out": 4.331319808959961, "created_at": "2025-01-15T03:45:50.856463+00:00"} {"global_step": 10476, "acc_step": 0, "speed/wps": 12970.759051941222, "speed/FLOPS": 203723594601463.56, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.893242120742798, "optim/lr": 0.002927245305299276, "optim/total_tokens": 2746220544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452081, "loss/out": 4.239492416381836, "created_at": "2025-01-15T03:45:55.916897+00:00"} {"global_step": 10477, "acc_step": 0, "speed/wps": 12970.371902402809, "speed/FLOPS": 203717513885963.56, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.9950931072235107, "optim/lr": 0.002927218942872332, "optim/total_tokens": 2746482688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 4.249619007110596, "created_at": "2025-01-15T03:46:00.976830+00:00"} {"global_step": 10478, "acc_step": 0, "speed/wps": 12967.539496300154, "speed/FLOPS": 203673027055987.53, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1420631408691406, "optim/lr": 0.0029271925757888353, "optim/total_tokens": 2746744832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449135, "loss/out": 4.288087844848633, "created_at": "2025-01-15T03:46:06.035624+00:00"} {"global_step": 10479, "acc_step": 0, "speed/wps": 12964.767170717314, "speed/FLOPS": 203629483873134.47, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1857742071151733, "optim/lr": 0.0029271662040488726, "optim/total_tokens": 2747006976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315917, "loss/out": 4.410912990570068, "created_at": "2025-01-15T03:46:11.098381+00:00"} {"global_step": 10480, "acc_step": 0, "speed/wps": 12961.521614569936, "speed/FLOPS": 203578507953978.38, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.4959917068481445, "optim/lr": 0.0029271398276525286, "optim/total_tokens": 2747269120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 383188, "loss/out": 4.304197311401367, "created_at": "2025-01-15T03:46:16.159498+00:00"} {"global_step": 10481, "acc_step": 0, "speed/wps": 12961.927883372957, "speed/FLOPS": 203584888963801.66, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.342401146888733, "optim/lr": 0.0029271134465998906, "optim/total_tokens": 2747531264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443307, "loss/out": 4.3960795402526855, "created_at": "2025-01-15T03:46:21.219070+00:00"} {"global_step": 10482, "acc_step": 0, "speed/wps": 12963.972358426747, "speed/FLOPS": 203617000253924.56, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2852259874343872, "optim/lr": 0.0029270870608910436, "optim/total_tokens": 2747793408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 4.31278133392334, "created_at": "2025-01-15T03:46:26.278007+00:00"} {"global_step": 10483, "acc_step": 0, "speed/wps": 12967.878167170797, "speed/FLOPS": 203678346347393.5, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.212798833847046, "optim/lr": 0.002927060670526075, "optim/total_tokens": 2748055552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465057, "loss/out": 4.409362316131592, "created_at": "2025-01-15T03:46:31.336996+00:00"} {"global_step": 10484, "acc_step": 0, "speed/wps": 12967.131226895339, "speed/FLOPS": 203666614623963.94, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0787924528121948, "optim/lr": 0.00292703427550507, "optim/total_tokens": 2748317696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443016, "loss/out": 4.272956371307373, "created_at": "2025-01-15T03:46:36.394576+00:00"} {"global_step": 10485, "acc_step": 0, "speed/wps": 12965.57024583063, "speed/FLOPS": 203642097271329.94, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2889070510864258, "optim/lr": 0.0029270078758281146, "optim/total_tokens": 2748579840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 500866, "loss/out": 4.343718528747559, "created_at": "2025-01-15T03:46:41.457382+00:00"} {"global_step": 10486, "acc_step": 0, "speed/wps": 12965.2314876696, "speed/FLOPS": 203636776608908.47, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1013129949569702, "optim/lr": 0.002926981471495296, "optim/total_tokens": 2748841984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424678, "loss/out": 4.32852840423584, "created_at": "2025-01-15T03:46:46.520200+00:00"} {"global_step": 10487, "acc_step": 0, "speed/wps": 12961.660097026896, "speed/FLOPS": 203580683011259.75, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.022627830505371, "optim/lr": 0.0029269550625066992, "optim/total_tokens": 2749104128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326583, "loss/out": 4.42836856842041, "created_at": "2025-01-15T03:46:51.580315+00:00"} {"global_step": 10488, "acc_step": 0, "speed/wps": 12967.508901608599, "speed/FLOPS": 203672546524315.12, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.9478741884231567, "optim/lr": 0.0029269286488624105, "optim/total_tokens": 2749366272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425702, "loss/out": 4.345974922180176, "created_at": "2025-01-15T03:46:56.638963+00:00"} {"global_step": 10489, "acc_step": 0, "speed/wps": 12966.039221718609, "speed/FLOPS": 203649463182091.03, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1303349733352661, "optim/lr": 0.002926902230562517, "optim/total_tokens": 2749628416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442208, "loss/out": 4.233494758605957, "created_at": "2025-01-15T03:47:01.702166+00:00"} {"global_step": 10490, "acc_step": 0, "speed/wps": 12964.57768791418, "speed/FLOPS": 203626507785334.56, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.050634503364563, "optim/lr": 0.002926875807607104, "optim/total_tokens": 2749890560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447263, "loss/out": 4.295646667480469, "created_at": "2025-01-15T03:47:06.766493+00:00"} {"global_step": 10491, "acc_step": 0, "speed/wps": 12955.063191617859, "speed/FLOPS": 203477069546712.84, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0011, "optim/grad_norm": 1.2026396989822388, "optim/lr": 0.0029268493799962576, "optim/total_tokens": 2750152704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471181, "loss/out": 4.221988201141357, "created_at": "2025-01-15T03:47:11.833704+00:00"} {"global_step": 10492, "acc_step": 0, "speed/wps": 12960.145371434908, "speed/FLOPS": 203556892164386.53, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3390381336212158, "optim/lr": 0.0029268229477300645, "optim/total_tokens": 2750414848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427066, "loss/out": 4.234983921051025, "created_at": "2025-01-15T03:47:16.898255+00:00"} {"global_step": 10493, "acc_step": 0, "speed/wps": 12964.41575482725, "speed/FLOPS": 203623964403684.97, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 5.891846179962158, "optim/lr": 0.0029267965108086113, "optim/total_tokens": 2750676992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 4.332371234893799, "created_at": "2025-01-15T03:47:21.960691+00:00"} {"global_step": 10494, "acc_step": 0, "speed/wps": 12963.115014925406, "speed/FLOPS": 203603534496122.56, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2838220596313477, "optim/lr": 0.0029267700692319834, "optim/total_tokens": 2750939136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439583, "loss/out": 4.280910968780518, "created_at": "2025-01-15T03:47:27.021082+00:00"} {"global_step": 10495, "acc_step": 0, "speed/wps": 12964.142575359154, "speed/FLOPS": 203619673744750.6, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0743125677108765, "optim/lr": 0.0029267436230002676, "optim/total_tokens": 2751201280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319909, "loss/out": 4.312020301818848, "created_at": "2025-01-15T03:47:32.080020+00:00"} {"global_step": 10496, "acc_step": 0, "speed/wps": 12965.18541016834, "speed/FLOPS": 203636052898434.06, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.343369245529175, "optim/lr": 0.0029267171721135497, "optim/total_tokens": 2751463424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 382664, "loss/out": 4.2627272605896, "created_at": "2025-01-15T03:47:37.145533+00:00"} {"global_step": 10497, "acc_step": 0, "speed/wps": 12955.22470749552, "speed/FLOPS": 203479606375517.6, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1215168237686157, "optim/lr": 0.0029266907165719164, "optim/total_tokens": 2751725568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443016, "loss/out": 4.261404037475586, "created_at": "2025-01-15T03:47:42.237464+00:00"} {"global_step": 10498, "acc_step": 0, "speed/wps": 12961.565094744077, "speed/FLOPS": 203579190869861.28, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.437984824180603, "optim/lr": 0.002926664256375454, "optim/total_tokens": 2751987712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 4.22765588760376, "created_at": "2025-01-15T03:47:47.298317+00:00"} {"global_step": 10499, "acc_step": 0, "speed/wps": 12961.240158405597, "speed/FLOPS": 203574087298156.12, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1250330209732056, "optim/lr": 0.002926637791524249, "optim/total_tokens": 2752249856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428403, "loss/out": 4.353487014770508, "created_at": "2025-01-15T03:47:52.360583+00:00"} {"global_step": 10500, "acc_step": 0, "speed/wps": 12958.731273639904, "speed/FLOPS": 203534681815341.97, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.315447211265564, "optim/lr": 0.0029266113220183876, "optim/total_tokens": 2752512000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 343860, "loss/out": 4.291903018951416, "created_at": "2025-01-15T03:47:57.426083+00:00"} {"global_step": 10501, "acc_step": 0, "speed/wps": 12964.28813350069, "speed/FLOPS": 203621959935381.25, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9228731393814087, "optim/lr": 0.002926584847857956, "optim/total_tokens": 2752774144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446459, "loss/out": 4.258161544799805, "created_at": "2025-01-15T03:48:02.484674+00:00"} {"global_step": 10502, "acc_step": 0, "speed/wps": 12963.971231129743, "speed/FLOPS": 203616982548176.03, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.79257869720459, "optim/lr": 0.0029265583690430406, "optim/total_tokens": 2753036288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439586, "loss/out": 4.328593730926514, "created_at": "2025-01-15T03:48:07.568601+00:00"} {"global_step": 10503, "acc_step": 0, "speed/wps": 12960.090830280977, "speed/FLOPS": 203556035520616.62, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.054969310760498, "optim/lr": 0.0029265318855737276, "optim/total_tokens": 2753298432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456567, "loss/out": 4.3181257247924805, "created_at": "2025-01-15T03:48:12.637616+00:00"} {"global_step": 10504, "acc_step": 0, "speed/wps": 12962.299323930823, "speed/FLOPS": 203590722948175.78, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.258929491043091, "optim/lr": 0.002926505397450104, "optim/total_tokens": 2753560576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457106, "loss/out": 4.225924015045166, "created_at": "2025-01-15T03:48:17.703068+00:00"} {"global_step": 10505, "acc_step": 0, "speed/wps": 12967.007587897744, "speed/FLOPS": 203664672703608.94, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2985589504241943, "optim/lr": 0.0029264789046722557, "optim/total_tokens": 2753822720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 4.17566442489624, "created_at": "2025-01-15T03:48:22.766188+00:00"} {"global_step": 10506, "acc_step": 0, "speed/wps": 12966.222012171778, "speed/FLOPS": 203652334157338.47, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5646706819534302, "optim/lr": 0.00292645240724027, "optim/total_tokens": 2754084864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429193, "loss/out": 4.341752052307129, "created_at": "2025-01-15T03:48:27.824337+00:00"} {"global_step": 10507, "acc_step": 0, "speed/wps": 12968.030285663768, "speed/FLOPS": 203680735577358.34, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.149053692817688, "optim/lr": 0.002926425905154232, "optim/total_tokens": 2754347008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 285262, "loss/out": 4.348725318908691, "created_at": "2025-01-15T03:48:32.881761+00:00"} {"global_step": 10508, "acc_step": 0, "speed/wps": 12962.126791612627, "speed/FLOPS": 203588013091034.25, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3221378326416016, "optim/lr": 0.002926399398414229, "optim/total_tokens": 2754609152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415331, "loss/out": 4.384684085845947, "created_at": "2025-01-15T03:48:37.942582+00:00"} {"global_step": 10509, "acc_step": 0, "speed/wps": 12964.750517181179, "speed/FLOPS": 203629222306465.1, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.90810227394104, "optim/lr": 0.002926372887020347, "optim/total_tokens": 2754871296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453660, "loss/out": 4.313522815704346, "created_at": "2025-01-15T03:48:43.007379+00:00"} {"global_step": 10510, "acc_step": 0, "speed/wps": 12960.195031733856, "speed/FLOPS": 203557672147621.9, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.1355082988739014, "optim/lr": 0.0029263463709726734, "optim/total_tokens": 2755133440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 4.221347808837891, "created_at": "2025-01-15T03:48:48.067708+00:00"} {"global_step": 10511, "acc_step": 0, "speed/wps": 12963.823291817365, "speed/FLOPS": 203614658957985.22, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.087907075881958, "optim/lr": 0.0029263198502712936, "optim/total_tokens": 2755395584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458715, "loss/out": 4.317056179046631, "created_at": "2025-01-15T03:48:53.132140+00:00"} {"global_step": 10512, "acc_step": 0, "speed/wps": 12961.979410116035, "speed/FLOPS": 203585698262106.88, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.331265687942505, "optim/lr": 0.002926293324916295, "optim/total_tokens": 2755657728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449696, "loss/out": 4.155558109283447, "created_at": "2025-01-15T03:48:58.201843+00:00"} {"global_step": 10513, "acc_step": 0, "speed/wps": 12965.576901393333, "speed/FLOPS": 203642201806087.88, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3072843551635742, "optim/lr": 0.002926266794907764, "optim/total_tokens": 2755919872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300816, "loss/out": 4.243319034576416, "created_at": "2025-01-15T03:49:03.266035+00:00"} {"global_step": 10514, "acc_step": 0, "speed/wps": 12962.10773094007, "speed/FLOPS": 203587713716979.62, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0726710557937622, "optim/lr": 0.002926240260245787, "optim/total_tokens": 2756182016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 4.35746955871582, "created_at": "2025-01-15T03:49:08.326873+00:00"} {"global_step": 10515, "acc_step": 0, "speed/wps": 12966.18104519817, "speed/FLOPS": 203651690714723.4, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 9.29032039642334, "optim/lr": 0.00292621372093045, "optim/total_tokens": 2756444160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473310, "loss/out": 4.278159141540527, "created_at": "2025-01-15T03:49:13.385327+00:00"} {"global_step": 10516, "acc_step": 0, "speed/wps": 12961.555337225185, "speed/FLOPS": 203579037614618.84, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3874831199645996, "optim/lr": 0.0029261871769618404, "optim/total_tokens": 2756706304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 4.2997307777404785, "created_at": "2025-01-15T03:49:18.448672+00:00"} {"global_step": 10517, "acc_step": 0, "speed/wps": 12965.461796477528, "speed/FLOPS": 203640393925214.44, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8581703901290894, "optim/lr": 0.002926160628340045, "optim/total_tokens": 2756968448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 4.258018493652344, "created_at": "2025-01-15T03:49:23.506843+00:00"} {"global_step": 10518, "acc_step": 0, "speed/wps": 12965.830257420159, "speed/FLOPS": 203646181110636.84, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.075526237487793, "optim/lr": 0.002926134075065149, "optim/total_tokens": 2757230592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 4.373497009277344, "created_at": "2025-01-15T03:49:28.567855+00:00"} {"global_step": 10519, "acc_step": 0, "speed/wps": 12964.881568195031, "speed/FLOPS": 203631280642720.6, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2967557907104492, "optim/lr": 0.002926107517137241, "optim/total_tokens": 2757492736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427066, "loss/out": 4.253925323486328, "created_at": "2025-01-15T03:49:33.626309+00:00"} {"global_step": 10520, "acc_step": 0, "speed/wps": 12959.856117603964, "speed/FLOPS": 203552349035495.12, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1607359647750854, "optim/lr": 0.002926080954556406, "optim/total_tokens": 2757754880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 4.356817722320557, "created_at": "2025-01-15T03:49:38.687316+00:00"} {"global_step": 10521, "acc_step": 0, "speed/wps": 12969.55940940396, "speed/FLOPS": 203704752566934.28, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.947054862976074, "optim/lr": 0.002926054387322731, "optim/total_tokens": 2758017024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470085, "loss/out": 4.2053680419921875, "created_at": "2025-01-15T03:49:43.747525+00:00"} {"global_step": 10522, "acc_step": 0, "speed/wps": 12969.092028875035, "speed/FLOPS": 203697411713479.34, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.680933952331543, "optim/lr": 0.0029260278154363043, "optim/total_tokens": 2758279168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449657, "loss/out": 4.198780059814453, "created_at": "2025-01-15T03:49:48.804540+00:00"} {"global_step": 10523, "acc_step": 0, "speed/wps": 12965.056201896912, "speed/FLOPS": 203634023504980.5, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2945659160614014, "optim/lr": 0.00292600123889721, "optim/total_tokens": 2758541312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327499, "loss/out": 4.177682876586914, "created_at": "2025-01-15T03:49:53.863103+00:00"} {"global_step": 10524, "acc_step": 0, "speed/wps": 12965.079795531616, "speed/FLOPS": 203634394075434.56, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3343803882598877, "optim/lr": 0.0029259746577055368, "optim/total_tokens": 2758803456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440624, "loss/out": 4.3510332107543945, "created_at": "2025-01-15T03:49:58.922900+00:00"} {"global_step": 10525, "acc_step": 0, "speed/wps": 12962.292153596867, "speed/FLOPS": 203590610328227.9, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2169979810714722, "optim/lr": 0.00292594807186137, "optim/total_tokens": 2759065600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457665, "loss/out": 4.285971641540527, "created_at": "2025-01-15T03:50:03.987965+00:00"} {"global_step": 10526, "acc_step": 0, "speed/wps": 12972.965467222319, "speed/FLOPS": 203758249385385.5, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.06730055809021, "optim/lr": 0.002925921481364798, "optim/total_tokens": 2759327744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472504, "loss/out": 4.324660301208496, "created_at": "2025-01-15T03:50:09.048395+00:00"} {"global_step": 10527, "acc_step": 0, "speed/wps": 12972.391569299949, "speed/FLOPS": 203749235529896.3, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1879481077194214, "optim/lr": 0.002925894886215906, "optim/total_tokens": 2759589888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437195, "loss/out": 4.149858474731445, "created_at": "2025-01-15T03:50:14.104357+00:00"} {"global_step": 10528, "acc_step": 0, "speed/wps": 12949.903422373, "speed/FLOPS": 203396028280456.4, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 16.54155921936035, "optim/lr": 0.0029258682864147815, "optim/total_tokens": 2759852032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 4.326020240783691, "created_at": "2025-01-15T03:50:19.168650+00:00"} {"global_step": 10529, "acc_step": 0, "speed/wps": 12961.541158170212, "speed/FLOPS": 203578814913076.44, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9966170191764832, "optim/lr": 0.0029258416819615113, "optim/total_tokens": 2760114176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449942, "loss/out": 4.2671098709106445, "created_at": "2025-01-15T03:50:24.228299+00:00"} {"global_step": 10530, "acc_step": 0, "speed/wps": 12965.050268267762, "speed/FLOPS": 203633930309181.03, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8539595603942871, "optim/lr": 0.002925815072856182, "optim/total_tokens": 2760376320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 4.299374580383301, "created_at": "2025-01-15T03:50:29.286691+00:00"} {"global_step": 10531, "acc_step": 0, "speed/wps": 12966.107688992, "speed/FLOPS": 203650538554703.56, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1476068496704102, "optim/lr": 0.002925788459098881, "optim/total_tokens": 2760638464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433179, "loss/out": 4.235065937042236, "created_at": "2025-01-15T03:50:34.346338+00:00"} {"global_step": 10532, "acc_step": 0, "speed/wps": 12944.936500863545, "speed/FLOPS": 203318015952885.03, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.001, "optim/grad_norm": 0.9446914196014404, "optim/lr": 0.002925761840689694, "optim/total_tokens": 2760900608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355503, "loss/out": 4.344833850860596, "created_at": "2025-01-15T03:50:39.417019+00:00"} {"global_step": 10533, "acc_step": 0, "speed/wps": 12966.810558425475, "speed/FLOPS": 203661578085000.1, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.2017415761947632, "optim/lr": 0.0029257352176287085, "optim/total_tokens": 2761162752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436381, "loss/out": 4.175110340118408, "created_at": "2025-01-15T03:50:44.475472+00:00"} {"global_step": 10534, "acc_step": 0, "speed/wps": 12965.065324504016, "speed/FLOPS": 203634166788060.84, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.191165566444397, "optim/lr": 0.0029257085899160114, "optim/total_tokens": 2761424896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 4.215967655181885, "created_at": "2025-01-15T03:50:49.538787+00:00"} {"global_step": 10535, "acc_step": 0, "speed/wps": 12970.382703593797, "speed/FLOPS": 203717683533510.44, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.001, "optim/grad_norm": 0.7416810393333435, "optim/lr": 0.0029256819575516897, "optim/total_tokens": 2761687040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 4.191911697387695, "created_at": "2025-01-15T03:50:54.597404+00:00"} {"global_step": 10536, "acc_step": 0, "speed/wps": 12963.80230120223, "speed/FLOPS": 203614329271530.4, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0442918539047241, "optim/lr": 0.0029256553205358297, "optim/total_tokens": 2761949184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466132, "loss/out": 4.194912910461426, "created_at": "2025-01-15T03:50:59.656510+00:00"} {"global_step": 10537, "acc_step": 0, "speed/wps": 12967.479958298221, "speed/FLOPS": 203672091929853.38, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1908483505249023, "optim/lr": 0.002925628678868519, "optim/total_tokens": 2762211328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462126, "loss/out": 4.2650556564331055, "created_at": "2025-01-15T03:51:04.717654+00:00"} {"global_step": 10538, "acc_step": 0, "speed/wps": 12963.905430283481, "speed/FLOPS": 203615949055465.66, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1449109315872192, "optim/lr": 0.0029256020325498446, "optim/total_tokens": 2762473472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331967, "loss/out": 4.282526969909668, "created_at": "2025-01-15T03:51:09.776334+00:00"} {"global_step": 10539, "acc_step": 0, "speed/wps": 12967.549081216821, "speed/FLOPS": 203673177600274.84, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.036305546760559, "optim/lr": 0.002925575381579893, "optim/total_tokens": 2762735616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366988, "loss/out": 4.189157009124756, "created_at": "2025-01-15T03:51:14.833710+00:00"} {"global_step": 10540, "acc_step": 0, "speed/wps": 12964.045762124475, "speed/FLOPS": 203618153159864.62, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7162354588508606, "optim/lr": 0.0029255487259587506, "optim/total_tokens": 2762997760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452081, "loss/out": 4.326875686645508, "created_at": "2025-01-15T03:51:19.900636+00:00"} {"global_step": 10541, "acc_step": 0, "speed/wps": 12967.072103633483, "speed/FLOPS": 203665686011892.62, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.4231066703796387, "optim/lr": 0.0029255220656865056, "optim/total_tokens": 2763259904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 4.268004417419434, "created_at": "2025-01-15T03:51:24.959550+00:00"} {"global_step": 10542, "acc_step": 0, "speed/wps": 12959.780985565414, "speed/FLOPS": 203551168983587.47, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1317352056503296, "optim/lr": 0.002925495400763244, "optim/total_tokens": 2763522048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440921, "loss/out": 4.161001205444336, "created_at": "2025-01-15T03:51:30.028451+00:00"} {"global_step": 10543, "acc_step": 0, "speed/wps": 12963.90437703592, "speed/FLOPS": 203615932512765.4, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2097628116607666, "optim/lr": 0.0029254687311890534, "optim/total_tokens": 2763784192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 4.1408586502075195, "created_at": "2025-01-15T03:51:35.089062+00:00"} {"global_step": 10544, "acc_step": 0, "speed/wps": 12965.578597535676, "speed/FLOPS": 203642228446334.12, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0817939043045044, "optim/lr": 0.002925442056964021, "optim/total_tokens": 2764046336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496846, "loss/out": 4.153507232666016, "created_at": "2025-01-15T03:51:40.147965+00:00"} {"global_step": 10545, "acc_step": 0, "speed/wps": 12960.404750144167, "speed/FLOPS": 203560966063439.12, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0908761024475098, "optim/lr": 0.0029254153780882324, "optim/total_tokens": 2764308480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451559, "loss/out": 4.326079368591309, "created_at": "2025-01-15T03:51:45.208251+00:00"} {"global_step": 10546, "acc_step": 0, "speed/wps": 12962.346944891493, "speed/FLOPS": 203591470900801.12, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6744545698165894, "optim/lr": 0.0029253886945617763, "optim/total_tokens": 2764570624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323417, "loss/out": 4.237971782684326, "created_at": "2025-01-15T03:51:50.272375+00:00"} {"global_step": 10547, "acc_step": 0, "speed/wps": 12970.275357098057, "speed/FLOPS": 203715997509285.28, "speed/curr_iter_time": 1.2727, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9281144142150879, "optim/lr": 0.00292536200638474, "optim/total_tokens": 2764832768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358017, "loss/out": 4.225338459014893, "created_at": "2025-01-15T03:51:55.333776+00:00"} {"global_step": 10548, "acc_step": 0, "speed/wps": 12961.847157135315, "speed/FLOPS": 203583621047310.0, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2470678091049194, "optim/lr": 0.0029253353135572082, "optim/total_tokens": 2765094912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452081, "loss/out": 4.282700538635254, "created_at": "2025-01-15T03:52:00.397819+00:00"} {"global_step": 10549, "acc_step": 0, "speed/wps": 12969.914979253148, "speed/FLOPS": 203710337280019.94, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0619758367538452, "optim/lr": 0.0029253086160792702, "optim/total_tokens": 2765357056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 4.310608863830566, "created_at": "2025-01-15T03:52:05.458015+00:00"} {"global_step": 10550, "acc_step": 0, "speed/wps": 12965.62467783171, "speed/FLOPS": 203642952200704.3, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7460033297538757, "optim/lr": 0.002925281913951012, "optim/total_tokens": 2765619200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466155, "loss/out": 4.319001197814941, "created_at": "2025-01-15T03:52:10.526234+00:00"} {"global_step": 10551, "acc_step": 0, "speed/wps": 12966.823874493972, "speed/FLOPS": 203661787232154.03, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5451313257217407, "optim/lr": 0.0029252552071725216, "optim/total_tokens": 2765881344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 4.255563735961914, "created_at": "2025-01-15T03:52:15.584113+00:00"} {"global_step": 10552, "acc_step": 0, "speed/wps": 12962.929128915104, "speed/FLOPS": 203600614900896.62, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9052211046218872, "optim/lr": 0.0029252284957438854, "optim/total_tokens": 2766143488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 4.265274524688721, "created_at": "2025-01-15T03:52:20.674562+00:00"} {"global_step": 10553, "acc_step": 0, "speed/wps": 12961.130291805572, "speed/FLOPS": 203572361692230.22, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3921310901641846, "optim/lr": 0.002925201779665191, "optim/total_tokens": 2766405632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433993, "loss/out": 4.271703720092773, "created_at": "2025-01-15T03:52:25.738775+00:00"} {"global_step": 10554, "acc_step": 0, "speed/wps": 12969.346903903683, "speed/FLOPS": 203701414876039.25, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7852864861488342, "optim/lr": 0.0029251750589365253, "optim/total_tokens": 2766667776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 4.245924472808838, "created_at": "2025-01-15T03:52:30.802409+00:00"} {"global_step": 10555, "acc_step": 0, "speed/wps": 12961.905540362264, "speed/FLOPS": 203584538036116.22, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.686970055103302, "optim/lr": 0.0029251483335579753, "optim/total_tokens": 2766929920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436096, "loss/out": 4.221105575561523, "created_at": "2025-01-15T03:52:35.862268+00:00"} {"global_step": 10556, "acc_step": 0, "speed/wps": 12965.859857518964, "speed/FLOPS": 203646646020863.38, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2834888696670532, "optim/lr": 0.0029251216035296288, "optim/total_tokens": 2767192064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448083, "loss/out": 4.285118103027344, "created_at": "2025-01-15T03:52:40.925214+00:00"} {"global_step": 10557, "acc_step": 0, "speed/wps": 12963.94186176818, "speed/FLOPS": 203616521261994.97, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9469100832939148, "optim/lr": 0.0029250948688515728, "optim/total_tokens": 2767454208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452847, "loss/out": 4.249902248382568, "created_at": "2025-01-15T03:52:45.986056+00:00"} {"global_step": 10558, "acc_step": 0, "speed/wps": 12961.741081174125, "speed/FLOPS": 203581954978575.88, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.7942782044410706, "optim/lr": 0.002925068129523894, "optim/total_tokens": 2767716352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 4.120052337646484, "created_at": "2025-01-15T03:52:51.050756+00:00"} {"global_step": 10559, "acc_step": 0, "speed/wps": 12967.054985395696, "speed/FLOPS": 203665417146445.25, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.1923348903656006, "optim/lr": 0.00292504138554668, "optim/total_tokens": 2767978496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440919, "loss/out": 4.368181228637695, "created_at": "2025-01-15T03:52:56.109756+00:00"} {"global_step": 10560, "acc_step": 0, "speed/wps": 12960.977253727315, "speed/FLOPS": 203569958018914.8, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7133269309997559, "optim/lr": 0.0029250146369200185, "optim/total_tokens": 2768240640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 339173, "loss/out": 4.227960586547852, "created_at": "2025-01-15T03:53:01.199454+00:00"} {"global_step": 10561, "acc_step": 0, "speed/wps": 12959.226732898302, "speed/FLOPS": 203542463683827.47, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.001, "optim/grad_norm": 1.2165555953979492, "optim/lr": 0.002924987883643996, "optim/total_tokens": 2768502784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 4.274274826049805, "created_at": "2025-01-15T03:53:06.260831+00:00"} {"global_step": 10562, "acc_step": 0, "speed/wps": 12962.349824509538, "speed/FLOPS": 203591516129159.53, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7980066537857056, "optim/lr": 0.002924961125718701, "optim/total_tokens": 2768764928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435329, "loss/out": 4.260889530181885, "created_at": "2025-01-15T03:53:11.320516+00:00"} {"global_step": 10563, "acc_step": 0, "speed/wps": 12960.346454433342, "speed/FLOPS": 203560050449193.9, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.065675973892212, "optim/lr": 0.002924934363144219, "optim/total_tokens": 2769027072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 4.231425762176514, "created_at": "2025-01-15T03:53:16.384636+00:00"} {"global_step": 10564, "acc_step": 0, "speed/wps": 12955.500669287272, "speed/FLOPS": 203483940734668.3, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2139426469802856, "optim/lr": 0.0029249075959206384, "optim/total_tokens": 2769289216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442208, "loss/out": 4.397397994995117, "created_at": "2025-01-15T03:53:21.452057+00:00"} {"global_step": 10565, "acc_step": 0, "speed/wps": 12964.71942448446, "speed/FLOPS": 203628733952937.28, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1455103158950806, "optim/lr": 0.0029248808240480467, "optim/total_tokens": 2769551360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329804, "loss/out": 4.218643665313721, "created_at": "2025-01-15T03:53:26.513493+00:00"} {"global_step": 10566, "acc_step": 0, "speed/wps": 12956.799366178815, "speed/FLOPS": 203504338553946.53, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0859935283660889, "optim/lr": 0.002924854047526531, "optim/total_tokens": 2769813504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430268, "loss/out": 4.320338726043701, "created_at": "2025-01-15T03:53:31.578989+00:00"} {"global_step": 10567, "acc_step": 0, "speed/wps": 12966.108350780534, "speed/FLOPS": 203650548949001.88, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6891363859176636, "optim/lr": 0.002924827266356178, "optim/total_tokens": 2770075648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 4.3747382164001465, "created_at": "2025-01-15T03:53:36.642235+00:00"} {"global_step": 10568, "acc_step": 0, "speed/wps": 12967.29016440374, "speed/FLOPS": 203669110955936.4, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.090942621231079, "optim/lr": 0.002924800480537076, "optim/total_tokens": 2770337792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431618, "loss/out": 4.185136795043945, "created_at": "2025-01-15T03:53:41.732959+00:00"} {"global_step": 10569, "acc_step": 0, "speed/wps": 12967.903365067075, "speed/FLOPS": 203678742114980.1, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.059817910194397, "optim/lr": 0.0029247736900693123, "optim/total_tokens": 2770599936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330091, "loss/out": 4.315260887145996, "created_at": "2025-01-15T03:53:46.790676+00:00"} {"global_step": 10570, "acc_step": 0, "speed/wps": 12959.959236420485, "speed/FLOPS": 203553968658208.7, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.001, "optim/grad_norm": 0.7621628046035767, "optim/lr": 0.002924746894952974, "optim/total_tokens": 2770862080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323417, "loss/out": 4.308514595031738, "created_at": "2025-01-15T03:53:51.853039+00:00"} {"global_step": 10571, "acc_step": 0, "speed/wps": 12961.865472121184, "speed/FLOPS": 203583908709330.53, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1116889715194702, "optim/lr": 0.0029247200951881486, "optim/total_tokens": 2771124224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295474, "loss/out": 4.236701965332031, "created_at": "2025-01-15T03:53:56.914135+00:00"} {"global_step": 10572, "acc_step": 0, "speed/wps": 12960.014031863306, "speed/FLOPS": 203554829295935.03, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.042294979095459, "optim/lr": 0.0029246932907749236, "optim/total_tokens": 2771386368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377054, "loss/out": 4.074683666229248, "created_at": "2025-01-15T03:54:01.975650+00:00"} {"global_step": 10573, "acc_step": 0, "speed/wps": 12963.741204545471, "speed/FLOPS": 203613369664580.4, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0209189653396606, "optim/lr": 0.0029246664817133864, "optim/total_tokens": 2771648512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435866, "loss/out": 4.329995632171631, "created_at": "2025-01-15T03:54:07.036413+00:00"} {"global_step": 10574, "acc_step": 0, "speed/wps": 12965.133989315353, "speed/FLOPS": 203635245263278.22, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.4829925298690796, "optim/lr": 0.0029246396680036243, "optim/total_tokens": 2771910656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 4.290709495544434, "created_at": "2025-01-15T03:54:12.096111+00:00"} {"global_step": 10575, "acc_step": 0, "speed/wps": 12961.988423338944, "speed/FLOPS": 203585839827158.12, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1417850255966187, "optim/lr": 0.0029246128496457253, "optim/total_tokens": 2772172800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443543, "loss/out": 4.365444660186768, "created_at": "2025-01-15T03:54:17.160643+00:00"} {"global_step": 10576, "acc_step": 0, "speed/wps": 12954.219033248573, "speed/FLOPS": 203463810879528.06, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.6738079786300659, "optim/lr": 0.002924586026639776, "optim/total_tokens": 2772434944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 4.1679277420043945, "created_at": "2025-01-15T03:54:22.231616+00:00"} {"global_step": 10577, "acc_step": 0, "speed/wps": 12964.596005840162, "speed/FLOPS": 203626795493533.6, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7232572436332703, "optim/lr": 0.0029245591989858653, "optim/total_tokens": 2772697088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325000, "loss/out": 4.175986289978027, "created_at": "2025-01-15T03:54:27.295429+00:00"} {"global_step": 10578, "acc_step": 0, "speed/wps": 12960.59580105224, "speed/FLOPS": 203563966780481.88, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0932644605636597, "optim/lr": 0.0029245323666840795, "optim/total_tokens": 2772959232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377054, "loss/out": 4.322822570800781, "created_at": "2025-01-15T03:54:32.361131+00:00"} {"global_step": 10579, "acc_step": 0, "speed/wps": 12964.282967268806, "speed/FLOPS": 203621878792610.38, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5864180326461792, "optim/lr": 0.0029245055297345062, "optim/total_tokens": 2773221376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454718, "loss/out": 4.249649524688721, "created_at": "2025-01-15T03:54:37.420717+00:00"} {"global_step": 10580, "acc_step": 0, "speed/wps": 12955.143680730705, "speed/FLOPS": 203478333738833.75, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1906020641326904, "optim/lr": 0.0029244786881372337, "optim/total_tokens": 2773483520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 4.280216693878174, "created_at": "2025-01-15T03:54:42.483195+00:00"} {"global_step": 10581, "acc_step": 0, "speed/wps": 12961.051831467674, "speed/FLOPS": 203571129364807.72, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1511894464492798, "optim/lr": 0.0029244518418923495, "optim/total_tokens": 2773745664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437195, "loss/out": 4.3522515296936035, "created_at": "2025-01-15T03:54:47.545812+00:00"} {"global_step": 10582, "acc_step": 0, "speed/wps": 12963.854430370175, "speed/FLOPS": 203615148031745.97, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7681204676628113, "optim/lr": 0.00292442499099994, "optim/total_tokens": 2774007808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 4.146934509277344, "created_at": "2025-01-15T03:54:52.608352+00:00"} {"global_step": 10583, "acc_step": 0, "speed/wps": 12963.000956670148, "speed/FLOPS": 203601743054491.28, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5419487953186035, "optim/lr": 0.0029243981354600945, "optim/total_tokens": 2774269952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441450, "loss/out": 4.161708831787109, "created_at": "2025-01-15T03:54:57.667589+00:00"} {"global_step": 10584, "acc_step": 0, "speed/wps": 12964.762027989957, "speed/FLOPS": 203629403099534.47, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1624191999435425, "optim/lr": 0.0029243712752728994, "optim/total_tokens": 2774532096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461913, "loss/out": 4.253904342651367, "created_at": "2025-01-15T03:55:02.726221+00:00"} {"global_step": 10585, "acc_step": 0, "speed/wps": 12960.148939898396, "speed/FLOPS": 203556948212009.2, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0466384887695312, "optim/lr": 0.002924344410438443, "optim/total_tokens": 2774794240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322598, "loss/out": 4.337368011474609, "created_at": "2025-01-15T03:55:07.791000+00:00"} {"global_step": 10586, "acc_step": 0, "speed/wps": 12962.66495462169, "speed/FLOPS": 203596465680605.06, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0542123317718506, "optim/lr": 0.0029243175409568124, "optim/total_tokens": 2775056384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362577, "loss/out": 4.1978864669799805, "created_at": "2025-01-15T03:55:12.850289+00:00"} {"global_step": 10587, "acc_step": 0, "speed/wps": 12960.456489160622, "speed/FLOPS": 203561778695789.4, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2265880107879639, "optim/lr": 0.0029242906668280957, "optim/total_tokens": 2775318528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443016, "loss/out": 4.2713093757629395, "created_at": "2025-01-15T03:55:17.911494+00:00"} {"global_step": 10588, "acc_step": 0, "speed/wps": 12954.924821466928, "speed/FLOPS": 203474896253351.97, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0180151462554932, "optim/lr": 0.0029242637880523806, "optim/total_tokens": 2775580672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 4.278908729553223, "created_at": "2025-01-15T03:55:22.977303+00:00"} {"global_step": 10589, "acc_step": 0, "speed/wps": 12966.144261869098, "speed/FLOPS": 203651112981994.78, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1396499872207642, "optim/lr": 0.0029242369046297545, "optim/total_tokens": 2775842816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 4.27894926071167, "created_at": "2025-01-15T03:55:28.038144+00:00"} {"global_step": 10590, "acc_step": 0, "speed/wps": 12965.80100386369, "speed/FLOPS": 203645721643333.9, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6236485242843628, "optim/lr": 0.002924210016560305, "optim/total_tokens": 2776104960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 4.2774505615234375, "created_at": "2025-01-15T03:55:33.101135+00:00"} {"global_step": 10591, "acc_step": 0, "speed/wps": 12965.98956679341, "speed/FLOPS": 203648683283257.78, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0337647199630737, "optim/lr": 0.002924183123844121, "optim/total_tokens": 2776367104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451559, "loss/out": 4.195183753967285, "created_at": "2025-01-15T03:55:38.165062+00:00"} {"global_step": 10592, "acc_step": 0, "speed/wps": 12965.826473204195, "speed/FLOPS": 203646121674325.1, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.253320336341858, "optim/lr": 0.0029241562264812884, "optim/total_tokens": 2776629248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293250, "loss/out": 4.255016326904297, "created_at": "2025-01-15T03:55:43.247851+00:00"} {"global_step": 10593, "acc_step": 0, "speed/wps": 12964.677513953256, "speed/FLOPS": 203628075690453.62, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7143465280532837, "optim/lr": 0.002924129324471896, "optim/total_tokens": 2776891392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 4.1667327880859375, "created_at": "2025-01-15T03:55:48.310632+00:00"} {"global_step": 10594, "acc_step": 0, "speed/wps": 12961.992916823234, "speed/FLOPS": 203585910403503.6, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8680347204208374, "optim/lr": 0.002924102417816032, "optim/total_tokens": 2777153536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442208, "loss/out": 4.276517868041992, "created_at": "2025-01-15T03:55:53.371111+00:00"} {"global_step": 10595, "acc_step": 0, "speed/wps": 12964.882800752139, "speed/FLOPS": 203631300001723.7, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2313512563705444, "optim/lr": 0.002924075506513783, "optim/total_tokens": 2777415680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 4.291059494018555, "created_at": "2025-01-15T03:55:58.430613+00:00"} {"global_step": 10596, "acc_step": 0, "speed/wps": 12969.39506429069, "speed/FLOPS": 203702171301096.5, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2540472745895386, "optim/lr": 0.002924048590565238, "optim/total_tokens": 2777677824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 4.237252712249756, "created_at": "2025-01-15T03:56:03.492767+00:00"} {"global_step": 10597, "acc_step": 0, "speed/wps": 12951.547022558943, "speed/FLOPS": 203421843279919.97, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1882152557373047, "optim/lr": 0.0029240216699704838, "optim/total_tokens": 2777939968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 4.20955228805542, "created_at": "2025-01-15T03:56:08.560767+00:00"} {"global_step": 10598, "acc_step": 0, "speed/wps": 12965.184098723697, "speed/FLOPS": 203636032300393.8, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.7661088109016418, "optim/lr": 0.0029239947447296088, "optim/total_tokens": 2778202112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431842, "loss/out": 4.206521987915039, "created_at": "2025-01-15T03:56:13.619104+00:00"} {"global_step": 10599, "acc_step": 0, "speed/wps": 12967.492604253144, "speed/FLOPS": 203672290551952.8, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6358019113540649, "optim/lr": 0.002923967814842701, "optim/total_tokens": 2778464256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 285549, "loss/out": 4.160770893096924, "created_at": "2025-01-15T03:56:18.680109+00:00"} {"global_step": 10600, "acc_step": 0, "speed/wps": 12969.506177796871, "speed/FLOPS": 203703916491397.03, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1455976963043213, "optim/lr": 0.002923940880309848, "optim/total_tokens": 2778726400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 378395, "loss/out": 4.20257568359375, "created_at": "2025-01-15T03:56:23.740553+00:00"} {"global_step": 10601, "acc_step": 0, "speed/wps": 12963.959332426279, "speed/FLOPS": 203616795662687.6, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0837955474853516, "optim/lr": 0.0029239139411311376, "optim/total_tokens": 2778988544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 4.346558570861816, "created_at": "2025-01-15T03:56:28.803995+00:00"} {"global_step": 10602, "acc_step": 0, "speed/wps": 12962.849128075151, "speed/FLOPS": 203599358377772.5, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.140688180923462, "optim/lr": 0.0029238869973066575, "optim/total_tokens": 2779250688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476715, "loss/out": 4.266712188720703, "created_at": "2025-01-15T03:56:33.867833+00:00"} {"global_step": 10603, "acc_step": 0, "speed/wps": 12959.445680259436, "speed/FLOPS": 203545902552999.28, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9814934730529785, "optim/lr": 0.002923860048836496, "optim/total_tokens": 2779512832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447555, "loss/out": 4.230587005615234, "created_at": "2025-01-15T03:56:38.938804+00:00"} {"global_step": 10604, "acc_step": 0, "speed/wps": 12962.992832755674, "speed/FLOPS": 203601615457251.03, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6432177424430847, "optim/lr": 0.0029238330957207406, "optim/total_tokens": 2779774976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429217, "loss/out": 4.2502875328063965, "created_at": "2025-01-15T03:56:43.999807+00:00"} {"global_step": 10605, "acc_step": 0, "speed/wps": 12957.613105288821, "speed/FLOPS": 203517119444863.9, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0548534393310547, "optim/lr": 0.00292380613795948, "optim/total_tokens": 2780037120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 287143, "loss/out": 4.160390853881836, "created_at": "2025-01-15T03:56:49.061778+00:00"} {"global_step": 10606, "acc_step": 0, "speed/wps": 12959.794101029427, "speed/FLOPS": 203551374979972.3, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0647664070129395, "optim/lr": 0.002923779175552802, "optim/total_tokens": 2780299264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450746, "loss/out": 4.149278163909912, "created_at": "2025-01-15T03:56:54.122957+00:00"} {"global_step": 10607, "acc_step": 0, "speed/wps": 12960.483856701307, "speed/FLOPS": 203562208540622.8, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1057251691818237, "optim/lr": 0.002923752208500794, "optim/total_tokens": 2780561408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 4.232489585876465, "created_at": "2025-01-15T03:56:59.183062+00:00"} {"global_step": 10608, "acc_step": 0, "speed/wps": 12961.337999283149, "speed/FLOPS": 203575624023585.53, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8435779213905334, "optim/lr": 0.0029237252368035445, "optim/total_tokens": 2780823552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 4.208357334136963, "created_at": "2025-01-15T03:57:04.248981+00:00"} {"global_step": 10609, "acc_step": 0, "speed/wps": 12964.098375704727, "speed/FLOPS": 203618979528441.12, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4647318422794342, "optim/lr": 0.0029236982604611407, "optim/total_tokens": 2781085696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442778, "loss/out": 4.238373756408691, "created_at": "2025-01-15T03:57:09.313719+00:00"} {"global_step": 10610, "acc_step": 0, "speed/wps": 12964.392958978522, "speed/FLOPS": 203623606363555.9, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.99269700050354, "optim/lr": 0.0029236712794736717, "optim/total_tokens": 2781347840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450990, "loss/out": 4.11305046081543, "created_at": "2025-01-15T03:57:14.381704+00:00"} {"global_step": 10611, "acc_step": 0, "speed/wps": 12964.083732515963, "speed/FLOPS": 203618749537039.72, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0041899681091309, "optim/lr": 0.0029236442938412243, "optim/total_tokens": 2781609984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 4.255886077880859, "created_at": "2025-01-15T03:57:19.443759+00:00"} {"global_step": 10612, "acc_step": 0, "speed/wps": 12962.245502446973, "speed/FLOPS": 203589877607813.4, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7208376526832581, "optim/lr": 0.002923617303563888, "optim/total_tokens": 2781872128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423627, "loss/out": 4.079983711242676, "created_at": "2025-01-15T03:57:24.504645+00:00"} {"global_step": 10613, "acc_step": 0, "speed/wps": 12965.056070516177, "speed/FLOPS": 203634021441465.53, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0783729553222656, "optim/lr": 0.00292359030864175, "optim/total_tokens": 2782134272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296240, "loss/out": 4.274422645568848, "created_at": "2025-01-15T03:57:29.563513+00:00"} {"global_step": 10614, "acc_step": 0, "speed/wps": 12963.057776596503, "speed/FLOPS": 203602635489513.44, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.2913652658462524, "optim/lr": 0.002923563309074898, "optim/total_tokens": 2782396416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434505, "loss/out": 4.245418548583984, "created_at": "2025-01-15T03:57:34.626266+00:00"} {"global_step": 10615, "acc_step": 0, "speed/wps": 12968.977702883723, "speed/FLOPS": 203695616066685.44, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9162883758544922, "optim/lr": 0.002923536304863421, "optim/total_tokens": 2782658560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456851, "loss/out": 4.20988130569458, "created_at": "2025-01-15T03:57:39.684053+00:00"} {"global_step": 10616, "acc_step": 0, "speed/wps": 12965.665844351726, "speed/FLOPS": 203643598777469.88, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1005523204803467, "optim/lr": 0.0029235092960074064, "optim/total_tokens": 2782920704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 4.234920501708984, "created_at": "2025-01-15T03:57:44.742136+00:00"} {"global_step": 10617, "acc_step": 0, "speed/wps": 12968.968063271008, "speed/FLOPS": 203695464663321.5, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9057855606079102, "optim/lr": 0.002923482282506943, "optim/total_tokens": 2783182848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441158, "loss/out": 4.152195930480957, "created_at": "2025-01-15T03:57:49.799705+00:00"} {"global_step": 10618, "acc_step": 0, "speed/wps": 12963.199222025274, "speed/FLOPS": 203604857084338.7, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0458484888076782, "optim/lr": 0.002923455264362118, "optim/total_tokens": 2783444992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 4.243740081787109, "created_at": "2025-01-15T03:57:54.858861+00:00"} {"global_step": 10619, "acc_step": 0, "speed/wps": 12967.548921109645, "speed/FLOPS": 203673175085571.62, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1169310808181763, "optim/lr": 0.0029234282415730208, "optim/total_tokens": 2783707136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456851, "loss/out": 4.262622833251953, "created_at": "2025-01-15T03:57:59.923032+00:00"} {"global_step": 10620, "acc_step": 0, "speed/wps": 12959.474000591668, "speed/FLOPS": 203546347362733.22, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2301117181777954, "optim/lr": 0.0029234012141397384, "optim/total_tokens": 2783969280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309312, "loss/out": 4.118058681488037, "created_at": "2025-01-15T03:58:04.984883+00:00"} {"global_step": 10621, "acc_step": 0, "speed/wps": 12965.12409026272, "speed/FLOPS": 203635089785053.9, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1024608612060547, "optim/lr": 0.00292337418206236, "optim/total_tokens": 2784231424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371118, "loss/out": 4.143997669219971, "created_at": "2025-01-15T03:58:10.043376+00:00"} {"global_step": 10622, "acc_step": 0, "speed/wps": 12965.017060110184, "speed/FLOPS": 203633408729433.47, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.027970314025879, "optim/lr": 0.0029233471453409725, "optim/total_tokens": 2784493568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444641, "loss/out": 4.211863994598389, "created_at": "2025-01-15T03:58:15.103620+00:00"} {"global_step": 10623, "acc_step": 0, "speed/wps": 12966.554698336115, "speed/FLOPS": 203657559450708.06, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7991952300071716, "optim/lr": 0.002923320103975665, "optim/total_tokens": 2784755712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457920, "loss/out": 4.185888767242432, "created_at": "2025-01-15T03:58:20.167481+00:00"} {"global_step": 10624, "acc_step": 0, "speed/wps": 12956.699809334552, "speed/FLOPS": 203502774876901.94, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6832506656646729, "optim/lr": 0.0029232930579665257, "optim/total_tokens": 2785017856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 4.292316436767578, "created_at": "2025-01-15T03:58:25.238074+00:00"} {"global_step": 10625, "acc_step": 0, "speed/wps": 12964.477032204837, "speed/FLOPS": 203624926849103.9, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1090677976608276, "optim/lr": 0.0029232660073136426, "optim/total_tokens": 2785280000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 4.177611827850342, "created_at": "2025-01-15T03:58:30.297643+00:00"} {"global_step": 10626, "acc_step": 0, "speed/wps": 12961.899931939637, "speed/FLOPS": 203584449948132.06, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 7.750130653381348, "optim/lr": 0.0029232389520171044, "optim/total_tokens": 2785542144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 4.146602153778076, "created_at": "2025-01-15T03:58:35.362292+00:00"} {"global_step": 10627, "acc_step": 0, "speed/wps": 12965.452924145595, "speed/FLOPS": 203640254573049.56, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.156846523284912, "optim/lr": 0.002923211892076999, "optim/total_tokens": 2785804288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424963, "loss/out": 4.304698944091797, "created_at": "2025-01-15T03:58:40.427513+00:00"} {"global_step": 10628, "acc_step": 0, "speed/wps": 12959.313431244933, "speed/FLOPS": 203543825400497.34, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0586791038513184, "optim/lr": 0.0029231848274934146, "optim/total_tokens": 2786066432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332493, "loss/out": 4.228754043579102, "created_at": "2025-01-15T03:58:45.488161+00:00"} {"global_step": 10629, "acc_step": 0, "speed/wps": 12957.650479681428, "speed/FLOPS": 203517706461057.56, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1323355436325073, "optim/lr": 0.0029231577582664397, "optim/total_tokens": 2786328576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439871, "loss/out": 4.113386154174805, "created_at": "2025-01-15T03:58:50.553425+00:00"} {"global_step": 10630, "acc_step": 0, "speed/wps": 12964.06130325606, "speed/FLOPS": 203618397254691.9, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7959734201431274, "optim/lr": 0.002923130684396162, "optim/total_tokens": 2786590720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 4.260449409484863, "created_at": "2025-01-15T03:58:55.618371+00:00"} {"global_step": 10631, "acc_step": 0, "speed/wps": 12964.139721216958, "speed/FLOPS": 203619628916525.4, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1499465703964233, "optim/lr": 0.002923103605882671, "optim/total_tokens": 2786852864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443022, "loss/out": 4.139590263366699, "created_at": "2025-01-15T03:59:00.679470+00:00"} {"global_step": 10632, "acc_step": 0, "speed/wps": 12963.161767855981, "speed/FLOPS": 203604268815142.53, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.179052472114563, "optim/lr": 0.0029230765227260544, "optim/total_tokens": 2787115008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467734, "loss/out": 4.37868595123291, "created_at": "2025-01-15T03:59:05.742989+00:00"} {"global_step": 10633, "acc_step": 0, "speed/wps": 12961.203221668013, "speed/FLOPS": 203573507155935.97, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.113520622253418, "optim/lr": 0.002923049434926401, "optim/total_tokens": 2787377152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465870, "loss/out": 4.243452072143555, "created_at": "2025-01-15T03:59:10.808451+00:00"} {"global_step": 10634, "acc_step": 0, "speed/wps": 12960.265640478647, "speed/FLOPS": 203558781154985.4, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5827738046646118, "optim/lr": 0.0029230223424837978, "optim/total_tokens": 2787639296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322837, "loss/out": 4.2136054039001465, "created_at": "2025-01-15T03:59:15.868689+00:00"} {"global_step": 10635, "acc_step": 0, "speed/wps": 12967.189832168968, "speed/FLOPS": 203667535100318.38, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.187153935432434, "optim/lr": 0.002922995245398335, "optim/total_tokens": 2787901440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291959, "loss/out": 4.218643665313721, "created_at": "2025-01-15T03:59:20.928431+00:00"} {"global_step": 10636, "acc_step": 0, "speed/wps": 12956.424054096404, "speed/FLOPS": 203498443761959.4, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 12.293766021728516, "optim/lr": 0.0029229681436701, "optim/total_tokens": 2788163584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351996, "loss/out": 4.12351131439209, "created_at": "2025-01-15T03:59:25.992269+00:00"} {"global_step": 10637, "acc_step": 0, "speed/wps": 12962.09354828886, "speed/FLOPS": 203587490958953.38, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5377453565597534, "optim/lr": 0.002922941037299181, "optim/total_tokens": 2788425728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 341043, "loss/out": 4.316032409667969, "created_at": "2025-01-15T03:59:31.054135+00:00"} {"global_step": 10638, "acc_step": 0, "speed/wps": 12962.293614883065, "speed/FLOPS": 203590633279735.66, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8758808374404907, "optim/lr": 0.0029229139262856676, "optim/total_tokens": 2788687872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 4.286914348602295, "created_at": "2025-01-15T03:59:36.118464+00:00"} {"global_step": 10639, "acc_step": 0, "speed/wps": 12957.587155218116, "speed/FLOPS": 203516711863344.34, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8658631443977356, "optim/lr": 0.002922886810629647, "optim/total_tokens": 2788950016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493946, "loss/out": 4.281424522399902, "created_at": "2025-01-15T03:59:41.180007+00:00"} {"global_step": 10640, "acc_step": 0, "speed/wps": 12964.636886363665, "speed/FLOPS": 203627437578330.94, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9634572863578796, "optim/lr": 0.0029228596903312083, "optim/total_tokens": 2789212160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443543, "loss/out": 4.346070289611816, "created_at": "2025-01-15T03:59:46.251551+00:00"} {"global_step": 10641, "acc_step": 0, "speed/wps": 12962.960800465491, "speed/FLOPS": 203601112346116.72, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1031261682510376, "optim/lr": 0.00292283256539044, "optim/total_tokens": 2789474304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 4.242938041687012, "created_at": "2025-01-15T03:59:51.312339+00:00"} {"global_step": 10642, "acc_step": 0, "speed/wps": 12963.553289722542, "speed/FLOPS": 203610418204064.28, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0311522483825684, "optim/lr": 0.00292280543580743, "optim/total_tokens": 2789736448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 339985, "loss/out": 4.195181846618652, "created_at": "2025-01-15T03:59:56.371836+00:00"} {"global_step": 10643, "acc_step": 0, "speed/wps": 12959.369223617932, "speed/FLOPS": 203544701696382.03, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2055063247680664, "optim/lr": 0.002922778301582268, "optim/total_tokens": 2789998592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 333837, "loss/out": 4.335334777832031, "created_at": "2025-01-15T04:00:01.432395+00:00"} {"global_step": 10644, "acc_step": 0, "speed/wps": 12960.311988552412, "speed/FLOPS": 203559509115172.8, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8597995042800903, "optim/lr": 0.002922751162715041, "optim/total_tokens": 2790260736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448890, "loss/out": 4.307543754577637, "created_at": "2025-01-15T04:00:06.497080+00:00"} {"global_step": 10645, "acc_step": 0, "speed/wps": 12968.823836489622, "speed/FLOPS": 203693199383531.5, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.197152853012085, "optim/lr": 0.002922724019205839, "optim/total_tokens": 2790522880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 4.075220584869385, "created_at": "2025-01-15T04:00:11.553982+00:00"} {"global_step": 10646, "acc_step": 0, "speed/wps": 12966.04448730361, "speed/FLOPS": 203649545885339.16, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0612943172454834, "optim/lr": 0.0029226968710547494, "optim/total_tokens": 2790785024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466654, "loss/out": 4.182765007019043, "created_at": "2025-01-15T04:00:16.617048+00:00"} {"global_step": 10647, "acc_step": 0, "speed/wps": 12965.54119889449, "speed/FLOPS": 203641641049283.2, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.94571852684021, "optim/lr": 0.002922669718261862, "optim/total_tokens": 2791047168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 4.35036039352417, "created_at": "2025-01-15T04:00:21.676411+00:00"} {"global_step": 10648, "acc_step": 0, "speed/wps": 12961.826528664124, "speed/FLOPS": 203583297048823.47, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9639400243759155, "optim/lr": 0.0029226425608272636, "optim/total_tokens": 2791309312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468515, "loss/out": 4.133401393890381, "created_at": "2025-01-15T04:00:26.736083+00:00"} {"global_step": 10649, "acc_step": 0, "speed/wps": 12958.596459549044, "speed/FLOPS": 203532564374791.06, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7110073566436768, "optim/lr": 0.0029226153987510447, "optim/total_tokens": 2791571456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424963, "loss/out": 4.211731910705566, "created_at": "2025-01-15T04:00:31.805601+00:00"} {"global_step": 10650, "acc_step": 0, "speed/wps": 12966.025470056711, "speed/FLOPS": 203649247193344.2, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.344019889831543, "optim/lr": 0.002922588232033293, "optim/total_tokens": 2791833600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335714, "loss/out": 4.218532085418701, "created_at": "2025-01-15T04:00:36.867684+00:00"} {"global_step": 10651, "acc_step": 0, "speed/wps": 12957.766242885993, "speed/FLOPS": 203519524681260.6, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8981586694717407, "optim/lr": 0.0029225610606740974, "optim/total_tokens": 2792095744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 346666, "loss/out": 4.186249732971191, "created_at": "2025-01-15T04:00:41.932908+00:00"} {"global_step": 10652, "acc_step": 0, "speed/wps": 12961.151600829091, "speed/FLOPS": 203572696379726.25, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.7523312568664551, "optim/lr": 0.0029225338846735464, "optim/total_tokens": 2792357888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441687, "loss/out": 4.31378698348999, "created_at": "2025-01-15T04:00:46.994925+00:00"} {"global_step": 10653, "acc_step": 0, "speed/wps": 12958.521036664297, "speed/FLOPS": 203531379754744.2, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.035060167312622, "optim/lr": 0.002922506704031728, "optim/total_tokens": 2792620032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 4.186607837677002, "created_at": "2025-01-15T04:00:52.055798+00:00"} {"global_step": 10654, "acc_step": 0, "speed/wps": 12958.515106629115, "speed/FLOPS": 203531286615392.94, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5044670701026917, "optim/lr": 0.0029224795187487323, "optim/total_tokens": 2792882176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436679, "loss/out": 4.311343193054199, "created_at": "2025-01-15T04:00:57.118127+00:00"} {"global_step": 10655, "acc_step": 0, "speed/wps": 12960.648334116793, "speed/FLOPS": 203564791884448.7, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3967339992523193, "optim/lr": 0.002922452328824647, "optim/total_tokens": 2793144320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 4.143279075622559, "created_at": "2025-01-15T04:01:02.178638+00:00"} {"global_step": 10656, "acc_step": 0, "speed/wps": 12965.821429975369, "speed/FLOPS": 203646042463486.25, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5550222396850586, "optim/lr": 0.002922425134259561, "optim/total_tokens": 2793406464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453124, "loss/out": 4.198128700256348, "created_at": "2025-01-15T04:01:07.238175+00:00"} {"global_step": 10657, "acc_step": 0, "speed/wps": 12964.245370695606, "speed/FLOPS": 203621288286765.16, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9447738528251648, "optim/lr": 0.002922397935053563, "optim/total_tokens": 2793668608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334369, "loss/out": 4.229076385498047, "created_at": "2025-01-15T04:01:12.299753+00:00"} {"global_step": 10658, "acc_step": 0, "speed/wps": 12959.322803528117, "speed/FLOPS": 203543972605083.62, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.6094481945037842, "optim/lr": 0.0029223707312067415, "optim/total_tokens": 2793930752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 376530, "loss/out": 4.05568790435791, "created_at": "2025-01-15T04:01:17.367083+00:00"} {"global_step": 10659, "acc_step": 0, "speed/wps": 12964.000125342398, "speed/FLOPS": 203617436371490.6, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0005, "optim/grad_norm": 0.7001274824142456, "optim/lr": 0.002922343522719186, "optim/total_tokens": 2794192896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 4.1536359786987305, "created_at": "2025-01-15T04:01:22.429394+00:00"} {"global_step": 10660, "acc_step": 0, "speed/wps": 12961.423498756407, "speed/FLOPS": 203576966910301.66, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.838932991027832, "optim/lr": 0.002922316309590985, "optim/total_tokens": 2794455040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453945, "loss/out": 4.108278274536133, "created_at": "2025-01-15T04:01:27.494434+00:00"} {"global_step": 10661, "acc_step": 0, "speed/wps": 12966.707196748865, "speed/FLOPS": 203659954647835.1, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8823164701461792, "optim/lr": 0.002922289091822227, "optim/total_tokens": 2794717184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467490, "loss/out": 4.332980155944824, "created_at": "2025-01-15T04:01:32.559523+00:00"} {"global_step": 10662, "acc_step": 0, "speed/wps": 12962.366820490204, "speed/FLOPS": 203591783074390.84, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7970668077468872, "optim/lr": 0.0029222618694130006, "optim/total_tokens": 2794979328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461385, "loss/out": 4.287816047668457, "created_at": "2025-01-15T04:01:37.623336+00:00"} {"global_step": 10663, "acc_step": 0, "speed/wps": 12963.716868260992, "speed/FLOPS": 203612987429791.1, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2094905376434326, "optim/lr": 0.0029222346423633957, "optim/total_tokens": 2795241472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439054, "loss/out": 4.163799285888672, "created_at": "2025-01-15T04:01:42.683443+00:00"} {"global_step": 10664, "acc_step": 0, "speed/wps": 12968.611226030358, "speed/FLOPS": 203689860044111.34, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3324254751205444, "optim/lr": 0.0029222074106734996, "optim/total_tokens": 2795503616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436381, "loss/out": 4.2395782470703125, "created_at": "2025-01-15T04:01:47.751688+00:00"} {"global_step": 10665, "acc_step": 0, "speed/wps": 12964.530112663777, "speed/FLOPS": 203625760550651.75, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8081458210945129, "optim/lr": 0.0029221801743434025, "optim/total_tokens": 2795765760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 396797, "loss/out": 4.230213642120361, "created_at": "2025-01-15T04:01:52.815813+00:00"} {"global_step": 10666, "acc_step": 0, "speed/wps": 12958.376615004878, "speed/FLOPS": 203529111414126.84, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9178382158279419, "optim/lr": 0.0029221529333731924, "optim/total_tokens": 2796027904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464271, "loss/out": 4.216465473175049, "created_at": "2025-01-15T04:01:57.876939+00:00"} {"global_step": 10667, "acc_step": 0, "speed/wps": 12965.919297027896, "speed/FLOPS": 203647579600029.6, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0683585405349731, "optim/lr": 0.002922125687762959, "optim/total_tokens": 2796290048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 4.179481506347656, "created_at": "2025-01-15T04:02:02.935878+00:00"} {"global_step": 10668, "acc_step": 0, "speed/wps": 12958.650193461046, "speed/FLOPS": 203533408339717.16, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1619359254837036, "optim/lr": 0.00292209843751279, "optim/total_tokens": 2796552192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447554, "loss/out": 4.272771835327148, "created_at": "2025-01-15T04:02:07.997300+00:00"} {"global_step": 10669, "acc_step": 0, "speed/wps": 12964.021630095429, "speed/FLOPS": 203617774133187.56, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1036635637283325, "optim/lr": 0.0029220711826227755, "optim/total_tokens": 2796814336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 4.2441301345825195, "created_at": "2025-01-15T04:02:13.056698+00:00"} {"global_step": 10670, "acc_step": 0, "speed/wps": 12962.229681532375, "speed/FLOPS": 203589629118609.44, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9382347464561462, "optim/lr": 0.0029220439230930035, "optim/total_tokens": 2797076480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 4.252084255218506, "created_at": "2025-01-15T04:02:18.121510+00:00"} {"global_step": 10671, "acc_step": 0, "speed/wps": 12965.131311505464, "speed/FLOPS": 203635203204594.22, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.03728187084198, "optim/lr": 0.0029220166589235632, "optim/total_tokens": 2797338624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297061, "loss/out": 4.1901774406433105, "created_at": "2025-01-15T04:02:23.187985+00:00"} {"global_step": 10672, "acc_step": 0, "speed/wps": 12961.383920522541, "speed/FLOPS": 203576345279752.6, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1294461488723755, "optim/lr": 0.0029219893901145447, "optim/total_tokens": 2797600768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445976, "loss/out": 4.3104448318481445, "created_at": "2025-01-15T04:02:28.247956+00:00"} {"global_step": 10673, "acc_step": 0, "speed/wps": 12964.667976275892, "speed/FLOPS": 203627925888124.47, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.157684326171875, "optim/lr": 0.0029219621166660346, "optim/total_tokens": 2797862912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467205, "loss/out": 4.220053672790527, "created_at": "2025-01-15T04:02:33.317817+00:00"} {"global_step": 10674, "acc_step": 0, "speed/wps": 12966.734237069011, "speed/FLOPS": 203660379353220.28, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9301350116729736, "optim/lr": 0.002921934838578124, "optim/total_tokens": 2798125056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432127, "loss/out": 4.158203125, "created_at": "2025-01-15T04:02:38.376864+00:00"} {"global_step": 10675, "acc_step": 0, "speed/wps": 12967.932458435207, "speed/FLOPS": 203679199066305.1, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.054703950881958, "optim/lr": 0.0029219075558509014, "optim/total_tokens": 2798387200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477516, "loss/out": 4.219352722167969, "created_at": "2025-01-15T04:02:43.434373+00:00"} {"global_step": 10676, "acc_step": 0, "speed/wps": 12954.110256222102, "speed/FLOPS": 203462102386852.03, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1068674325942993, "optim/lr": 0.002921880268484455, "optim/total_tokens": 2798649344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 4.079288482666016, "created_at": "2025-01-15T04:02:48.501719+00:00"} {"global_step": 10677, "acc_step": 0, "speed/wps": 12961.913572185837, "speed/FLOPS": 203584664186942.3, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.02853262424469, "optim/lr": 0.002921852976478875, "optim/total_tokens": 2798911488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467490, "loss/out": 4.164952754974365, "created_at": "2025-01-15T04:02:53.561420+00:00"} {"global_step": 10678, "acc_step": 0, "speed/wps": 12962.724309920044, "speed/FLOPS": 203597397937128.44, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0762766599655151, "optim/lr": 0.0029218256798342495, "optim/total_tokens": 2799173632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 4.15045166015625, "created_at": "2025-01-15T04:02:58.620883+00:00"} {"global_step": 10679, "acc_step": 0, "speed/wps": 12960.62319305235, "speed/FLOPS": 203564397009484.16, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6478856801986694, "optim/lr": 0.002921798378550668, "optim/total_tokens": 2799435776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297882, "loss/out": 4.246373653411865, "created_at": "2025-01-15T04:03:03.680950+00:00"} {"global_step": 10680, "acc_step": 0, "speed/wps": 12959.626795525719, "speed/FLOPS": 203548747217134.72, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2221695184707642, "optim/lr": 0.002921771072628219, "optim/total_tokens": 2799697920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413141, "loss/out": 4.209890365600586, "created_at": "2025-01-15T04:03:08.746163+00:00"} {"global_step": 10681, "acc_step": 0, "speed/wps": 12973.872126076467, "speed/FLOPS": 203772489708569.72, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5884672403335571, "optim/lr": 0.002921743762066993, "optim/total_tokens": 2799960064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428932, "loss/out": 4.073266983032227, "created_at": "2025-01-15T04:03:13.803277+00:00"} {"global_step": 10682, "acc_step": 0, "speed/wps": 12965.923869687935, "speed/FLOPS": 203647651419939.0, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0190410614013672, "optim/lr": 0.0029217164468670776, "optim/total_tokens": 2800222208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 4.143115043640137, "created_at": "2025-01-15T04:03:18.866926+00:00"} {"global_step": 10683, "acc_step": 0, "speed/wps": 12954.006770220161, "speed/FLOPS": 203460476996986.9, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9532247185707092, "optim/lr": 0.0029216891270285625, "optim/total_tokens": 2800484352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435044, "loss/out": 4.182796955108643, "created_at": "2025-01-15T04:03:23.933926+00:00"} {"global_step": 10684, "acc_step": 0, "speed/wps": 12958.223076412234, "speed/FLOPS": 203526699879545.25, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 21.111900329589844, "optim/lr": 0.0029216618025515373, "optim/total_tokens": 2800746496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 4.209685325622559, "created_at": "2025-01-15T04:03:29.000914+00:00"} {"global_step": 10685, "acc_step": 0, "speed/wps": 12964.26731334861, "speed/FLOPS": 203621632926282.94, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.415342330932617, "optim/lr": 0.00292163447343609, "optim/total_tokens": 2801008640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423627, "loss/out": 4.345008850097656, "created_at": "2025-01-15T04:03:34.061215+00:00"} {"global_step": 10686, "acc_step": 0, "speed/wps": 12965.230374486437, "speed/FLOPS": 203636759124837.22, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.203604817390442, "optim/lr": 0.0029216071396823105, "optim/total_tokens": 2801270784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289551, "loss/out": 4.187005996704102, "created_at": "2025-01-15T04:03:39.121239+00:00"} {"global_step": 10687, "acc_step": 0, "speed/wps": 12966.648182484047, "speed/FLOPS": 203659027747712.03, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5333865880966187, "optim/lr": 0.002921579801290288, "optim/total_tokens": 2801532928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 378395, "loss/out": 4.267511367797852, "created_at": "2025-01-15T04:03:44.179569+00:00"} {"global_step": 10688, "acc_step": 0, "speed/wps": 12960.59063537561, "speed/FLOPS": 203563885646432.03, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2796380519866943, "optim/lr": 0.002921552458260112, "optim/total_tokens": 2801795072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427066, "loss/out": 4.242990493774414, "created_at": "2025-01-15T04:03:49.244043+00:00"} {"global_step": 10689, "acc_step": 0, "speed/wps": 12961.054135179143, "speed/FLOPS": 203571165547761.97, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.038244605064392, "optim/lr": 0.002921525110591871, "optim/total_tokens": 2802057216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 4.094938278198242, "created_at": "2025-01-15T04:03:54.303999+00:00"} {"global_step": 10690, "acc_step": 0, "speed/wps": 12967.059042729377, "speed/FLOPS": 203665480872446.1, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.6938793659210205, "optim/lr": 0.002921497758285655, "optim/total_tokens": 2802319360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472504, "loss/out": 4.239964008331299, "created_at": "2025-01-15T04:03:59.362282+00:00"} {"global_step": 10691, "acc_step": 0, "speed/wps": 12965.513984381734, "speed/FLOPS": 203641213607963.7, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8719539642333984, "optim/lr": 0.0029214704013415527, "optim/total_tokens": 2802581504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 4.260421276092529, "created_at": "2025-01-15T04:04:04.423347+00:00"} {"global_step": 10692, "acc_step": 0, "speed/wps": 12966.448882195764, "speed/FLOPS": 203655897462817.25, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2047233581542969, "optim/lr": 0.002921443039759653, "optim/total_tokens": 2802843648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446505, "loss/out": 4.169981479644775, "created_at": "2025-01-15T04:04:09.482141+00:00"} {"global_step": 10693, "acc_step": 0, "speed/wps": 12965.19242126579, "speed/FLOPS": 203636163017353.78, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0005, "optim/grad_norm": 1.0465080738067627, "optim/lr": 0.002921415673540046, "optim/total_tokens": 2803105792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461385, "loss/out": 4.247374534606934, "created_at": "2025-01-15T04:04:14.540522+00:00"} {"global_step": 10694, "acc_step": 0, "speed/wps": 12968.473648992533, "speed/FLOPS": 203687699207681.66, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5020277500152588, "optim/lr": 0.0029213883026828207, "optim/total_tokens": 2803367936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 4.153408527374268, "created_at": "2025-01-15T04:04:19.601960+00:00"} {"global_step": 10695, "acc_step": 0, "speed/wps": 12968.167697952116, "speed/FLOPS": 203682893826171.12, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1152573823928833, "optim/lr": 0.002921360927188066, "optim/total_tokens": 2803630080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 341575, "loss/out": 4.225218296051025, "created_at": "2025-01-15T04:04:24.661691+00:00"} {"global_step": 10696, "acc_step": 0, "speed/wps": 12968.407083050175, "speed/FLOPS": 203686653698085.72, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5265600681304932, "optim/lr": 0.002921333547055872, "optim/total_tokens": 2803892224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322156, "loss/out": 4.318099021911621, "created_at": "2025-01-15T04:04:29.724445+00:00"} {"global_step": 10697, "acc_step": 0, "speed/wps": 12969.862538377227, "speed/FLOPS": 203709513624002.28, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0627317428588867, "optim/lr": 0.0029213061622863274, "optim/total_tokens": 2804154368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 423001, "loss/out": 4.331440448760986, "created_at": "2025-01-15T04:04:34.780962+00:00"} {"global_step": 10698, "acc_step": 0, "speed/wps": 12958.03786994249, "speed/FLOPS": 203523790957437.53, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.6373038291931152, "optim/lr": 0.0029212787728795215, "optim/total_tokens": 2804416512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459237, "loss/out": 4.274257659912109, "created_at": "2025-01-15T04:04:39.842146+00:00"} {"global_step": 10699, "acc_step": 0, "speed/wps": 12963.494688656987, "speed/FLOPS": 203609497793803.44, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1491491794586182, "optim/lr": 0.002921251378835544, "optim/total_tokens": 2804678656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 4.218632221221924, "created_at": "2025-01-15T04:04:44.901281+00:00"} {"global_step": 10700, "acc_step": 0, "speed/wps": 12961.966994571227, "speed/FLOPS": 203585503258914.9, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1620056629180908, "optim/lr": 0.002921223980154484, "optim/total_tokens": 2804940800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439583, "loss/out": 4.199672698974609, "created_at": "2025-01-15T04:04:49.969347+00:00"} {"global_step": 10701, "acc_step": 0, "speed/wps": 12961.147750111946, "speed/FLOPS": 203572635898922.03, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 6.91680908203125, "optim/lr": 0.002921196576836432, "optim/total_tokens": 2805202944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 4.268209457397461, "created_at": "2025-01-15T04:04:55.036095+00:00"} {"global_step": 10702, "acc_step": 0, "speed/wps": 12962.651231606982, "speed/FLOPS": 203596250141801.66, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.001, "optim/grad_norm": 1.1052980422973633, "optim/lr": 0.002921169168881475, "optim/total_tokens": 2805465088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436381, "loss/out": 4.169729232788086, "created_at": "2025-01-15T04:05:00.097069+00:00"} {"global_step": 10703, "acc_step": 0, "speed/wps": 12964.51961023186, "speed/FLOPS": 203625595595526.66, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9667961001396179, "optim/lr": 0.002921141756289705, "optim/total_tokens": 2805727232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437717, "loss/out": 4.288505554199219, "created_at": "2025-01-15T04:05:05.157095+00:00"} {"global_step": 10704, "acc_step": 0, "speed/wps": 12957.290829292027, "speed/FLOPS": 203512057657480.62, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.142866611480713, "optim/lr": 0.00292111433906121, "optim/total_tokens": 2805989376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323123, "loss/out": 4.216305255889893, "created_at": "2025-01-15T04:05:10.222561+00:00"} {"global_step": 10705, "acc_step": 0, "speed/wps": 12964.341673388943, "speed/FLOPS": 203622800852897.78, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0368776321411133, "optim/lr": 0.0029210869171960798, "optim/total_tokens": 2806251520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364683, "loss/out": 4.208750247955322, "created_at": "2025-01-15T04:05:15.281145+00:00"} {"global_step": 10706, "acc_step": 0, "speed/wps": 12963.54802377254, "speed/FLOPS": 203610335495083.34, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.35794997215271, "optim/lr": 0.0029210594906944037, "optim/total_tokens": 2806513664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457902, "loss/out": 4.246128559112549, "created_at": "2025-01-15T04:05:20.345573+00:00"} {"global_step": 10707, "acc_step": 0, "speed/wps": 12964.173328522493, "speed/FLOPS": 203620156765440.1, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9954395890235901, "optim/lr": 0.002921032059556271, "optim/total_tokens": 2806775808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491336, "loss/out": 4.326906204223633, "created_at": "2025-01-15T04:05:25.408542+00:00"} {"global_step": 10708, "acc_step": 0, "speed/wps": 12962.296363152604, "speed/FLOPS": 203590676445085.28, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3790555000305176, "optim/lr": 0.0029210046237817724, "optim/total_tokens": 2807037952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449127, "loss/out": 4.113638877868652, "created_at": "2025-01-15T04:05:30.469614+00:00"} {"global_step": 10709, "acc_step": 0, "speed/wps": 12961.324978576069, "speed/FLOPS": 203575419515488.53, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.666015386581421, "optim/lr": 0.002920977183370996, "optim/total_tokens": 2807300096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478850, "loss/out": 4.1796674728393555, "created_at": "2025-01-15T04:05:35.537610+00:00"} {"global_step": 10710, "acc_step": 0, "speed/wps": 12965.396642215086, "speed/FLOPS": 203639370587991.44, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.013353943824768, "optim/lr": 0.002920949738324032, "optim/total_tokens": 2807562240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444119, "loss/out": 4.19658088684082, "created_at": "2025-01-15T04:05:40.596113+00:00"} {"global_step": 10711, "acc_step": 0, "speed/wps": 12963.890906946148, "speed/FLOPS": 203615720946495.7, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0638879537582397, "optim/lr": 0.0029209222886409695, "optim/total_tokens": 2807824384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293306, "loss/out": 4.254059314727783, "created_at": "2025-01-15T04:05:45.656072+00:00"} {"global_step": 10712, "acc_step": 0, "speed/wps": 12963.00893970361, "speed/FLOPS": 203601868439001.62, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 13.616304397583008, "optim/lr": 0.002920894834321899, "optim/total_tokens": 2808086528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 397147, "loss/out": 4.1048173904418945, "created_at": "2025-01-15T04:05:50.718040+00:00"} {"global_step": 10713, "acc_step": 0, "speed/wps": 12967.09250507641, "speed/FLOPS": 203666006444588.3, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0734387636184692, "optim/lr": 0.002920867375366909, "optim/total_tokens": 2808348672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461100, "loss/out": 4.2474164962768555, "created_at": "2025-01-15T04:05:55.776166+00:00"} {"global_step": 10714, "acc_step": 0, "speed/wps": 12967.787002458937, "speed/FLOPS": 203676914480320.3, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3769851922988892, "optim/lr": 0.0029208399117760897, "optim/total_tokens": 2808610816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483333, "loss/out": 4.257410049438477, "created_at": "2025-01-15T04:06:00.833474+00:00"} {"global_step": 10715, "acc_step": 0, "speed/wps": 12970.236955971488, "speed/FLOPS": 203715394366823.84, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2652794122695923, "optim/lr": 0.002920812443549531, "optim/total_tokens": 2808872960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456028, "loss/out": 4.20425271987915, "created_at": "2025-01-15T04:06:05.895082+00:00"} {"global_step": 10716, "acc_step": 0, "speed/wps": 12965.604622626064, "speed/FLOPS": 203642637206140.9, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0822839736938477, "optim/lr": 0.0029207849706873214, "optim/total_tokens": 2809135104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443022, "loss/out": 4.303685188293457, "created_at": "2025-01-15T04:06:10.959382+00:00"} {"global_step": 10717, "acc_step": 0, "speed/wps": 12955.414782823364, "speed/FLOPS": 203482591769732.25, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.733306884765625, "optim/lr": 0.0029207574931895517, "optim/total_tokens": 2809397248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 4.178233623504639, "created_at": "2025-01-15T04:06:16.027380+00:00"} {"global_step": 10718, "acc_step": 0, "speed/wps": 12962.421044126835, "speed/FLOPS": 203592634731115.22, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.001, "optim/grad_norm": 1.538905143737793, "optim/lr": 0.0029207300110563107, "optim/total_tokens": 2809659392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455791, "loss/out": 4.198970794677734, "created_at": "2025-01-15T04:06:21.090814+00:00"} {"global_step": 10719, "acc_step": 0, "speed/wps": 12951.35394998328, "speed/FLOPS": 203418810809809.66, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5218392610549927, "optim/lr": 0.002920702524287688, "optim/total_tokens": 2809921536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 284929, "loss/out": 4.155703544616699, "created_at": "2025-01-15T04:06:26.158219+00:00"} {"global_step": 10720, "acc_step": 0, "speed/wps": 12955.3432442372, "speed/FLOPS": 203481468157932.22, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.010808229446411, "optim/lr": 0.002920675032883775, "optim/total_tokens": 2810183680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436096, "loss/out": 4.172366142272949, "created_at": "2025-01-15T04:06:31.226203+00:00"} {"global_step": 10721, "acc_step": 0, "speed/wps": 12964.498230419056, "speed/FLOPS": 203625259796187.62, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0014, "optim/grad_norm": 1.0033918619155884, "optim/lr": 0.002920647536844659, "optim/total_tokens": 2810445824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 4.226139068603516, "created_at": "2025-01-15T04:06:36.288402+00:00"} {"global_step": 10722, "acc_step": 0, "speed/wps": 12962.161329226128, "speed/FLOPS": 203588555551713.94, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.001, "optim/grad_norm": 0.6610222458839417, "optim/lr": 0.002920620036170431, "optim/total_tokens": 2810707968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440399, "loss/out": 4.245438575744629, "created_at": "2025-01-15T04:06:41.350153+00:00"} {"global_step": 10723, "acc_step": 0, "speed/wps": 12958.875543311773, "speed/FLOPS": 203536947768786.0, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7524574398994446, "optim/lr": 0.0029205925308611806, "optim/total_tokens": 2810970112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493135, "loss/out": 4.229001045227051, "created_at": "2025-01-15T04:06:46.415498+00:00"} {"global_step": 10724, "acc_step": 0, "speed/wps": 12961.789693636047, "speed/FLOPS": 203582718504090.88, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0009, "optim/grad_norm": 3.7367448806762695, "optim/lr": 0.0029205650209169977, "optim/total_tokens": 2811232256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448368, "loss/out": 4.216795921325684, "created_at": "2025-01-15T04:06:51.475816+00:00"} {"global_step": 10725, "acc_step": 0, "speed/wps": 12960.624119250107, "speed/FLOPS": 203564411556692.66, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1897575855255127, "optim/lr": 0.002920537506337972, "optim/total_tokens": 2811494400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309387, "loss/out": 4.259891510009766, "created_at": "2025-01-15T04:06:56.540821+00:00"} {"global_step": 10726, "acc_step": 0, "speed/wps": 12966.6404458467, "speed/FLOPS": 203658906233191.25, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1196328401565552, "optim/lr": 0.0029205099871241927, "optim/total_tokens": 2811756544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447791, "loss/out": 4.266635894775391, "created_at": "2025-01-15T04:07:01.602028+00:00"} {"global_step": 10727, "acc_step": 0, "speed/wps": 12968.37671848115, "speed/FLOPS": 203686176780803.9, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0850627422332764, "optim/lr": 0.0029204824632757496, "optim/total_tokens": 2812018688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457095, "loss/out": 4.213718414306641, "created_at": "2025-01-15T04:07:06.662589+00:00"} {"global_step": 10728, "acc_step": 0, "speed/wps": 12961.756662143529, "speed/FLOPS": 203582199699110.88, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.245375394821167, "optim/lr": 0.002920454934792734, "optim/total_tokens": 2812280832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482002, "loss/out": 4.082197189331055, "created_at": "2025-01-15T04:07:11.722218+00:00"} {"global_step": 10729, "acc_step": 0, "speed/wps": 12962.582874854848, "speed/FLOPS": 203595176505077.38, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7883347868919373, "optim/lr": 0.0029204274016752336, "optim/total_tokens": 2812542976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 4.292023658752441, "created_at": "2025-01-15T04:07:16.787982+00:00"} {"global_step": 10730, "acc_step": 0, "speed/wps": 12959.465733123048, "speed/FLOPS": 203546217510777.84, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0604041814804077, "optim/lr": 0.0029203998639233397, "optim/total_tokens": 2812805120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453945, "loss/out": 4.225377559661865, "created_at": "2025-01-15T04:07:21.849437+00:00"} {"global_step": 10731, "acc_step": 0, "speed/wps": 12949.56249881678, "speed/FLOPS": 203390673607527.9, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8020002841949463, "optim/lr": 0.0029203723215371416, "optim/total_tokens": 2813067264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476183, "loss/out": 4.186727523803711, "created_at": "2025-01-15T04:07:26.914510+00:00"} {"global_step": 10732, "acc_step": 0, "speed/wps": 12944.566409502382, "speed/FLOPS": 203312203159499.12, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0015, "optim/grad_norm": 1.1424463987350464, "optim/lr": 0.002920344774516729, "optim/total_tokens": 2813329408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451808, "loss/out": 4.168774604797363, "created_at": "2025-01-15T04:07:31.981807+00:00"} {"global_step": 10733, "acc_step": 0, "speed/wps": 12960.897816486631, "speed/FLOPS": 203568710347891.7, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.112781047821045, "optim/lr": 0.0029203172228621926, "optim/total_tokens": 2813591552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306907, "loss/out": 4.151101589202881, "created_at": "2025-01-15T04:07:37.044844+00:00"} {"global_step": 10734, "acc_step": 0, "speed/wps": 12958.611789549832, "speed/FLOPS": 203532805153519.06, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6482800841331482, "optim/lr": 0.0029202896665736218, "optim/total_tokens": 2813853696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 385632, "loss/out": 4.216610908508301, "created_at": "2025-01-15T04:07:42.106388+00:00"} {"global_step": 10735, "acc_step": 0, "speed/wps": 12964.746040893764, "speed/FLOPS": 203629152000220.2, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9791932702064514, "optim/lr": 0.0029202621056511058, "optim/total_tokens": 2814115840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448368, "loss/out": 4.306554794311523, "created_at": "2025-01-15T04:07:47.165507+00:00"} {"global_step": 10736, "acc_step": 0, "speed/wps": 12962.69618073716, "speed/FLOPS": 203596956129657.7, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8714660406112671, "optim/lr": 0.0029202345400947353, "optim/total_tokens": 2814377984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 4.14457893371582, "created_at": "2025-01-15T04:07:52.225127+00:00"} {"global_step": 10737, "acc_step": 0, "speed/wps": 12962.61452322272, "speed/FLOPS": 203595673586184.22, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0394139289855957, "optim/lr": 0.0029202069699046, "optim/total_tokens": 2814640128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443543, "loss/out": 4.166630744934082, "created_at": "2025-01-15T04:07:57.288095+00:00"} {"global_step": 10738, "acc_step": 0, "speed/wps": 12960.969416430107, "speed/FLOPS": 203569834923392.6, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9666879177093506, "optim/lr": 0.00292017939508079, "optim/total_tokens": 2814902272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 4.178534984588623, "created_at": "2025-01-15T04:08:02.348140+00:00"} {"global_step": 10739, "acc_step": 0, "speed/wps": 12967.294862288163, "speed/FLOPS": 203669184742666.84, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6500180959701538, "optim/lr": 0.002920151815623396, "optim/total_tokens": 2815164416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484138, "loss/out": 4.118917465209961, "created_at": "2025-01-15T04:08:07.411578+00:00"} {"global_step": 10740, "acc_step": 0, "speed/wps": 12962.378943107185, "speed/FLOPS": 203591973476748.75, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9958715438842773, "optim/lr": 0.002920124231532506, "optim/total_tokens": 2815426560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 422528, "loss/out": 4.217686653137207, "created_at": "2025-01-15T04:08:12.474258+00:00"} {"global_step": 10741, "acc_step": 0, "speed/wps": 12962.15439781199, "speed/FLOPS": 203588446684330.12, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9503918290138245, "optim/lr": 0.0029200966428082117, "optim/total_tokens": 2815688704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296475, "loss/out": 4.128312110900879, "created_at": "2025-01-15T04:08:17.533922+00:00"} {"global_step": 10742, "acc_step": 0, "speed/wps": 12966.183247999636, "speed/FLOPS": 203651725312747.4, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0550490617752075, "optim/lr": 0.002920069049450603, "optim/total_tokens": 2815950848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322443, "loss/out": 4.10799503326416, "created_at": "2025-01-15T04:08:22.592176+00:00"} {"global_step": 10743, "acc_step": 0, "speed/wps": 12963.522534724447, "speed/FLOPS": 203609935154557.4, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1013305187225342, "optim/lr": 0.002920041451459769, "optim/total_tokens": 2816212992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329951, "loss/out": 4.226925849914551, "created_at": "2025-01-15T04:08:27.652262+00:00"} {"global_step": 10744, "acc_step": 0, "speed/wps": 12966.526784287273, "speed/FLOPS": 203657121022213.1, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9216855764389038, "optim/lr": 0.0029200138488358, "optim/total_tokens": 2816475136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365072, "loss/out": 4.241168975830078, "created_at": "2025-01-15T04:08:32.716904+00:00"} {"global_step": 10745, "acc_step": 0, "speed/wps": 12959.627721581071, "speed/FLOPS": 203548761762106.56, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.070878267288208, "optim/lr": 0.0029199862415787868, "optim/total_tokens": 2816737280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454182, "loss/out": 4.220542907714844, "created_at": "2025-01-15T04:08:37.782597+00:00"} {"global_step": 10746, "acc_step": 0, "speed/wps": 12963.726246914464, "speed/FLOPS": 203613134734431.5, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0011, "optim/grad_norm": 1.0206040143966675, "optim/lr": 0.002919958629688819, "optim/total_tokens": 2816999424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462672, "loss/out": 4.240159511566162, "created_at": "2025-01-15T04:08:42.842016+00:00"} {"global_step": 10747, "acc_step": 0, "speed/wps": 12966.599371099346, "speed/FLOPS": 203658261097841.44, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.9052278995513916, "optim/lr": 0.0029199310131659867, "optim/total_tokens": 2817261568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446218, "loss/out": 4.248578071594238, "created_at": "2025-01-15T04:08:47.899834+00:00"} {"global_step": 10748, "acc_step": 0, "speed/wps": 12961.50709657324, "speed/FLOPS": 203578279928890.44, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1346940994262695, "optim/lr": 0.0029199033920103796, "optim/total_tokens": 2817523712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478811, "loss/out": 4.2215118408203125, "created_at": "2025-01-15T04:08:52.965667+00:00"} {"global_step": 10749, "acc_step": 0, "speed/wps": 12963.384578453792, "speed/FLOPS": 203607768361754.97, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0993552207946777, "optim/lr": 0.002919875766222089, "optim/total_tokens": 2817785856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331148, "loss/out": 4.200050354003906, "created_at": "2025-01-15T04:08:58.029115+00:00"} {"global_step": 10750, "acc_step": 0, "speed/wps": 12957.415158696445, "speed/FLOPS": 203514010421623.5, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.049181342124939, "optim/lr": 0.002919848135801204, "optim/total_tokens": 2818048000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 378395, "loss/out": 4.1165947914123535, "created_at": "2025-01-15T04:09:03.091624+00:00"} {"global_step": 10751, "acc_step": 0, "speed/wps": 12962.18370415853, "speed/FLOPS": 203588906980773.7, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6905903220176697, "optim/lr": 0.0029198205007478144, "optim/total_tokens": 2818310144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445404, "loss/out": 4.267010688781738, "created_at": "2025-01-15T04:09:08.152276+00:00"} {"global_step": 10752, "acc_step": 0, "speed/wps": 12965.495195636706, "speed/FLOPS": 203640918504904.56, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8066617250442505, "optim/lr": 0.0029197928610620117, "optim/total_tokens": 2818572288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487328, "loss/out": 4.233098983764648, "created_at": "2025-01-15T04:09:13.210976+00:00"} {"global_step": 10753, "acc_step": 0, "speed/wps": 12960.559612757294, "speed/FLOPS": 203563398393581.8, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9936315417289734, "optim/lr": 0.0029197652167438845, "optim/total_tokens": 2818834432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446742, "loss/out": 4.146462440490723, "created_at": "2025-01-15T04:09:18.272708+00:00"} {"global_step": 10754, "acc_step": 0, "speed/wps": 12959.858781301096, "speed/FLOPS": 203552390872518.88, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.4830981492996216, "optim/lr": 0.002919737567793525, "optim/total_tokens": 2819096576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468018, "loss/out": 4.215142250061035, "created_at": "2025-01-15T04:09:23.333820+00:00"} {"global_step": 10755, "acc_step": 0, "speed/wps": 12954.46082507569, "speed/FLOPS": 203467608552429.72, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0318312644958496, "optim/lr": 0.0029197099142110215, "optim/total_tokens": 2819358720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 4.170035362243652, "created_at": "2025-01-15T04:09:28.400481+00:00"} {"global_step": 10756, "acc_step": 0, "speed/wps": 12962.225546038577, "speed/FLOPS": 203589564164996.6, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1456390619277954, "optim/lr": 0.0029196822559964647, "optim/total_tokens": 2819620864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461605, "loss/out": 4.262664318084717, "created_at": "2025-01-15T04:09:33.459945+00:00"} {"global_step": 10757, "acc_step": 0, "speed/wps": 12961.956773253056, "speed/FLOPS": 203585342719067.47, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0056382417678833, "optim/lr": 0.0029196545931499452, "optim/total_tokens": 2819883008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431320, "loss/out": 4.467921257019043, "created_at": "2025-01-15T04:09:38.524619+00:00"} {"global_step": 10758, "acc_step": 0, "speed/wps": 12965.816391528508, "speed/FLOPS": 203645963327754.72, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.642271876335144, "optim/lr": 0.0029196269256715536, "optim/total_tokens": 2820145152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318851, "loss/out": 4.105312347412109, "created_at": "2025-01-15T04:09:43.587178+00:00"} {"global_step": 10759, "acc_step": 0, "speed/wps": 12959.2941783696, "speed/FLOPS": 203543523007634.03, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0575493574142456, "optim/lr": 0.0029195992535613795, "optim/total_tokens": 2820407296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 400880, "loss/out": 4.173003196716309, "created_at": "2025-01-15T04:09:48.652952+00:00"} {"global_step": 10760, "acc_step": 0, "speed/wps": 12961.825370702121, "speed/FLOPS": 203583278861439.0, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.322181224822998, "optim/lr": 0.0029195715768195133, "optim/total_tokens": 2820669440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467182, "loss/out": 4.208499908447266, "created_at": "2025-01-15T04:09:53.714190+00:00"} {"global_step": 10761, "acc_step": 0, "speed/wps": 12958.74096963338, "speed/FLOPS": 203534834104243.22, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9000197649002075, "optim/lr": 0.002919543895446046, "optim/total_tokens": 2820931584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 4.250683307647705, "created_at": "2025-01-15T04:09:58.780210+00:00"} {"global_step": 10762, "acc_step": 0, "speed/wps": 12960.951770013096, "speed/FLOPS": 203569557762164.72, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.5547412633895874, "optim/lr": 0.0029195162094410665, "optim/total_tokens": 2821193728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458430, "loss/out": 4.0181803703308105, "created_at": "2025-01-15T04:10:03.840181+00:00"} {"global_step": 10763, "acc_step": 0, "speed/wps": 12963.502598265357, "speed/FLOPS": 203609622025071.8, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.5280423164367676, "optim/lr": 0.002919488518804667, "optim/total_tokens": 2821455872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 4.341672420501709, "created_at": "2025-01-15T04:10:08.899386+00:00"} {"global_step": 10764, "acc_step": 0, "speed/wps": 12961.136525033859, "speed/FLOPS": 203572459593645.8, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1363282203674316, "optim/lr": 0.002919460823536936, "optim/total_tokens": 2821718016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427351, "loss/out": 4.227696418762207, "created_at": "2025-01-15T04:10:13.959967+00:00"} {"global_step": 10765, "acc_step": 0, "speed/wps": 12958.434678229189, "speed/FLOPS": 203530023376851.88, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.0921845436096191, "optim/lr": 0.002919433123637965, "optim/total_tokens": 2821980160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334608, "loss/out": 4.260444641113281, "created_at": "2025-01-15T04:10:19.025562+00:00"} {"global_step": 10766, "acc_step": 0, "speed/wps": 12960.657916010361, "speed/FLOPS": 203564942381254.06, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5205367803573608, "optim/lr": 0.0029194054191078443, "optim/total_tokens": 2822242304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448361, "loss/out": 4.128640174865723, "created_at": "2025-01-15T04:10:24.087298+00:00"} {"global_step": 10767, "acc_step": 0, "speed/wps": 12966.238507024476, "speed/FLOPS": 203652593231666.4, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0011, "optim/grad_norm": 1.4434993267059326, "optim/lr": 0.002919377709946663, "optim/total_tokens": 2822504448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 4.134778022766113, "created_at": "2025-01-15T04:10:29.148873+00:00"} {"global_step": 10768, "acc_step": 0, "speed/wps": 12969.603194515135, "speed/FLOPS": 203705440272272.3, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6135656833648682, "optim/lr": 0.002919349996154514, "optim/total_tokens": 2822766592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457380, "loss/out": 4.1622772216796875, "created_at": "2025-01-15T04:10:34.206926+00:00"} {"global_step": 10769, "acc_step": 0, "speed/wps": 12968.306090637327, "speed/FLOPS": 203685067473463.8, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0382977724075317, "optim/lr": 0.002919322277731485, "optim/total_tokens": 2823028736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434807, "loss/out": 4.24697208404541, "created_at": "2025-01-15T04:10:39.267515+00:00"} {"global_step": 10770, "acc_step": 0, "speed/wps": 12963.362870436149, "speed/FLOPS": 203607427407507.94, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7448397278785706, "optim/lr": 0.002919294554677668, "optim/total_tokens": 2823290880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 4.211421966552734, "created_at": "2025-01-15T04:10:44.329315+00:00"} {"global_step": 10771, "acc_step": 0, "speed/wps": 12962.567924672341, "speed/FLOPS": 203594941691917.4, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0780436992645264, "optim/lr": 0.002919266826993154, "optim/total_tokens": 2823553024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441679, "loss/out": 4.169285297393799, "created_at": "2025-01-15T04:10:49.392620+00:00"} {"global_step": 10772, "acc_step": 0, "speed/wps": 12962.906827823532, "speed/FLOPS": 203600264631608.6, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0256584882736206, "optim/lr": 0.002919239094678032, "optim/total_tokens": 2823815168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 4.225315093994141, "created_at": "2025-01-15T04:10:54.451779+00:00"} {"global_step": 10773, "acc_step": 0, "speed/wps": 12964.451725807765, "speed/FLOPS": 203624529377363.25, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0803172588348389, "optim/lr": 0.002919211357732393, "optim/total_tokens": 2824077312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456045, "loss/out": 4.265953063964844, "created_at": "2025-01-15T04:10:59.539434+00:00"} {"global_step": 10774, "acc_step": 0, "speed/wps": 12966.121022729996, "speed/FLOPS": 203650747979381.25, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0359139442443848, "optim/lr": 0.0029191836161563275, "optim/total_tokens": 2824339456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443543, "loss/out": 4.301148414611816, "created_at": "2025-01-15T04:11:04.597393+00:00"} {"global_step": 10775, "acc_step": 0, "speed/wps": 12964.751283930827, "speed/FLOPS": 203629234349321.94, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.114537000656128, "optim/lr": 0.0029191558699499264, "optim/total_tokens": 2824601600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464799, "loss/out": 4.1717915534973145, "created_at": "2025-01-15T04:11:09.657482+00:00"} {"global_step": 10776, "acc_step": 0, "speed/wps": 12962.750617263784, "speed/FLOPS": 203597811130111.94, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.5896717309951782, "optim/lr": 0.00291912811911328, "optim/total_tokens": 2824863744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 4.1777191162109375, "created_at": "2025-01-15T04:11:14.717512+00:00"} {"global_step": 10777, "acc_step": 0, "speed/wps": 12967.33792240471, "speed/FLOPS": 203669861060968.0, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.710283100605011, "optim/lr": 0.0029191003636464784, "optim/total_tokens": 2825125888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448077, "loss/out": 4.297733783721924, "created_at": "2025-01-15T04:11:19.775102+00:00"} {"global_step": 10778, "acc_step": 0, "speed/wps": 12964.026437864833, "speed/FLOPS": 203617849645812.66, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.026152491569519, "optim/lr": 0.0029190726035496126, "optim/total_tokens": 2825388032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 4.227474212646484, "created_at": "2025-01-15T04:11:24.837277+00:00"} {"global_step": 10779, "acc_step": 0, "speed/wps": 12961.429586635039, "speed/FLOPS": 203577062528801.2, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0571799278259277, "optim/lr": 0.0029190448388227734, "optim/total_tokens": 2825650176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301877, "loss/out": 4.21607780456543, "created_at": "2025-01-15T04:11:29.899945+00:00"} {"global_step": 10780, "acc_step": 0, "speed/wps": 12960.68294978869, "speed/FLOPS": 203565335571142.72, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8950626254081726, "optim/lr": 0.0029190170694660513, "optim/total_tokens": 2825912320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355086, "loss/out": 4.259764671325684, "created_at": "2025-01-15T04:11:34.964208+00:00"} {"global_step": 10781, "acc_step": 0, "speed/wps": 12962.825986711316, "speed/FLOPS": 203598994910853.9, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.6452252864837646, "optim/lr": 0.002918989295479536, "optim/total_tokens": 2826174464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431842, "loss/out": 4.208089828491211, "created_at": "2025-01-15T04:11:40.027890+00:00"} {"global_step": 10782, "acc_step": 0, "speed/wps": 12965.979308072108, "speed/FLOPS": 203648522155942.78, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0252768993377686, "optim/lr": 0.0029189615168633194, "optim/total_tokens": 2826436608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490809, "loss/out": 4.146270751953125, "created_at": "2025-01-15T04:11:45.092877+00:00"} {"global_step": 10783, "acc_step": 0, "speed/wps": 12965.210024289705, "speed/FLOPS": 203636439497033.44, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.775744915008545, "optim/lr": 0.0029189337336174913, "optim/total_tokens": 2826698752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 4.225865840911865, "created_at": "2025-01-15T04:11:50.152176+00:00"} {"global_step": 10784, "acc_step": 0, "speed/wps": 12963.91988678034, "speed/FLOPS": 203616176114614.16, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7827219367027283, "optim/lr": 0.002918905945742143, "optim/total_tokens": 2826960896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473564, "loss/out": 4.231600284576416, "created_at": "2025-01-15T04:11:55.212482+00:00"} {"global_step": 10785, "acc_step": 0, "speed/wps": 12963.58717340702, "speed/FLOPS": 203610950393890.06, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0899417400360107, "optim/lr": 0.002918878153237364, "optim/total_tokens": 2827223040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439064, "loss/out": 4.2417707443237305, "created_at": "2025-01-15T04:12:00.272329+00:00"} {"global_step": 10786, "acc_step": 0, "speed/wps": 12955.120991414717, "speed/FLOPS": 203477977371947.75, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0022, "optim/grad_norm": 1.0423606634140015, "optim/lr": 0.002918850356103246, "optim/total_tokens": 2827485184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475134, "loss/out": 4.232024669647217, "created_at": "2025-01-15T04:12:05.334787+00:00"} {"global_step": 10787, "acc_step": 0, "speed/wps": 12954.008976041052, "speed/FLOPS": 203460511642435.1, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1651313304901123, "optim/lr": 0.00291882255433988, "optim/total_tokens": 2827747328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437432, "loss/out": 4.250051498413086, "created_at": "2025-01-15T04:12:10.401190+00:00"} {"global_step": 10788, "acc_step": 0, "speed/wps": 12961.161307606702, "speed/FLOPS": 203572848838007.16, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7138673067092896, "optim/lr": 0.0029187947479473557, "optim/total_tokens": 2828009472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293306, "loss/out": 4.190991401672363, "created_at": "2025-01-15T04:12:15.461599+00:00"} {"global_step": 10789, "acc_step": 0, "speed/wps": 12968.663475142586, "speed/FLOPS": 203690680688216.72, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8755661845207214, "optim/lr": 0.0029187669369257646, "optim/total_tokens": 2828271616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440624, "loss/out": 4.223990440368652, "created_at": "2025-01-15T04:12:20.521265+00:00"} {"global_step": 10790, "acc_step": 0, "speed/wps": 12972.533314152986, "speed/FLOPS": 203751461827591.1, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.7526825070381165, "optim/lr": 0.0029187391212751965, "optim/total_tokens": 2828533760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457665, "loss/out": 4.24552059173584, "created_at": "2025-01-15T04:12:25.578442+00:00"} {"global_step": 10791, "acc_step": 0, "speed/wps": 12959.755895601578, "speed/FLOPS": 203550774911228.2, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0042667388916016, "optim/lr": 0.0029187113009957436, "optim/total_tokens": 2828795904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448361, "loss/out": 4.158650875091553, "created_at": "2025-01-15T04:12:30.642409+00:00"} {"global_step": 10792, "acc_step": 0, "speed/wps": 12966.914220428125, "speed/FLOPS": 203663206239198.22, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0502705574035645, "optim/lr": 0.002918683476087495, "optim/total_tokens": 2829058048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459528, "loss/out": 4.158938884735107, "created_at": "2025-01-15T04:12:35.700269+00:00"} {"global_step": 10793, "acc_step": 0, "speed/wps": 12962.520918142056, "speed/FLOPS": 203594203389765.53, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6597921252250671, "optim/lr": 0.0029186556465505427, "optim/total_tokens": 2829320192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 4.13862419128418, "created_at": "2025-01-15T04:12:40.760853+00:00"} {"global_step": 10794, "acc_step": 0, "speed/wps": 12971.490298897577, "speed/FLOPS": 203735079839751.75, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6721914410591125, "optim/lr": 0.0029186278123849766, "optim/total_tokens": 2829582336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448083, "loss/out": 4.204519271850586, "created_at": "2025-01-15T04:12:45.822236+00:00"} {"global_step": 10795, "acc_step": 0, "speed/wps": 12950.412023057752, "speed/FLOPS": 203404016553101.0, "speed/curr_iter_time": 1.2787, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9381447434425354, "optim/lr": 0.0029185999735908885, "optim/total_tokens": 2829844480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313596, "loss/out": 4.221877574920654, "created_at": "2025-01-15T04:12:50.886369+00:00"} {"global_step": 10796, "acc_step": 0, "speed/wps": 12959.925541164812, "speed/FLOPS": 203553439427916.22, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3771275281906128, "optim/lr": 0.002918572130168369, "optim/total_tokens": 2830106624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332731, "loss/out": 4.174645900726318, "created_at": "2025-01-15T04:12:55.947367+00:00"} {"global_step": 10797, "acc_step": 0, "speed/wps": 12968.462210461928, "speed/FLOPS": 203687519549840.3, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1481629610061646, "optim/lr": 0.0029185442821175085, "optim/total_tokens": 2830368768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 361473, "loss/out": 4.2719645500183105, "created_at": "2025-01-15T04:13:01.006820+00:00"} {"global_step": 10798, "acc_step": 0, "speed/wps": 12960.18139033048, "speed/FLOPS": 203557457890636.62, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0869346857070923, "optim/lr": 0.002918516429438398, "optim/total_tokens": 2830630912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440628, "loss/out": 4.2014479637146, "created_at": "2025-01-15T04:13:06.067325+00:00"} {"global_step": 10799, "acc_step": 0, "speed/wps": 12966.564518337365, "speed/FLOPS": 203657713687321.8, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7053640484809875, "optim/lr": 0.0029184885721311284, "optim/total_tokens": 2830893056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 4.268490791320801, "created_at": "2025-01-15T04:13:11.130416+00:00"} {"global_step": 10800, "acc_step": 0, "speed/wps": 12958.062640667062, "speed/FLOPS": 203524180015705.5, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.001, "optim/grad_norm": 0.9906105399131775, "optim/lr": 0.00291846071019579, "optim/total_tokens": 2831155200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447839, "loss/out": 4.283977508544922, "created_at": "2025-01-15T04:13:16.192377+00:00"} {"global_step": 10801, "acc_step": 0, "speed/wps": 12963.083977833145, "speed/FLOPS": 203603047015939.2, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0746248960494995, "optim/lr": 0.0029184328436324754, "optim/total_tokens": 2831417344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 4.277174472808838, "created_at": "2025-01-15T04:13:21.252212+00:00"} {"global_step": 10802, "acc_step": 0, "speed/wps": 12955.120760061856, "speed/FLOPS": 203477973738233.16, "speed/curr_iter_time": 1.2776, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.116584300994873, "optim/lr": 0.002918404972441274, "optim/total_tokens": 2831679488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446218, "loss/out": 4.251398086547852, "created_at": "2025-01-15T04:13:26.315150+00:00"} {"global_step": 10803, "acc_step": 0, "speed/wps": 12966.132330506716, "speed/FLOPS": 203650925583553.16, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0542314052581787, "optim/lr": 0.0029183770966222772, "optim/total_tokens": 2831941632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330862, "loss/out": 4.2565107345581055, "created_at": "2025-01-15T04:13:31.377402+00:00"} {"global_step": 10804, "acc_step": 0, "speed/wps": 12965.728371043077, "speed/FLOPS": 203644580845078.22, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5577839016914368, "optim/lr": 0.002918349216175576, "optim/total_tokens": 2832203776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355618, "loss/out": 4.191771030426025, "created_at": "2025-01-15T04:13:36.438059+00:00"} {"global_step": 10805, "acc_step": 0, "speed/wps": 12958.781672441137, "speed/FLOPS": 203535473397770.56, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.6502096056938171, "optim/lr": 0.002918321331101261, "optim/total_tokens": 2832465920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424393, "loss/out": 4.2683563232421875, "created_at": "2025-01-15T04:13:41.502281+00:00"} {"global_step": 10806, "acc_step": 0, "speed/wps": 12958.303714562024, "speed/FLOPS": 203527966412496.12, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6988818049430847, "optim/lr": 0.0029182934413994237, "optim/total_tokens": 2832728064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 4.112422943115234, "created_at": "2025-01-15T04:13:46.564109+00:00"} {"global_step": 10807, "acc_step": 0, "speed/wps": 12953.310145179576, "speed/FLOPS": 203449535543461.38, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0414625406265259, "optim/lr": 0.002918265547070155, "optim/total_tokens": 2832990208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424963, "loss/out": 4.123038291931152, "created_at": "2025-01-15T04:13:51.627450+00:00"} {"global_step": 10808, "acc_step": 0, "speed/wps": 12964.659812054328, "speed/FLOPS": 203627797657806.0, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2601667642593384, "optim/lr": 0.002918237648113545, "optim/total_tokens": 2833252352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 4.075863361358643, "created_at": "2025-01-15T04:13:56.685994+00:00"} {"global_step": 10809, "acc_step": 0, "speed/wps": 12963.104691499673, "speed/FLOPS": 203603372352535.56, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.9433244466781616, "optim/lr": 0.0029182097445296863, "optim/total_tokens": 2833514496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 4.096036911010742, "created_at": "2025-01-15T04:14:01.751262+00:00"} {"global_step": 10810, "acc_step": 0, "speed/wps": 12964.978370026349, "speed/FLOPS": 203632801048501.1, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1187736988067627, "optim/lr": 0.0029181818363186694, "optim/total_tokens": 2833776640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471976, "loss/out": 4.219244003295898, "created_at": "2025-01-15T04:14:06.810907+00:00"} {"global_step": 10811, "acc_step": 0, "speed/wps": 12961.974377042312, "speed/FLOPS": 203585619210767.88, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.22128164768219, "optim/lr": 0.0029181539234805847, "optim/total_tokens": 2834038784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426015, "loss/out": 4.261694431304932, "created_at": "2025-01-15T04:14:11.871021+00:00"} {"global_step": 10812, "acc_step": 0, "speed/wps": 12955.596563031908, "speed/FLOPS": 203485446877699.62, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0257365703582764, "optim/lr": 0.002918126006015523, "optim/total_tokens": 2834300928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310665, "loss/out": 4.2126359939575195, "created_at": "2025-01-15T04:14:16.937186+00:00"} {"global_step": 10813, "acc_step": 0, "speed/wps": 12969.244950299806, "speed/FLOPS": 203699813554591.44, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7370638847351074, "optim/lr": 0.002918098083923577, "optim/total_tokens": 2834563072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445404, "loss/out": 4.208524703979492, "created_at": "2025-01-15T04:14:21.997344+00:00"} {"global_step": 10814, "acc_step": 0, "speed/wps": 12965.872273355282, "speed/FLOPS": 203646841028633.9, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1892966032028198, "optim/lr": 0.002918070157204837, "optim/total_tokens": 2834825216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454182, "loss/out": 4.24471378326416, "created_at": "2025-01-15T04:14:27.055464+00:00"} {"global_step": 10815, "acc_step": 0, "speed/wps": 12967.589320815889, "speed/FLOPS": 203673809618473.2, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0750551223754883, "optim/lr": 0.0029180422258593933, "optim/total_tokens": 2835087360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 4.1167683601379395, "created_at": "2025-01-15T04:14:32.118281+00:00"} {"global_step": 10816, "acc_step": 0, "speed/wps": 12968.802609786755, "speed/FLOPS": 203692865988994.4, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.140915870666504, "optim/lr": 0.0029180142898873377, "optim/total_tokens": 2835349504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438536, "loss/out": 4.1500396728515625, "created_at": "2025-01-15T04:14:37.175702+00:00"} {"global_step": 10817, "acc_step": 0, "speed/wps": 12962.117541819143, "speed/FLOPS": 203587867810317.03, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6914119720458984, "optim/lr": 0.0029179863492887616, "optim/total_tokens": 2835611648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461100, "loss/out": 4.164947986602783, "created_at": "2025-01-15T04:14:42.253388+00:00"} {"global_step": 10818, "acc_step": 0, "speed/wps": 12962.204622677098, "speed/FLOPS": 203589235534852.84, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4156519174575806, "optim/lr": 0.0029179584040637556, "optim/total_tokens": 2835873792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322689, "loss/out": 4.166193008422852, "created_at": "2025-01-15T04:14:47.317919+00:00"} {"global_step": 10819, "acc_step": 0, "speed/wps": 12960.601809379084, "speed/FLOPS": 203564061149511.28, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.940659761428833, "optim/lr": 0.0029179304542124116, "optim/total_tokens": 2836135936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440909, "loss/out": 4.1783623695373535, "created_at": "2025-01-15T04:14:52.392604+00:00"} {"global_step": 10820, "acc_step": 0, "speed/wps": 12967.66983421909, "speed/FLOPS": 203675074192107.5, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3465683460235596, "optim/lr": 0.0029179024997348204, "optim/total_tokens": 2836398080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461100, "loss/out": 4.163189888000488, "created_at": "2025-01-15T04:14:57.454447+00:00"} {"global_step": 10821, "acc_step": 0, "speed/wps": 12965.24249051605, "speed/FLOPS": 203636949423731.44, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7952226996421814, "optim/lr": 0.002917874540631073, "optim/total_tokens": 2836660224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479331, "loss/out": 4.225516319274902, "created_at": "2025-01-15T04:15:02.514331+00:00"} {"global_step": 10822, "acc_step": 0, "speed/wps": 12962.636185752837, "speed/FLOPS": 203596013825987.16, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0093181133270264, "optim/lr": 0.002917846576901261, "optim/total_tokens": 2836922368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447263, "loss/out": 4.236456871032715, "created_at": "2025-01-15T04:15:07.583656+00:00"} {"global_step": 10823, "acc_step": 0, "speed/wps": 12966.87873277927, "speed/FLOPS": 203662648856907.8, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.3714044094085693, "optim/lr": 0.002917818608545475, "optim/total_tokens": 2837184512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445404, "loss/out": 4.20258092880249, "created_at": "2025-01-15T04:15:12.641576+00:00"} {"global_step": 10824, "acc_step": 0, "speed/wps": 12964.254167333145, "speed/FLOPS": 203621426450045.56, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5935484766960144, "optim/lr": 0.002917790635563807, "optim/total_tokens": 2837446656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 4.222908973693848, "created_at": "2025-01-15T04:15:17.702117+00:00"} {"global_step": 10825, "acc_step": 0, "speed/wps": 12964.46712221275, "speed/FLOPS": 203624771199060.44, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0413943529129028, "optim/lr": 0.002917762657956348, "optim/total_tokens": 2837708800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 4.142933368682861, "created_at": "2025-01-15T04:15:22.764927+00:00"} {"global_step": 10826, "acc_step": 0, "speed/wps": 12959.803020514843, "speed/FLOPS": 203551515072747.44, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0501272678375244, "optim/lr": 0.002917734675723189, "optim/total_tokens": 2837970944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324181, "loss/out": 4.157168388366699, "created_at": "2025-01-15T04:15:27.826110+00:00"} {"global_step": 10827, "acc_step": 0, "speed/wps": 12959.494220611334, "speed/FLOPS": 203546664945927.28, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0363017320632935, "optim/lr": 0.0029177066888644216, "optim/total_tokens": 2838233088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304748, "loss/out": 4.148941516876221, "created_at": "2025-01-15T04:15:32.890933+00:00"} {"global_step": 10828, "acc_step": 0, "speed/wps": 12961.050862238015, "speed/FLOPS": 203571114141724.06, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0999641418457031, "optim/lr": 0.0029176786973801375, "optim/total_tokens": 2838495232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431320, "loss/out": 4.336122035980225, "created_at": "2025-01-15T04:15:37.950857+00:00"} {"global_step": 10829, "acc_step": 0, "speed/wps": 12968.842460403257, "speed/FLOPS": 203693491897687.06, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7832970023155212, "optim/lr": 0.002917650701270427, "optim/total_tokens": 2838757376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439302, "loss/out": 4.161571502685547, "created_at": "2025-01-15T04:15:43.017421+00:00"} {"global_step": 10830, "acc_step": 0, "speed/wps": 12964.121973018666, "speed/FLOPS": 203619350156682.62, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0355406999588013, "optim/lr": 0.0029176227005353822, "optim/total_tokens": 2839019520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 4.084837913513184, "created_at": "2025-01-15T04:15:48.082024+00:00"} {"global_step": 10831, "acc_step": 0, "speed/wps": 12964.368842376287, "speed/FLOPS": 203623227579178.12, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 50.63325119018555, "optim/lr": 0.0029175946951750945, "optim/total_tokens": 2839281664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420140, "loss/out": 4.197884559631348, "created_at": "2025-01-15T04:15:53.143054+00:00"} {"global_step": 10832, "acc_step": 0, "speed/wps": 12960.96113034276, "speed/FLOPS": 203569704779004.53, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.225994348526001, "optim/lr": 0.002917566685189655, "optim/total_tokens": 2839543808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 4.047482013702393, "created_at": "2025-01-15T04:15:58.211423+00:00"} {"global_step": 10833, "acc_step": 0, "speed/wps": 12961.30968274035, "speed/FLOPS": 203575179273369.88, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7493919134140015, "optim/lr": 0.002917538670579155, "optim/total_tokens": 2839805952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316975, "loss/out": 4.136632442474365, "created_at": "2025-01-15T04:16:03.271317+00:00"} {"global_step": 10834, "acc_step": 0, "speed/wps": 12962.463157358894, "speed/FLOPS": 203593296177294.34, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1233607530593872, "optim/lr": 0.0029175106513436863, "optim/total_tokens": 2840068096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 350526, "loss/out": 4.261462211608887, "created_at": "2025-01-15T04:16:08.330515+00:00"} {"global_step": 10835, "acc_step": 0, "speed/wps": 12959.034753422251, "speed/FLOPS": 203539448382348.66, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.826661229133606, "optim/lr": 0.0029174826274833395, "optim/total_tokens": 2840330240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420921, "loss/out": 4.252758979797363, "created_at": "2025-01-15T04:16:13.391353+00:00"} {"global_step": 10836, "acc_step": 0, "speed/wps": 12958.427323652599, "speed/FLOPS": 203529907863120.28, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8732558488845825, "optim/lr": 0.002917454598998207, "optim/total_tokens": 2840592384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450225, "loss/out": 4.247304439544678, "created_at": "2025-01-15T04:16:18.452614+00:00"} {"global_step": 10837, "acc_step": 0, "speed/wps": 12961.475620732217, "speed/FLOPS": 203577785557554.84, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1337766647338867, "optim/lr": 0.00291742656588838, "optim/total_tokens": 2840854528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438240, "loss/out": 4.22134256362915, "created_at": "2025-01-15T04:16:23.512340+00:00"} {"global_step": 10838, "acc_step": 0, "speed/wps": 12958.267290896822, "speed/FLOPS": 203527394328782.75, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.815396785736084, "optim/lr": 0.0029173985281539497, "optim/total_tokens": 2841116672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 4.261567115783691, "created_at": "2025-01-15T04:16:28.576767+00:00"} {"global_step": 10839, "acc_step": 0, "speed/wps": 12962.478491734066, "speed/FLOPS": 203593537024728.06, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.529702365398407, "optim/lr": 0.0029173704857950074, "optim/total_tokens": 2841378816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443835, "loss/out": 4.135065078735352, "created_at": "2025-01-15T04:16:33.641588+00:00"} {"global_step": 10840, "acc_step": 0, "speed/wps": 12958.58684252161, "speed/FLOPS": 203532413326160.03, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5690314173698425, "optim/lr": 0.002917342438811645, "optim/total_tokens": 2841640960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478811, "loss/out": 4.223031520843506, "created_at": "2025-01-15T04:16:38.704868+00:00"} {"global_step": 10841, "acc_step": 0, "speed/wps": 12958.050125247231, "speed/FLOPS": 203523983443839.06, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7329362034797668, "optim/lr": 0.002917314387203954, "optim/total_tokens": 2841903104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431320, "loss/out": 4.206932544708252, "created_at": "2025-01-15T04:16:43.766069+00:00"} {"global_step": 10842, "acc_step": 0, "speed/wps": 12959.24951296924, "speed/FLOPS": 203542821476144.66, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0624297857284546, "optim/lr": 0.0029172863309720257, "optim/total_tokens": 2842165248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309558, "loss/out": 4.21182918548584, "created_at": "2025-01-15T04:16:48.827400+00:00"} {"global_step": 10843, "acc_step": 0, "speed/wps": 12957.786016155378, "speed/FLOPS": 203519835247627.6, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5591498613357544, "optim/lr": 0.002917258270115952, "optim/total_tokens": 2842427392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429438, "loss/out": 4.255756378173828, "created_at": "2025-01-15T04:16:53.893154+00:00"} {"global_step": 10844, "acc_step": 0, "speed/wps": 12959.081605798701, "speed/FLOPS": 203540184263302.72, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.0481879711151123, "optim/lr": 0.002917230204635824, "optim/total_tokens": 2842689536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469069, "loss/out": 4.145689487457275, "created_at": "2025-01-15T04:16:58.958603+00:00"} {"global_step": 10845, "acc_step": 0, "speed/wps": 12960.128247611365, "speed/FLOPS": 203556623211207.16, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1415766477584839, "optim/lr": 0.0029172021345317332, "optim/total_tokens": 2842951680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444119, "loss/out": 4.1860456466674805, "created_at": "2025-01-15T04:17:04.020693+00:00"} {"global_step": 10846, "acc_step": 0, "speed/wps": 12966.767893668155, "speed/FLOPS": 203660907976358.84, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8944815397262573, "optim/lr": 0.002917174059803772, "optim/total_tokens": 2843213824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460578, "loss/out": 4.1938557624816895, "created_at": "2025-01-15T04:17:09.078383+00:00"} {"global_step": 10847, "acc_step": 0, "speed/wps": 12964.932167723357, "speed/FLOPS": 203632075377844.0, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0212074518203735, "optim/lr": 0.002917145980452031, "optim/total_tokens": 2843475968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 4.117624759674072, "created_at": "2025-01-15T04:17:14.137435+00:00"} {"global_step": 10848, "acc_step": 0, "speed/wps": 12964.065836391672, "speed/FLOPS": 203618468453815.88, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7058281898498535, "optim/lr": 0.0029171178964766025, "optim/total_tokens": 2843738112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445166, "loss/out": 4.131937503814697, "created_at": "2025-01-15T04:17:19.220261+00:00"} {"global_step": 10849, "acc_step": 0, "speed/wps": 12961.716814473823, "speed/FLOPS": 203581573836701.7, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.685754120349884, "optim/lr": 0.0029170898078775776, "optim/total_tokens": 2844000256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 343860, "loss/out": 4.20121955871582, "created_at": "2025-01-15T04:17:24.281130+00:00"} {"global_step": 10850, "acc_step": 0, "speed/wps": 12965.749941255048, "speed/FLOPS": 203644919634897.8, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.4469282329082489, "optim/lr": 0.0029170617146550486, "optim/total_tokens": 2844262400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 349184, "loss/out": 4.056356906890869, "created_at": "2025-01-15T04:17:29.344962+00:00"} {"global_step": 10851, "acc_step": 0, "speed/wps": 12962.040363623239, "speed/FLOPS": 203586655620696.4, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5444652438163757, "optim/lr": 0.0029170336168091067, "optim/total_tokens": 2844524544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443022, "loss/out": 4.175023078918457, "created_at": "2025-01-15T04:17:34.409182+00:00"} {"global_step": 10852, "acc_step": 0, "speed/wps": 12947.608171437978, "speed/FLOPS": 203359978210516.56, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.4693889319896698, "optim/lr": 0.002917005514339843, "optim/total_tokens": 2844786688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463770, "loss/out": 4.019641876220703, "created_at": "2025-01-15T04:17:39.476016+00:00"} {"global_step": 10853, "acc_step": 0, "speed/wps": 12955.160722217084, "speed/FLOPS": 203478601398794.78, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.44256189465522766, "optim/lr": 0.0029169774072473507, "optim/total_tokens": 2845048832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469875, "loss/out": 4.162297248840332, "created_at": "2025-01-15T04:17:44.542918+00:00"} {"global_step": 10854, "acc_step": 0, "speed/wps": 12953.827640993997, "speed/FLOPS": 203457663526340.62, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6441679000854492, "optim/lr": 0.002916949295531721, "optim/total_tokens": 2845310976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445166, "loss/out": 4.083741188049316, "created_at": "2025-01-15T04:17:49.606151+00:00"} {"global_step": 10855, "acc_step": 0, "speed/wps": 12964.636998627217, "speed/FLOPS": 203627439341584.28, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5631834864616394, "optim/lr": 0.0029169211791930444, "optim/total_tokens": 2845573120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331674, "loss/out": 4.1344895362854, "created_at": "2025-01-15T04:17:54.669554+00:00"} {"global_step": 10856, "acc_step": 0, "speed/wps": 12957.374898708305, "speed/FLOPS": 203513378083186.75, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1499656438827515, "optim/lr": 0.0029168930582314143, "optim/total_tokens": 2845835264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358303, "loss/out": 4.214978218078613, "created_at": "2025-01-15T04:17:59.732127+00:00"} {"global_step": 10857, "acc_step": 0, "speed/wps": 12961.198248880033, "speed/FLOPS": 203573429051467.44, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5074992775917053, "optim/lr": 0.002916864932646921, "optim/total_tokens": 2846097408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437723, "loss/out": 4.190692901611328, "created_at": "2025-01-15T04:18:04.821929+00:00"} {"global_step": 10858, "acc_step": 0, "speed/wps": 12968.102480643467, "speed/FLOPS": 203681869498720.8, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.0240058898925781, "optim/lr": 0.002916836802439657, "optim/total_tokens": 2846359552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454995, "loss/out": 4.255882740020752, "created_at": "2025-01-15T04:18:09.879114+00:00"} {"global_step": 10859, "acc_step": 0, "speed/wps": 12961.725700793319, "speed/FLOPS": 203581713408560.47, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.4453893005847931, "optim/lr": 0.0029168086676097144, "optim/total_tokens": 2846621696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479622, "loss/out": 4.086665630340576, "created_at": "2025-01-15T04:18:14.942754+00:00"} {"global_step": 10860, "acc_step": 0, "speed/wps": 12954.092990952471, "speed/FLOPS": 203461831212067.53, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.4837375581264496, "optim/lr": 0.002916780528157184, "optim/total_tokens": 2846883840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 318692, "loss/out": 4.180350303649902, "created_at": "2025-01-15T04:18:20.009642+00:00"} {"global_step": 10861, "acc_step": 0, "speed/wps": 12967.230093433911, "speed/FLOPS": 203668167458809.1, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6027737259864807, "optim/lr": 0.0029167523840821588, "optim/total_tokens": 2847145984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348945, "loss/out": 4.165285110473633, "created_at": "2025-01-15T04:18:25.067895+00:00"} {"global_step": 10862, "acc_step": 0, "speed/wps": 12961.927217235918, "speed/FLOPS": 203584878501204.1, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.45048123598098755, "optim/lr": 0.00291672423538473, "optim/total_tokens": 2847408128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 388560, "loss/out": 4.113968849182129, "created_at": "2025-01-15T04:18:30.128650+00:00"} {"global_step": 10863, "acc_step": 0, "speed/wps": 12963.25025260464, "speed/FLOPS": 203605658589710.97, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8740158081054688, "optim/lr": 0.0029166960820649897, "optim/total_tokens": 2847670272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348373, "loss/out": 4.090770721435547, "created_at": "2025-01-15T04:18:35.190979+00:00"} {"global_step": 10864, "acc_step": 0, "speed/wps": 12960.17603164182, "speed/FLOPS": 203557373725067.62, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.356925368309021, "optim/lr": 0.002916667924123029, "optim/total_tokens": 2847932416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443022, "loss/out": 4.22886848449707, "created_at": "2025-01-15T04:18:40.251893+00:00"} {"global_step": 10865, "acc_step": 0, "speed/wps": 12960.74849592098, "speed/FLOPS": 203566365063220.16, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8526561856269836, "optim/lr": 0.0029166397615589404, "optim/total_tokens": 2848194560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 4.136093616485596, "created_at": "2025-01-15T04:18:45.317273+00:00"} {"global_step": 10866, "acc_step": 0, "speed/wps": 12966.534016649772, "speed/FLOPS": 203657234616404.44, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.41810524463653564, "optim/lr": 0.0029166115943728157, "optim/total_tokens": 2848456704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451797, "loss/out": 4.121170997619629, "created_at": "2025-01-15T04:18:50.375410+00:00"} {"global_step": 10867, "acc_step": 0, "speed/wps": 12958.805594051106, "speed/FLOPS": 203535849119526.84, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.04464852809906, "optim/lr": 0.0029165834225647466, "optim/total_tokens": 2848718848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451275, "loss/out": 4.099258899688721, "created_at": "2025-01-15T04:18:55.437535+00:00"} {"global_step": 10868, "acc_step": 0, "speed/wps": 12960.976437292013, "speed/FLOPS": 203569945195676.5, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1547353267669678, "optim/lr": 0.002916555246134826, "optim/total_tokens": 2848980992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 4.096046447753906, "created_at": "2025-01-15T04:19:00.505819+00:00"} {"global_step": 10869, "acc_step": 0, "speed/wps": 12962.907837925923, "speed/FLOPS": 203600280496654.66, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.001, "optim/grad_norm": 1.0255012512207031, "optim/lr": 0.0029165270650831445, "optim/total_tokens": 2849243136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325819, "loss/out": 4.215068340301514, "created_at": "2025-01-15T04:19:05.565179+00:00"} {"global_step": 10870, "acc_step": 0, "speed/wps": 12963.73845805148, "speed/FLOPS": 203613326527118.22, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2424219846725464, "optim/lr": 0.002916498879409794, "optim/total_tokens": 2849505280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356961, "loss/out": 4.363546848297119, "created_at": "2025-01-15T04:19:10.624730+00:00"} {"global_step": 10871, "acc_step": 0, "speed/wps": 12960.686475579865, "speed/FLOPS": 203565390948538.1, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.683344304561615, "optim/lr": 0.002916470689114868, "optim/total_tokens": 2849767424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429454, "loss/out": 4.153740882873535, "created_at": "2025-01-15T04:19:15.688968+00:00"} {"global_step": 10872, "acc_step": 0, "speed/wps": 12964.894113535292, "speed/FLOPS": 203631477684528.53, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1757352352142334, "optim/lr": 0.0029164424941984573, "optim/total_tokens": 2850029568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482806, "loss/out": 4.2216877937316895, "created_at": "2025-01-15T04:19:20.748549+00:00"} {"global_step": 10873, "acc_step": 0, "speed/wps": 12966.767532873075, "speed/FLOPS": 203660902309576.34, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.016497015953064, "optim/lr": 0.002916414294660654, "optim/total_tokens": 2850291712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428623, "loss/out": 4.258398056030273, "created_at": "2025-01-15T04:19:25.810869+00:00"} {"global_step": 10874, "acc_step": 0, "speed/wps": 12961.144777923279, "speed/FLOPS": 203572589216614.8, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2629406452178955, "optim/lr": 0.0029163860905015502, "optim/total_tokens": 2850553856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 4.095219612121582, "created_at": "2025-01-15T04:19:30.870776+00:00"} {"global_step": 10875, "acc_step": 0, "speed/wps": 12962.17777793372, "speed/FLOPS": 203588813901269.56, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8489832282066345, "optim/lr": 0.0029163578817212378, "optim/total_tokens": 2850816000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443306, "loss/out": 4.20731258392334, "created_at": "2025-01-15T04:19:35.931011+00:00"} {"global_step": 10876, "acc_step": 0, "speed/wps": 12958.848621670422, "speed/FLOPS": 203536524927414.34, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4098016321659088, "optim/lr": 0.0029163296683198097, "optim/total_tokens": 2851078144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449420, "loss/out": 4.170604705810547, "created_at": "2025-01-15T04:19:40.997350+00:00"} {"global_step": 10877, "acc_step": 0, "speed/wps": 12963.37549164548, "speed/FLOPS": 203607625640943.84, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4891746938228607, "optim/lr": 0.0029163014502973563, "optim/total_tokens": 2851340288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295180, "loss/out": 4.1753034591674805, "created_at": "2025-01-15T04:19:46.056521+00:00"} {"global_step": 10878, "acc_step": 0, "speed/wps": 12961.351208619064, "speed/FLOPS": 203575831494357.62, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5569739937782288, "optim/lr": 0.0029162732276539712, "optim/total_tokens": 2851602432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 397433, "loss/out": 4.381202697753906, "created_at": "2025-01-15T04:19:51.121632+00:00"} {"global_step": 10879, "acc_step": 0, "speed/wps": 12962.549593374399, "speed/FLOPS": 203594653773693.38, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7287975549697876, "optim/lr": 0.0029162450003897456, "optim/total_tokens": 2851864576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473310, "loss/out": 4.082887649536133, "created_at": "2025-01-15T04:19:56.181500+00:00"} {"global_step": 10880, "acc_step": 0, "speed/wps": 12967.208962750123, "speed/FLOPS": 203667835572383.6, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3234608173370361, "optim/lr": 0.0029162167685047717, "optim/total_tokens": 2852126720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 4.279518127441406, "created_at": "2025-01-15T04:20:01.244175+00:00"} {"global_step": 10881, "acc_step": 0, "speed/wps": 12960.007818809669, "speed/FLOPS": 203554731711390.03, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5618685483932495, "optim/lr": 0.002916188531999142, "optim/total_tokens": 2852388864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455517, "loss/out": 4.077588081359863, "created_at": "2025-01-15T04:20:06.305721+00:00"} {"global_step": 10882, "acc_step": 0, "speed/wps": 12960.804272005065, "speed/FLOPS": 203567241103265.1, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5247994065284729, "optim/lr": 0.0029161602908729487, "optim/total_tokens": 2852651008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 4.282756805419922, "created_at": "2025-01-15T04:20:11.370491+00:00"} {"global_step": 10883, "acc_step": 0, "speed/wps": 12961.529297351286, "speed/FLOPS": 203578628622616.72, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0158727169036865, "optim/lr": 0.0029161320451262835, "optim/total_tokens": 2852913152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438769, "loss/out": 4.071196556091309, "created_at": "2025-01-15T04:20:16.435937+00:00"} {"global_step": 10884, "acc_step": 0, "speed/wps": 12965.748371684651, "speed/FLOPS": 203644894982637.9, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 3.4733128547668457, "optim/lr": 0.0029161037947592386, "optim/total_tokens": 2853175296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 280160, "loss/out": 4.166801929473877, "created_at": "2025-01-15T04:20:21.494075+00:00"} {"global_step": 10885, "acc_step": 0, "speed/wps": 12962.96327440145, "speed/FLOPS": 203601151202680.53, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7669926285743713, "optim/lr": 0.002916075539771906, "optim/total_tokens": 2853437440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 400357, "loss/out": 4.1315507888793945, "created_at": "2025-01-15T04:20:26.554328+00:00"} {"global_step": 10886, "acc_step": 0, "speed/wps": 12965.676286527632, "speed/FLOPS": 203643762786191.53, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1730455160140991, "optim/lr": 0.0029160472801643784, "optim/total_tokens": 2853699584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442208, "loss/out": 4.136398792266846, "created_at": "2025-01-15T04:20:31.615778+00:00"} {"global_step": 10887, "acc_step": 0, "speed/wps": 12963.461660414745, "speed/FLOPS": 203608979039873.47, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9096952676773071, "optim/lr": 0.0029160190159367477, "optim/total_tokens": 2853961728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482002, "loss/out": 4.218632221221924, "created_at": "2025-01-15T04:20:36.674701+00:00"} {"global_step": 10888, "acc_step": 0, "speed/wps": 12965.596385556599, "speed/FLOPS": 203642507831646.03, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.5765456557273865, "optim/lr": 0.0029159907470891065, "optim/total_tokens": 2854223872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440637, "loss/out": 3.8733932971954346, "created_at": "2025-01-15T04:20:41.736862+00:00"} {"global_step": 10889, "acc_step": 0, "speed/wps": 12961.642150293746, "speed/FLOPS": 203580401133154.25, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.02323579788208, "optim/lr": 0.0029159624736215467, "optim/total_tokens": 2854486016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438769, "loss/out": 4.1782002449035645, "created_at": "2025-01-15T04:20:46.796568+00:00"} {"global_step": 10890, "acc_step": 0, "speed/wps": 12969.378147878146, "speed/FLOPS": 203701905605591.9, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6962994337081909, "optim/lr": 0.0029159341955341603, "optim/total_tokens": 2854748160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458734, "loss/out": 4.218618392944336, "created_at": "2025-01-15T04:20:51.856476+00:00"} {"global_step": 10891, "acc_step": 0, "speed/wps": 12961.500507261344, "speed/FLOPS": 203578176434692.47, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0281449556350708, "optim/lr": 0.0029159059128270395, "optim/total_tokens": 2855010304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 4.108121395111084, "created_at": "2025-01-15T04:20:56.920506+00:00"} {"global_step": 10892, "acc_step": 0, "speed/wps": 12964.71128169562, "speed/FLOPS": 203628606059249.2, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7809324264526367, "optim/lr": 0.0029158776255002778, "optim/total_tokens": 2855272448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432365, "loss/out": 4.07706356048584, "created_at": "2025-01-15T04:21:01.980501+00:00"} {"global_step": 10893, "acc_step": 0, "speed/wps": 12959.80825955468, "speed/FLOPS": 203551597359067.28, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9709311127662659, "optim/lr": 0.0029158493335539664, "optim/total_tokens": 2855534592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313835, "loss/out": 4.06028938293457, "created_at": "2025-01-15T04:21:07.058427+00:00"} {"global_step": 10894, "acc_step": 0, "speed/wps": 12965.375734892366, "speed/FLOPS": 203639042209758.47, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.5204497575759888, "optim/lr": 0.0029158210369881967, "optim/total_tokens": 2855796736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 4.082136154174805, "created_at": "2025-01-15T04:21:12.116732+00:00"} {"global_step": 10895, "acc_step": 0, "speed/wps": 12952.463392468022, "speed/FLOPS": 203436236128566.25, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7477443218231201, "optim/lr": 0.0029157927358030634, "optim/total_tokens": 2856058880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 4.089508056640625, "created_at": "2025-01-15T04:21:17.181100+00:00"} {"global_step": 10896, "acc_step": 0, "speed/wps": 12949.654541808599, "speed/FLOPS": 203392119269194.75, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6547660231590271, "optim/lr": 0.002915764429998657, "optim/total_tokens": 2856321024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472789, "loss/out": 4.086793422698975, "created_at": "2025-01-15T04:21:22.247101+00:00"} {"global_step": 10897, "acc_step": 0, "speed/wps": 12960.10836932933, "speed/FLOPS": 203556310995472.2, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7180027365684509, "optim/lr": 0.0029157361195750707, "optim/total_tokens": 2856583168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448077, "loss/out": 4.227431297302246, "created_at": "2025-01-15T04:21:27.312838+00:00"} {"global_step": 10898, "acc_step": 0, "speed/wps": 12959.61322453385, "speed/FLOPS": 203548534066058.94, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.45042285323143005, "optim/lr": 0.0029157078045323965, "optim/total_tokens": 2856845312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315714, "loss/out": 4.065018653869629, "created_at": "2025-01-15T04:21:32.374340+00:00"} {"global_step": 10899, "acc_step": 0, "speed/wps": 12963.49038040818, "speed/FLOPS": 203609430126835.78, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7315306067466736, "optim/lr": 0.002915679484870727, "optim/total_tokens": 2857107456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 375098, "loss/out": 4.08681583404541, "created_at": "2025-01-15T04:21:37.434373+00:00"} {"global_step": 10900, "acc_step": 0, "speed/wps": 12960.199655257142, "speed/FLOPS": 203557744766408.5, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5806229114532471, "optim/lr": 0.0029156511605901545, "optim/total_tokens": 2857369600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437432, "loss/out": 4.088057041168213, "created_at": "2025-01-15T04:21:42.497689+00:00"} {"global_step": 10901, "acc_step": 0, "speed/wps": 12959.89864141586, "speed/FLOPS": 203553016930394.8, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7161477208137512, "optim/lr": 0.002915622831690771, "optim/total_tokens": 2857631744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 4.381673336029053, "created_at": "2025-01-15T04:21:47.558285+00:00"} {"global_step": 10902, "acc_step": 0, "speed/wps": 12961.130575893592, "speed/FLOPS": 203572366154222.94, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0402730703353882, "optim/lr": 0.0029155944981726694, "optim/total_tokens": 2857893888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 4.238512992858887, "created_at": "2025-01-15T04:21:52.622844+00:00"} {"global_step": 10903, "acc_step": 0, "speed/wps": 12965.528024017352, "speed/FLOPS": 203641434119733.56, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6040650010108948, "optim/lr": 0.002915566160035942, "optim/total_tokens": 2858156032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478613, "loss/out": 4.182727336883545, "created_at": "2025-01-15T04:21:57.681189+00:00"} {"global_step": 10904, "acc_step": 0, "speed/wps": 12960.998777099745, "speed/FLOPS": 203570296073055.06, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6777793765068054, "optim/lr": 0.0029155378172806817, "optim/total_tokens": 2858418176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434515, "loss/out": 4.129805564880371, "created_at": "2025-01-15T04:22:02.772111+00:00"} {"global_step": 10905, "acc_step": 0, "speed/wps": 12964.440117620923, "speed/FLOPS": 203624347054837.6, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5415771007537842, "optim/lr": 0.00291550946990698, "optim/total_tokens": 2858680320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 277512, "loss/out": 4.088846683502197, "created_at": "2025-01-15T04:22:07.831962+00:00"} {"global_step": 10906, "acc_step": 0, "speed/wps": 12958.748422409088, "speed/FLOPS": 203534951160326.88, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3448875844478607, "optim/lr": 0.00291548111791493, "optim/total_tokens": 2858942464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452330, "loss/out": 4.152294635772705, "created_at": "2025-01-15T04:22:12.896840+00:00"} {"global_step": 10907, "acc_step": 0, "speed/wps": 12960.194344292368, "speed/FLOPS": 203557661350408.7, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4756842851638794, "optim/lr": 0.0029154527613046246, "optim/total_tokens": 2859204608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 4.216087341308594, "created_at": "2025-01-15T04:22:17.957219+00:00"} {"global_step": 10908, "acc_step": 0, "speed/wps": 12959.940136771866, "speed/FLOPS": 203553668671981.47, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7133898735046387, "optim/lr": 0.0029154244000761565, "optim/total_tokens": 2859466752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455270, "loss/out": 4.223623275756836, "created_at": "2025-01-15T04:22:23.023922+00:00"} {"global_step": 10909, "acc_step": 0, "speed/wps": 12957.37557869299, "speed/FLOPS": 203513388763280.62, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3789829015731812, "optim/lr": 0.0029153960342296168, "optim/total_tokens": 2859728896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 4.162035942077637, "created_at": "2025-01-15T04:22:28.089580+00:00"} {"global_step": 10910, "acc_step": 0, "speed/wps": 12963.032172464476, "speed/FLOPS": 203602233341436.2, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6728804111480713, "optim/lr": 0.0029153676637650987, "optim/total_tokens": 2859991040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461385, "loss/out": 4.158708095550537, "created_at": "2025-01-15T04:22:33.153775+00:00"} {"global_step": 10911, "acc_step": 0, "speed/wps": 12967.13012294395, "speed/FLOPS": 203666597284890.4, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7291936874389648, "optim/lr": 0.0029153392886826952, "optim/total_tokens": 2860253184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323655, "loss/out": 4.14210319519043, "created_at": "2025-01-15T04:22:38.212469+00:00"} {"global_step": 10912, "acc_step": 0, "speed/wps": 12963.693441986708, "speed/FLOPS": 203612619487962.4, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1264170408248901, "optim/lr": 0.0029153109089824986, "optim/total_tokens": 2860515328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365501, "loss/out": 4.15303373336792, "created_at": "2025-01-15T04:22:43.276362+00:00"} {"global_step": 10913, "acc_step": 0, "speed/wps": 12968.000465349623, "speed/FLOPS": 203680267208347.28, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.9069828391075134, "optim/lr": 0.0029152825246646016, "optim/total_tokens": 2860777472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432656, "loss/out": 4.223642349243164, "created_at": "2025-01-15T04:22:48.334719+00:00"} {"global_step": 10914, "acc_step": 0, "speed/wps": 12960.927129087488, "speed/FLOPS": 203569170742567.88, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6336939334869385, "optim/lr": 0.0029152541357290967, "optim/total_tokens": 2861039616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487849, "loss/out": 4.084141731262207, "created_at": "2025-01-15T04:22:53.395086+00:00"} {"global_step": 10915, "acc_step": 0, "speed/wps": 12966.488164038172, "speed/FLOPS": 203656514438131.2, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.49003830552101135, "optim/lr": 0.0029152257421760767, "optim/total_tokens": 2861301760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463750, "loss/out": 4.107274055480957, "created_at": "2025-01-15T04:22:58.458384+00:00"} {"global_step": 10916, "acc_step": 0, "speed/wps": 12965.885561217223, "speed/FLOPS": 203647049732765.12, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3827778100967407, "optim/lr": 0.0029151973440056335, "optim/total_tokens": 2861563904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454173, "loss/out": 4.097044944763184, "created_at": "2025-01-15T04:23:03.522398+00:00"} {"global_step": 10917, "acc_step": 0, "speed/wps": 12964.368352736386, "speed/FLOPS": 203623219888710.66, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3780142366886139, "optim/lr": 0.002915168941217861, "optim/total_tokens": 2861826048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427066, "loss/out": 4.074191093444824, "created_at": "2025-01-15T04:23:08.582605+00:00"} {"global_step": 10918, "acc_step": 0, "speed/wps": 12962.31630545267, "speed/FLOPS": 203590989666311.44, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5839301943778992, "optim/lr": 0.0029151405338128512, "optim/total_tokens": 2862088192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495052, "loss/out": 4.196306228637695, "created_at": "2025-01-15T04:23:13.643322+00:00"} {"global_step": 10919, "acc_step": 0, "speed/wps": 12961.915558655033, "speed/FLOPS": 203584695387170.72, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.39321017265319824, "optim/lr": 0.002915112121790696, "optim/total_tokens": 2862350336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443835, "loss/out": 4.098835468292236, "created_at": "2025-01-15T04:23:18.704109+00:00"} {"global_step": 10920, "acc_step": 0, "speed/wps": 12958.670277272364, "speed/FLOPS": 203533723783571.97, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.9407681226730347, "optim/lr": 0.0029150837051514903, "optim/total_tokens": 2862612480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308499, "loss/out": 4.1161956787109375, "created_at": "2025-01-15T04:23:23.768796+00:00"} {"global_step": 10921, "acc_step": 0, "speed/wps": 12953.894543813956, "speed/FLOPS": 203458714327062.2, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7048449516296387, "optim/lr": 0.0029150552838953245, "optim/total_tokens": 2862874624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 411556, "loss/out": 4.083755016326904, "created_at": "2025-01-15T04:23:28.835657+00:00"} {"global_step": 10922, "acc_step": 0, "speed/wps": 12961.782767394428, "speed/FLOPS": 203582609717948.62, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5755632519721985, "optim/lr": 0.002915026858022292, "optim/total_tokens": 2863136768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457902, "loss/out": 4.147747993469238, "created_at": "2025-01-15T04:23:33.895354+00:00"} {"global_step": 10923, "acc_step": 0, "speed/wps": 12962.446999236086, "speed/FLOPS": 203593042391772.06, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9333162903785706, "optim/lr": 0.0029149984275324862, "optim/total_tokens": 2863398912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 4.237165927886963, "created_at": "2025-01-15T04:23:38.959759+00:00"} {"global_step": 10924, "acc_step": 0, "speed/wps": 12965.435201018081, "speed/FLOPS": 203639976206977.84, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9141840934753418, "optim/lr": 0.0029149699924259993, "optim/total_tokens": 2863661056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429217, "loss/out": 4.107105255126953, "created_at": "2025-01-15T04:23:44.019176+00:00"} {"global_step": 10925, "acc_step": 0, "speed/wps": 12959.35038585193, "speed/FLOPS": 203544405823381.2, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4958142638206482, "optim/lr": 0.002914941552702924, "optim/total_tokens": 2863923200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 4.1873345375061035, "created_at": "2025-01-15T04:23:49.080092+00:00"} {"global_step": 10926, "acc_step": 0, "speed/wps": 12963.892225294123, "speed/FLOPS": 203615741652962.25, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8442298769950867, "optim/lr": 0.0029149131083633538, "optim/total_tokens": 2864185344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459528, "loss/out": 4.174558639526367, "created_at": "2025-01-15T04:23:54.148685+00:00"} {"global_step": 10927, "acc_step": 0, "speed/wps": 12967.452479956251, "speed/FLOPS": 203671660344733.53, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.739312469959259, "optim/lr": 0.002914884659407381, "optim/total_tokens": 2864447488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316527, "loss/out": 4.052023887634277, "created_at": "2025-01-15T04:23:59.206849+00:00"} {"global_step": 10928, "acc_step": 0, "speed/wps": 12962.608751813497, "speed/FLOPS": 203595582938271.62, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0034788846969604, "optim/lr": 0.002914856205835098, "optim/total_tokens": 2864709632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352687, "loss/out": 4.080249786376953, "created_at": "2025-01-15T04:24:04.271558+00:00"} {"global_step": 10929, "acc_step": 0, "speed/wps": 12963.546438017913, "speed/FLOPS": 203610310588627.84, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.4059244692325592, "optim/lr": 0.0029148277476465983, "optim/total_tokens": 2864971776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 404376, "loss/out": 4.130008697509766, "created_at": "2025-01-15T04:24:09.335603+00:00"} {"global_step": 10930, "acc_step": 0, "speed/wps": 12961.03722619473, "speed/FLOPS": 203570899968926.4, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.40745601058006287, "optim/lr": 0.0029147992848419747, "optim/total_tokens": 2865233920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433179, "loss/out": 4.087079048156738, "created_at": "2025-01-15T04:24:14.400780+00:00"} {"global_step": 10931, "acc_step": 0, "speed/wps": 12958.312059294549, "speed/FLOPS": 203528097477987.28, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3874635100364685, "optim/lr": 0.0029147708174213196, "optim/total_tokens": 2865496064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 4.154355049133301, "created_at": "2025-01-15T04:24:19.467078+00:00"} {"global_step": 10932, "acc_step": 0, "speed/wps": 12966.591259399138, "speed/FLOPS": 203658133692443.03, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.3659382462501526, "optim/lr": 0.0029147423453847257, "optim/total_tokens": 2865758208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462963, "loss/out": 4.129593849182129, "created_at": "2025-01-15T04:24:24.525030+00:00"} {"global_step": 10933, "acc_step": 0, "speed/wps": 12963.284597994038, "speed/FLOPS": 203606198031247.0, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3025131225585938, "optim/lr": 0.0029147138687322873, "optim/total_tokens": 2866020352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 4.111689567565918, "created_at": "2025-01-15T04:24:29.589419+00:00"} {"global_step": 10934, "acc_step": 0, "speed/wps": 12966.68016848819, "speed/FLOPS": 203659530131860.47, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8912249207496643, "optim/lr": 0.0029146853874640957, "optim/total_tokens": 2866282496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332206, "loss/out": 4.15235710144043, "created_at": "2025-01-15T04:24:34.647268+00:00"} {"global_step": 10935, "acc_step": 0, "speed/wps": 12963.685610957229, "speed/FLOPS": 203612496490883.47, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.7291898727416992, "optim/lr": 0.002914656901580245, "optim/total_tokens": 2866544640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331010, "loss/out": 4.154811382293701, "created_at": "2025-01-15T04:24:39.706285+00:00"} {"global_step": 10936, "acc_step": 0, "speed/wps": 12958.593801143508, "speed/FLOPS": 203532522620879.06, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7893363833427429, "optim/lr": 0.0029146284110808276, "optim/total_tokens": 2866806784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440108, "loss/out": 4.037137508392334, "created_at": "2025-01-15T04:24:44.774037+00:00"} {"global_step": 10937, "acc_step": 0, "speed/wps": 12960.510037985774, "speed/FLOPS": 203562619753672.22, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2889326810836792, "optim/lr": 0.002914599915965936, "optim/total_tokens": 2867068928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486044, "loss/out": 4.156072616577148, "created_at": "2025-01-15T04:24:49.843233+00:00"} {"global_step": 10938, "acc_step": 0, "speed/wps": 12964.923045303638, "speed/FLOPS": 203631932097706.8, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5017609596252441, "optim/lr": 0.0029145714162356635, "optim/total_tokens": 2867331072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433708, "loss/out": 4.187744140625, "created_at": "2025-01-15T04:24:54.906131+00:00"} {"global_step": 10939, "acc_step": 0, "speed/wps": 12958.62165241042, "speed/FLOPS": 203532960063297.6, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5515334010124207, "optim/lr": 0.002914542911890104, "optim/total_tokens": 2867593216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 4.157537460327148, "created_at": "2025-01-15T04:24:59.967704+00:00"} {"global_step": 10940, "acc_step": 0, "speed/wps": 12969.562677082075, "speed/FLOPS": 203704803890309.3, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.45710277557373047, "optim/lr": 0.002914514402929349, "optim/total_tokens": 2867855360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426829, "loss/out": 3.9578537940979004, "created_at": "2025-01-15T04:25:05.030322+00:00"} {"global_step": 10941, "acc_step": 0, "speed/wps": 12964.425347069204, "speed/FLOPS": 203624115063025.97, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.40902215242385864, "optim/lr": 0.002914485889353493, "optim/total_tokens": 2868117504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325094, "loss/out": 4.082169532775879, "created_at": "2025-01-15T04:25:10.089318+00:00"} {"global_step": 10942, "acc_step": 0, "speed/wps": 12959.416603478761, "speed/FLOPS": 203545445862203.12, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5516433715820312, "optim/lr": 0.0029144573711626276, "optim/total_tokens": 2868379648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 358542, "loss/out": 4.17333984375, "created_at": "2025-01-15T04:25:15.153020+00:00"} {"global_step": 10943, "acc_step": 0, "speed/wps": 12965.990367136814, "speed/FLOPS": 203648695853750.72, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.44816747307777405, "optim/lr": 0.002914428848356847, "optim/total_tokens": 2868641792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 389608, "loss/out": 4.038091659545898, "created_at": "2025-01-15T04:25:20.215538+00:00"} {"global_step": 10944, "acc_step": 0, "speed/wps": 12963.553096279406, "speed/FLOPS": 203610415165774.0, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4236953556537628, "optim/lr": 0.002914400320936244, "optim/total_tokens": 2868903936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463248, "loss/out": 4.155824661254883, "created_at": "2025-01-15T04:25:25.282319+00:00"} {"global_step": 10945, "acc_step": 0, "speed/wps": 12956.37168647532, "speed/FLOPS": 203497621256509.66, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3206177055835724, "optim/lr": 0.002914371788900911, "optim/total_tokens": 2869166080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440919, "loss/out": 4.19564151763916, "created_at": "2025-01-15T04:25:30.348370+00:00"} {"global_step": 10946, "acc_step": 0, "speed/wps": 12959.399426687853, "speed/FLOPS": 203545176077098.5, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.46306276321411133, "optim/lr": 0.0029143432522509413, "optim/total_tokens": 2869428224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487565, "loss/out": 4.06984806060791, "created_at": "2025-01-15T04:25:35.410303+00:00"} {"global_step": 10947, "acc_step": 0, "speed/wps": 12965.25012037755, "speed/FLOPS": 203637069261190.84, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6457648873329163, "optim/lr": 0.0029143147109864293, "optim/total_tokens": 2869690368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452610, "loss/out": 4.016430377960205, "created_at": "2025-01-15T04:25:40.471750+00:00"} {"global_step": 10948, "acc_step": 0, "speed/wps": 12964.57830655238, "speed/FLOPS": 203626517501897.6, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0397372245788574, "optim/lr": 0.002914286165107466, "optim/total_tokens": 2869952512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309387, "loss/out": 4.0427045822143555, "created_at": "2025-01-15T04:25:45.530463+00:00"} {"global_step": 10949, "acc_step": 0, "speed/wps": 12967.753075308643, "speed/FLOPS": 203676381607804.16, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6960597634315491, "optim/lr": 0.0029142576146141464, "optim/total_tokens": 2870214656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 413950, "loss/out": 4.139379501342773, "created_at": "2025-01-15T04:25:50.588259+00:00"} {"global_step": 10950, "acc_step": 0, "speed/wps": 12966.516022967324, "speed/FLOPS": 203656952000895.7, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.44362589716911316, "optim/lr": 0.0029142290595065624, "optim/total_tokens": 2870476800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433708, "loss/out": 4.182387828826904, "created_at": "2025-01-15T04:25:55.652476+00:00"} {"global_step": 10951, "acc_step": 0, "speed/wps": 12962.36700912244, "speed/FLOPS": 203591786037119.28, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1870123147964478, "optim/lr": 0.0029142004997848078, "optim/total_tokens": 2870738944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489708, "loss/out": 4.166157245635986, "created_at": "2025-01-15T04:26:00.713071+00:00"} {"global_step": 10952, "acc_step": 0, "speed/wps": 12965.998208118293, "speed/FLOPS": 203648819007139.5, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.8796250820159912, "optim/lr": 0.0029141719354489753, "optim/total_tokens": 2871001088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442256, "loss/out": 4.180898189544678, "created_at": "2025-01-15T04:26:05.775941+00:00"} {"global_step": 10953, "acc_step": 0, "speed/wps": 12965.275663996135, "speed/FLOPS": 203637470458821.34, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0883359909057617, "optim/lr": 0.002914143366499159, "optim/total_tokens": 2871263232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468777, "loss/out": 4.077607154846191, "created_at": "2025-01-15T04:26:10.834376+00:00"} {"global_step": 10954, "acc_step": 0, "speed/wps": 12957.254933825527, "speed/FLOPS": 203511493869853.7, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8130970001220703, "optim/lr": 0.0029141147929354514, "optim/total_tokens": 2871525376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466369, "loss/out": 4.162784576416016, "created_at": "2025-01-15T04:26:15.896062+00:00"} {"global_step": 10955, "acc_step": 0, "speed/wps": 12952.350620627372, "speed/FLOPS": 203434464891844.12, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7300714254379272, "optim/lr": 0.0029140862147579456, "optim/total_tokens": 2871787520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450994, "loss/out": 4.216254711151123, "created_at": "2025-01-15T04:26:20.965539+00:00"} {"global_step": 10956, "acc_step": 0, "speed/wps": 12964.441739421909, "speed/FLOPS": 203624372527450.66, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6236068606376648, "optim/lr": 0.0029140576319667353, "optim/total_tokens": 2872049664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 336239, "loss/out": 4.111250400543213, "created_at": "2025-01-15T04:26:26.025617+00:00"} {"global_step": 10957, "acc_step": 0, "speed/wps": 12966.432568390645, "speed/FLOPS": 203655641232090.34, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9975705742835999, "optim/lr": 0.0029140290445619135, "optim/total_tokens": 2872311808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313309, "loss/out": 4.017452716827393, "created_at": "2025-01-15T04:26:31.083474+00:00"} {"global_step": 10958, "acc_step": 0, "speed/wps": 12960.478877304144, "speed/FLOPS": 203562130332347.94, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.89366215467453, "optim/lr": 0.0029140004525435733, "optim/total_tokens": 2872573952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 378157, "loss/out": 4.115205764770508, "created_at": "2025-01-15T04:26:36.144316+00:00"} {"global_step": 10959, "acc_step": 0, "speed/wps": 12963.196838745367, "speed/FLOPS": 203604819651652.78, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3411259949207306, "optim/lr": 0.0029139718559118085, "optim/total_tokens": 2872836096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438769, "loss/out": 4.074761390686035, "created_at": "2025-01-15T04:26:41.204996+00:00"} {"global_step": 10960, "acc_step": 0, "speed/wps": 12971.730100796844, "speed/FLOPS": 203738846258102.06, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4194749593734741, "optim/lr": 0.002913943254666712, "optim/total_tokens": 2873098240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 4.070496559143066, "created_at": "2025-01-15T04:26:46.266455+00:00"} {"global_step": 10961, "acc_step": 0, "speed/wps": 12961.86120793135, "speed/FLOPS": 203583841734369.56, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5365265607833862, "optim/lr": 0.0029139146488083778, "optim/total_tokens": 2873360384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434515, "loss/out": 4.1330156326293945, "created_at": "2025-01-15T04:26:51.331819+00:00"} {"global_step": 10962, "acc_step": 0, "speed/wps": 12967.421042142138, "speed/FLOPS": 203671166570662.8, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5445302724838257, "optim/lr": 0.002913886038336898, "optim/total_tokens": 2873622528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498178, "loss/out": 4.083285331726074, "created_at": "2025-01-15T04:26:56.390724+00:00"} {"global_step": 10963, "acc_step": 0, "speed/wps": 12965.076998306693, "speed/FLOPS": 203634350141173.38, "speed/curr_iter_time": 1.2729, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3642837405204773, "optim/lr": 0.0029138574232523665, "optim/total_tokens": 2873884672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 4.14336633682251, "created_at": "2025-01-15T04:27:01.456805+00:00"} {"global_step": 10964, "acc_step": 0, "speed/wps": 12962.700457415754, "speed/FLOPS": 203597023300771.78, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6458616256713867, "optim/lr": 0.0029138288035548774, "optim/total_tokens": 2874146816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 348373, "loss/out": 4.0916852951049805, "created_at": "2025-01-15T04:27:06.516213+00:00"} {"global_step": 10965, "acc_step": 0, "speed/wps": 12964.139499095483, "speed/FLOPS": 203619625427802.4, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1453477144241333, "optim/lr": 0.0029138001792445227, "optim/total_tokens": 2874408960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454196, "loss/out": 4.0590996742248535, "created_at": "2025-01-15T04:27:11.575620+00:00"} {"global_step": 10966, "acc_step": 0, "speed/wps": 12961.232199036514, "speed/FLOPS": 203573962285327.38, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6954178214073181, "optim/lr": 0.002913771550321397, "optim/total_tokens": 2874671104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462963, "loss/out": 4.032994270324707, "created_at": "2025-01-15T04:27:16.635435+00:00"} {"global_step": 10967, "acc_step": 0, "speed/wps": 12959.611837840714, "speed/FLOPS": 203548512286137.72, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.260552167892456, "optim/lr": 0.002913742916785593, "optim/total_tokens": 2874933248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486809, "loss/out": 4.056330680847168, "created_at": "2025-01-15T04:27:21.698654+00:00"} {"global_step": 10968, "acc_step": 0, "speed/wps": 12963.706856736533, "speed/FLOPS": 203612830185042.2, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6307398676872253, "optim/lr": 0.0029137142786372044, "optim/total_tokens": 2875195392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443016, "loss/out": 4.170615196228027, "created_at": "2025-01-15T04:27:26.758151+00:00"} {"global_step": 10969, "acc_step": 0, "speed/wps": 12962.93388572709, "speed/FLOPS": 203600689613165.44, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1523041725158691, "optim/lr": 0.0029136856358763256, "optim/total_tokens": 2875457536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440390, "loss/out": 4.104063034057617, "created_at": "2025-01-15T04:27:31.817342+00:00"} {"global_step": 10970, "acc_step": 0, "speed/wps": 12959.2397541632, "speed/FLOPS": 203542668200685.8, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9636353850364685, "optim/lr": 0.002913656988503048, "optim/total_tokens": 2875719680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468255, "loss/out": 4.1884026527404785, "created_at": "2025-01-15T04:27:36.877985+00:00"} {"global_step": 10971, "acc_step": 0, "speed/wps": 12969.558278744691, "speed/FLOPS": 203704734808376.75, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1562632322311401, "optim/lr": 0.002913628336517466, "optim/total_tokens": 2875981824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 4.086490631103516, "created_at": "2025-01-15T04:27:41.940318+00:00"} {"global_step": 10972, "acc_step": 0, "speed/wps": 12960.506211520476, "speed/FLOPS": 203562559653776.6, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6108744740486145, "optim/lr": 0.0029135996799196737, "optim/total_tokens": 2876243968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 322976, "loss/out": 4.14570426940918, "created_at": "2025-01-15T04:27:47.005652+00:00"} {"global_step": 10973, "acc_step": 0, "speed/wps": 12957.216201778538, "speed/FLOPS": 203510885529832.3, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7955120205879211, "optim/lr": 0.0029135710187097636, "optim/total_tokens": 2876506112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372174, "loss/out": 4.041593074798584, "created_at": "2025-01-15T04:27:52.067100+00:00"} {"global_step": 10974, "acc_step": 0, "speed/wps": 12961.591526305921, "speed/FLOPS": 203579606013860.8, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8694877028465271, "optim/lr": 0.0029135423528878304, "optim/total_tokens": 2876768256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439583, "loss/out": 4.064530372619629, "created_at": "2025-01-15T04:27:57.131326+00:00"} {"global_step": 10975, "acc_step": 0, "speed/wps": 12954.16883209238, "speed/FLOPS": 203463022401386.44, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5288857817649841, "optim/lr": 0.0029135136824539666, "optim/total_tokens": 2877030400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 4.109708786010742, "created_at": "2025-01-15T04:28:02.198363+00:00"} {"global_step": 10976, "acc_step": 0, "speed/wps": 12954.534385821567, "speed/FLOPS": 203468763925026.78, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0015, "optim/grad_norm": 0.8056230545043945, "optim/lr": 0.0029134850074082662, "optim/total_tokens": 2877292544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 4.085905075073242, "created_at": "2025-01-15T04:28:07.264005+00:00"} {"global_step": 10977, "acc_step": 0, "speed/wps": 12955.515681884859, "speed/FLOPS": 203484176528142.88, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6477541923522949, "optim/lr": 0.0029134563277508226, "optim/total_tokens": 2877554688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445170, "loss/out": 4.144643783569336, "created_at": "2025-01-15T04:28:12.329510+00:00"} {"global_step": 10978, "acc_step": 0, "speed/wps": 12960.437235275836, "speed/FLOPS": 203561476287071.25, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9133967161178589, "optim/lr": 0.0029134276434817294, "optim/total_tokens": 2877816832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304215, "loss/out": 4.176438331604004, "created_at": "2025-01-15T04:28:17.392211+00:00"} {"global_step": 10979, "acc_step": 0, "speed/wps": 12960.468383821977, "speed/FLOPS": 203561965517790.97, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0390686988830566, "optim/lr": 0.00291339895460108, "optim/total_tokens": 2878078976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 377340, "loss/out": 4.190551280975342, "created_at": "2025-01-15T04:28:22.454135+00:00"} {"global_step": 10980, "acc_step": 0, "speed/wps": 12964.006509409235, "speed/FLOPS": 203617536642032.9, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.9002231359481812, "optim/lr": 0.002913370261108968, "optim/total_tokens": 2878341120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450462, "loss/out": 4.144235610961914, "created_at": "2025-01-15T04:28:27.514341+00:00"} {"global_step": 10981, "acc_step": 0, "speed/wps": 12962.288204305429, "speed/FLOPS": 203590548299178.97, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1229610443115234, "optim/lr": 0.002913341563005488, "optim/total_tokens": 2878603264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455232, "loss/out": 4.166469573974609, "created_at": "2025-01-15T04:28:32.578044+00:00"} {"global_step": 10982, "acc_step": 0, "speed/wps": 12951.240387018775, "speed/FLOPS": 203417027147400.25, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5412350296974182, "optim/lr": 0.0029133128602907322, "optim/total_tokens": 2878865408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 4.104530334472656, "created_at": "2025-01-15T04:28:37.641893+00:00"} {"global_step": 10983, "acc_step": 0, "speed/wps": 12962.698836050395, "speed/FLOPS": 203596997835000.8, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0473111867904663, "optim/lr": 0.0029132841529647953, "optim/total_tokens": 2879127552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 4.110114097595215, "created_at": "2025-01-15T04:28:42.701274+00:00"} {"global_step": 10984, "acc_step": 0, "speed/wps": 12962.480497478751, "speed/FLOPS": 203593568527704.5, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6168608665466309, "optim/lr": 0.0029132554410277707, "optim/total_tokens": 2879389696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429217, "loss/out": 4.168568134307861, "created_at": "2025-01-15T04:28:47.776047+00:00"} {"global_step": 10985, "acc_step": 0, "speed/wps": 12966.449171295491, "speed/FLOPS": 203655902003525.72, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7187342047691345, "optim/lr": 0.002913226724479751, "optim/total_tokens": 2879651840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441157, "loss/out": 3.920039176940918, "created_at": "2025-01-15T04:28:52.838582+00:00"} {"global_step": 10986, "acc_step": 0, "speed/wps": 12966.808173817497, "speed/FLOPS": 203661540631455.0, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7986570596694946, "optim/lr": 0.002913198003320832, "optim/total_tokens": 2879913984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 283388, "loss/out": 4.121416091918945, "created_at": "2025-01-15T04:28:57.897590+00:00"} {"global_step": 10987, "acc_step": 0, "speed/wps": 12957.56157507821, "speed/FLOPS": 203516310092096.3, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5692523121833801, "optim/lr": 0.0029131692775511055, "optim/total_tokens": 2880176128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449135, "loss/out": 4.013664722442627, "created_at": "2025-01-15T04:29:02.965204+00:00"} {"global_step": 10988, "acc_step": 0, "speed/wps": 12968.224524943182, "speed/FLOPS": 203683786372154.28, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7311814427375793, "optim/lr": 0.002913140547170666, "optim/total_tokens": 2880438272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465870, "loss/out": 4.06338357925415, "created_at": "2025-01-15T04:29:08.028338+00:00"} {"global_step": 10989, "acc_step": 0, "speed/wps": 12960.059919852687, "speed/FLOPS": 203555550029864.97, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8652700781822205, "optim/lr": 0.0029131118121796078, "optim/total_tokens": 2880700416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444641, "loss/out": 4.134487152099609, "created_at": "2025-01-15T04:29:13.094154+00:00"} {"global_step": 10990, "acc_step": 0, "speed/wps": 12960.710974749187, "speed/FLOPS": 203565775741657.66, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5484272837638855, "optim/lr": 0.0029130830725780234, "optim/total_tokens": 2880962560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476467, "loss/out": 4.070272922515869, "created_at": "2025-01-15T04:29:18.154203+00:00"} {"global_step": 10991, "acc_step": 0, "speed/wps": 12959.823998180926, "speed/FLOPS": 203551844555819.6, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6828428506851196, "optim/lr": 0.002913054328366008, "optim/total_tokens": 2881224704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438240, "loss/out": 4.012066841125488, "created_at": "2025-01-15T04:29:23.215770+00:00"} {"global_step": 10992, "acc_step": 0, "speed/wps": 12952.030043864062, "speed/FLOPS": 203429429793257.7, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.49635791778564453, "optim/lr": 0.002913025579543654, "optim/total_tokens": 2881486848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 316000, "loss/out": 4.1835808753967285, "created_at": "2025-01-15T04:29:28.284799+00:00"} {"global_step": 10993, "acc_step": 0, "speed/wps": 12960.49123748305, "speed/FLOPS": 203562324465942.3, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6142615675926208, "optim/lr": 0.002912996826111056, "optim/total_tokens": 2881748992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449942, "loss/out": 4.04374885559082, "created_at": "2025-01-15T04:29:33.345120+00:00"} {"global_step": 10994, "acc_step": 0, "speed/wps": 12956.2436056269, "speed/FLOPS": 203495609570783.7, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5549435615539551, "optim/lr": 0.0029129680680683077, "optim/total_tokens": 2882011136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 4.225530624389648, "created_at": "2025-01-15T04:29:38.411777+00:00"} {"global_step": 10995, "acc_step": 0, "speed/wps": 12966.951220785955, "speed/FLOPS": 203663787380661.75, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5610783100128174, "optim/lr": 0.002912939305415502, "optim/total_tokens": 2882273280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459765, "loss/out": 3.993777275085449, "created_at": "2025-01-15T04:29:43.470156+00:00"} {"global_step": 10996, "acc_step": 0, "speed/wps": 12967.797371579736, "speed/FLOPS": 203677077341611.12, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6627143621444702, "optim/lr": 0.002912910538152735, "optim/total_tokens": 2882535424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462672, "loss/out": 4.2381486892700195, "created_at": "2025-01-15T04:29:48.533306+00:00"} {"global_step": 10997, "acc_step": 0, "speed/wps": 12962.678415000344, "speed/FLOPS": 203596677094348.3, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1303256750106812, "optim/lr": 0.002912881766280099, "optim/total_tokens": 2882797568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452847, "loss/out": 4.1966023445129395, "created_at": "2025-01-15T04:29:53.592640+00:00"} {"global_step": 10998, "acc_step": 0, "speed/wps": 12959.37351958136, "speed/FLOPS": 203544769170390.97, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9920458793640137, "optim/lr": 0.0029128529897976866, "optim/total_tokens": 2883059712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 4.124749660491943, "created_at": "2025-01-15T04:29:58.656592+00:00"} {"global_step": 10999, "acc_step": 0, "speed/wps": 12958.194799726427, "speed/FLOPS": 203526255755338.4, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.816227376461029, "optim/lr": 0.0029128242087055945, "optim/total_tokens": 2883321856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306381, "loss/out": 4.106748580932617, "created_at": "2025-01-15T04:30:03.723308+00:00"} {"global_step": 11000, "acc_step": 0, "speed/wps": 12955.728302070598, "speed/FLOPS": 203487516020330.66, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0270930528640747, "optim/lr": 0.0029127954230039146, "optim/total_tokens": 2883584000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431014, "loss/out": 4.102629661560059, "created_at": "2025-01-15T04:30:08.791491+00:00"} {"global_step": 11001, "acc_step": 0, "speed/wps": 12294.800552866076, "speed/FLOPS": 193106737509172.66, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.2754, "optim/grad_norm": 0.5926496982574463, "optim/lr": 0.0029127666326927417, "optim/total_tokens": 2883846144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313596, "loss/out": 4.094992637634277, "created_at": "2025-01-15T04:30:14.125666+00:00"} {"global_step": 11002, "acc_step": 0, "speed/wps": 12968.852824041607, "speed/FLOPS": 203693654672868.47, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5660413503646851, "optim/lr": 0.0029127378377721692, "optim/total_tokens": 2884108288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 416106, "loss/out": 4.054675579071045, "created_at": "2025-01-15T04:30:19.183716+00:00"} {"global_step": 11003, "acc_step": 0, "speed/wps": 12963.12896817304, "speed/FLOPS": 203603753651050.44, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7444628477096558, "optim/lr": 0.0029127090382422916, "optim/total_tokens": 2884370432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451797, "loss/out": 4.2084527015686035, "created_at": "2025-01-15T04:30:24.244604+00:00"} {"global_step": 11004, "acc_step": 0, "speed/wps": 12965.897613967043, "speed/FLOPS": 203647239037763.25, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.35810980200767517, "optim/lr": 0.0029126802341032025, "optim/total_tokens": 2884632576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 3.991267204284668, "created_at": "2025-01-15T04:30:29.303550+00:00"} {"global_step": 11005, "acc_step": 0, "speed/wps": 12967.808216278321, "speed/FLOPS": 203677247672504.6, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.9745291471481323, "optim/lr": 0.0029126514253549956, "optim/total_tokens": 2884894720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435053, "loss/out": 4.200801849365234, "created_at": "2025-01-15T04:30:34.362085+00:00"} {"global_step": 11006, "acc_step": 0, "speed/wps": 12963.174441210767, "speed/FLOPS": 203604467867594.44, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8449081182479858, "optim/lr": 0.002912622611997766, "optim/total_tokens": 2885156864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468018, "loss/out": 4.094478607177734, "created_at": "2025-01-15T04:30:39.423853+00:00"} {"global_step": 11007, "acc_step": 0, "speed/wps": 12962.844514600914, "speed/FLOPS": 203599285916820.03, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.2615228593349457, "optim/lr": 0.0029125937940316063, "optim/total_tokens": 2885419008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467490, "loss/out": 4.167706489562988, "created_at": "2025-01-15T04:30:44.490141+00:00"} {"global_step": 11008, "acc_step": 0, "speed/wps": 12957.581135370117, "speed/FLOPS": 203516617313359.38, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.0266255140304565, "optim/lr": 0.0029125649714566116, "optim/total_tokens": 2885681152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317586, "loss/out": 4.190086841583252, "created_at": "2025-01-15T04:30:49.551961+00:00"} {"global_step": 11009, "acc_step": 0, "speed/wps": 12962.329917922378, "speed/FLOPS": 203591203468853.7, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9535771608352661, "optim/lr": 0.002912536144272875, "optim/total_tokens": 2885943296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323942, "loss/out": 3.9981484413146973, "created_at": "2025-01-15T04:30:54.612378+00:00"} {"global_step": 11010, "acc_step": 0, "speed/wps": 12966.575392034934, "speed/FLOPS": 203657884473684.16, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.614669680595398, "optim/lr": 0.002912507312480492, "optim/total_tokens": 2886205440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312011, "loss/out": 4.166929721832275, "created_at": "2025-01-15T04:30:59.676584+00:00"} {"global_step": 11011, "acc_step": 0, "speed/wps": 12960.641557095993, "speed/FLOPS": 203564685442024.4, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.9563581943511963, "optim/lr": 0.002912478476079555, "optim/total_tokens": 2886467584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460593, "loss/out": 4.141689300537109, "created_at": "2025-01-15T04:31:04.760465+00:00"} {"global_step": 11012, "acc_step": 0, "speed/wps": 12960.97077954143, "speed/FLOPS": 203569856332929.06, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.49582475423812866, "optim/lr": 0.0029124496350701593, "optim/total_tokens": 2886729728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465057, "loss/out": 4.239588737487793, "created_at": "2025-01-15T04:31:09.824754+00:00"} {"global_step": 11013, "acc_step": 0, "speed/wps": 12963.601001096278, "speed/FLOPS": 203611167576751.1, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.417510062456131, "optim/lr": 0.002912420789452398, "optim/total_tokens": 2886991872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 4.221535682678223, "created_at": "2025-01-15T04:31:14.885691+00:00"} {"global_step": 11014, "acc_step": 0, "speed/wps": 12966.575571232477, "speed/FLOPS": 203657887288227.8, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.32495468854904175, "optim/lr": 0.0029123919392263662, "optim/total_tokens": 2887254016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441679, "loss/out": 4.173498153686523, "created_at": "2025-01-15T04:31:19.949316+00:00"} {"global_step": 11015, "acc_step": 0, "speed/wps": 12963.14884617003, "speed/FLOPS": 203604065862308.4, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.33643803000450134, "optim/lr": 0.0029123630843921576, "optim/total_tokens": 2887516160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331250, "loss/out": 4.176877021789551, "created_at": "2025-01-15T04:31:25.013455+00:00"} {"global_step": 11016, "acc_step": 0, "speed/wps": 12965.328235005307, "speed/FLOPS": 203638296158760.5, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.771347165107727, "optim/lr": 0.002912334224949866, "optim/total_tokens": 2887778304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436151, "loss/out": 4.106851577758789, "created_at": "2025-01-15T04:31:30.076965+00:00"} {"global_step": 11017, "acc_step": 0, "speed/wps": 12963.582296705688, "speed/FLOPS": 203610873798594.38, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0215517282485962, "optim/lr": 0.002912305360899586, "optim/total_tokens": 2888040448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485469, "loss/out": 4.219759941101074, "created_at": "2025-01-15T04:31:35.141265+00:00"} {"global_step": 11018, "acc_step": 0, "speed/wps": 12960.3338701525, "speed/FLOPS": 203559852795770.7, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8312155604362488, "optim/lr": 0.0029122764922414114, "optim/total_tokens": 2888302592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441157, "loss/out": 4.070333480834961, "created_at": "2025-01-15T04:31:40.206254+00:00"} {"global_step": 11019, "acc_step": 0, "speed/wps": 12965.518637978874, "speed/FLOPS": 203641286699101.47, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1163256168365479, "optim/lr": 0.0029122476189754366, "optim/total_tokens": 2888564736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 4.198456764221191, "created_at": "2025-01-15T04:31:45.264712+00:00"} {"global_step": 11020, "acc_step": 0, "speed/wps": 12962.525543325002, "speed/FLOPS": 203594276034619.38, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4989238977432251, "optim/lr": 0.0029122187411017566, "optim/total_tokens": 2888826880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434003, "loss/out": 3.9702000617980957, "created_at": "2025-01-15T04:31:50.329452+00:00"} {"global_step": 11021, "acc_step": 0, "speed/wps": 12965.076807206828, "speed/FLOPS": 203634347139687.4, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9822360277175903, "optim/lr": 0.002912189858620464, "optim/total_tokens": 2889089024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293783, "loss/out": 4.147010803222656, "created_at": "2025-01-15T04:31:55.387767+00:00"} {"global_step": 11022, "acc_step": 0, "speed/wps": 12969.333934029342, "speed/FLOPS": 203701211166340.25, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3085170984268188, "optim/lr": 0.0029121609715316546, "optim/total_tokens": 2889351168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305105, "loss/out": 4.126302719116211, "created_at": "2025-01-15T04:32:00.445796+00:00"} {"global_step": 11023, "acc_step": 0, "speed/wps": 12970.782400951959, "speed/FLOPS": 203723961329762.1, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.843988835811615, "optim/lr": 0.0029121320798354217, "optim/total_tokens": 2889613312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 310665, "loss/out": 4.0531840324401855, "created_at": "2025-01-15T04:32:05.501931+00:00"} {"global_step": 11024, "acc_step": 0, "speed/wps": 12962.44555224198, "speed/FLOPS": 203593019664741.22, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6069895029067993, "optim/lr": 0.00291210318353186, "optim/total_tokens": 2889875456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 359074, "loss/out": 4.07327938079834, "created_at": "2025-01-15T04:32:10.566042+00:00"} {"global_step": 11025, "acc_step": 0, "speed/wps": 12946.292885680683, "speed/FLOPS": 203339319840306.22, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7659196853637695, "optim/lr": 0.0029120742826210634, "optim/total_tokens": 2890137600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 4.054195880889893, "created_at": "2025-01-15T04:32:15.636132+00:00"} {"global_step": 11026, "acc_step": 0, "speed/wps": 12953.202070116275, "speed/FLOPS": 203447838076081.53, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.124348759651184, "optim/lr": 0.0029120453771031263, "optim/total_tokens": 2890399744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443306, "loss/out": 4.100634574890137, "created_at": "2025-01-15T04:32:20.701632+00:00"} {"global_step": 11027, "acc_step": 0, "speed/wps": 12969.33879593644, "speed/FLOPS": 203701287529272.22, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5399869084358215, "optim/lr": 0.0029120164669781427, "optim/total_tokens": 2890661888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450990, "loss/out": 4.242703914642334, "created_at": "2025-01-15T04:32:25.760416+00:00"} {"global_step": 11028, "acc_step": 0, "speed/wps": 12961.682828353561, "speed/FLOPS": 203581040037981.16, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1418448686599731, "optim/lr": 0.002911987552246208, "optim/total_tokens": 2890924032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460034, "loss/out": 3.948529005050659, "created_at": "2025-01-15T04:32:30.820179+00:00"} {"global_step": 11029, "acc_step": 0, "speed/wps": 12961.352127758746, "speed/FLOPS": 203575845930709.34, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6324238777160645, "optim/lr": 0.0029119586329074154, "optim/total_tokens": 2891186176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437195, "loss/out": 4.080111980438232, "created_at": "2025-01-15T04:32:35.881936+00:00"} {"global_step": 11030, "acc_step": 0, "speed/wps": 12960.014332610579, "speed/FLOPS": 203554834019584.22, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.5446470975875854, "optim/lr": 0.00291192970896186, "optim/total_tokens": 2891448320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309845, "loss/out": 4.218240261077881, "created_at": "2025-01-15T04:32:40.945758+00:00"} {"global_step": 11031, "acc_step": 0, "speed/wps": 12957.532385067128, "speed/FLOPS": 203515851622861.0, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.43700334429740906, "optim/lr": 0.002911900780409636, "optim/total_tokens": 2891710464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429470, "loss/out": 4.100796699523926, "created_at": "2025-01-15T04:32:46.016575+00:00"} {"global_step": 11032, "acc_step": 0, "speed/wps": 12964.786769490416, "speed/FLOPS": 203629791698797.56, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5193464756011963, "optim/lr": 0.0029118718472508375, "optim/total_tokens": 2891972608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 3.973407506942749, "created_at": "2025-01-15T04:32:51.077047+00:00"} {"global_step": 11033, "acc_step": 0, "speed/wps": 12961.55011825977, "speed/FLOPS": 203578955643595.4, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.49996572732925415, "optim/lr": 0.002911842909485559, "optim/total_tokens": 2892234752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443545, "loss/out": 4.191974639892578, "created_at": "2025-01-15T04:32:56.137016+00:00"} {"global_step": 11034, "acc_step": 0, "speed/wps": 12963.389512321406, "speed/FLOPS": 203607845854925.7, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4346548318862915, "optim/lr": 0.002911813967113895, "optim/total_tokens": 2892496896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453660, "loss/out": 4.120179176330566, "created_at": "2025-01-15T04:33:01.197311+00:00"} {"global_step": 11035, "acc_step": 0, "speed/wps": 12961.821672387747, "speed/FLOPS": 203583220774329.75, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.45638254284858704, "optim/lr": 0.00291178502013594, "optim/total_tokens": 2892759040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 4.118159770965576, "created_at": "2025-01-15T04:33:06.258193+00:00"} {"global_step": 11036, "acc_step": 0, "speed/wps": 12966.813843833697, "speed/FLOPS": 203661629686850.84, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9536907076835632, "optim/lr": 0.0029117560685517885, "optim/total_tokens": 2893021184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466392, "loss/out": 4.201968193054199, "created_at": "2025-01-15T04:33:11.316554+00:00"} {"global_step": 11037, "acc_step": 0, "speed/wps": 12954.898548262796, "speed/FLOPS": 203474483596577.9, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6988568902015686, "optim/lr": 0.002911727112361535, "optim/total_tokens": 2893283328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464271, "loss/out": 4.051587104797363, "created_at": "2025-01-15T04:33:16.378937+00:00"} {"global_step": 11038, "acc_step": 0, "speed/wps": 12956.802699001319, "speed/FLOPS": 203504390900503.72, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8972316384315491, "optim/lr": 0.002911698151565273, "optim/total_tokens": 2893545472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320721, "loss/out": 4.074088096618652, "created_at": "2025-01-15T04:33:21.445829+00:00"} {"global_step": 11039, "acc_step": 0, "speed/wps": 12963.714864521917, "speed/FLOPS": 203612955958315.5, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8591752052307129, "optim/lr": 0.0029116691861630984, "optim/total_tokens": 2893807616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371642, "loss/out": 3.9276015758514404, "created_at": "2025-01-15T04:33:26.504756+00:00"} {"global_step": 11040, "acc_step": 0, "speed/wps": 12960.67483832785, "speed/FLOPS": 203565208169503.94, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0011, "optim/grad_norm": 0.560859203338623, "optim/lr": 0.0029116402161551047, "optim/total_tokens": 2894069760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438769, "loss/out": 4.191895484924316, "created_at": "2025-01-15T04:33:31.568858+00:00"} {"global_step": 11041, "acc_step": 0, "speed/wps": 12961.172403812989, "speed/FLOPS": 203573023119174.66, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7195098996162415, "optim/lr": 0.0029116112415413873, "optim/total_tokens": 2894331904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 4.030617713928223, "created_at": "2025-01-15T04:33:36.629120+00:00"} {"global_step": 11042, "acc_step": 0, "speed/wps": 12959.412436366487, "speed/FLOPS": 203545380411978.62, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7692204117774963, "optim/lr": 0.0029115822623220398, "optim/total_tokens": 2894594048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429454, "loss/out": 4.080696105957031, "created_at": "2025-01-15T04:33:41.690422+00:00"} {"global_step": 11043, "acc_step": 0, "speed/wps": 12961.394043054222, "speed/FLOPS": 203576504268022.47, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0254535675048828, "optim/lr": 0.002911553278497157, "optim/total_tokens": 2894856192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 4.006200313568115, "created_at": "2025-01-15T04:33:46.753761+00:00"} {"global_step": 11044, "acc_step": 0, "speed/wps": 12953.74757361094, "speed/FLOPS": 203456405958065.62, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0456827878952026, "optim/lr": 0.0029115242900668342, "optim/total_tokens": 2895118336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435866, "loss/out": 4.20887565612793, "created_at": "2025-01-15T04:33:51.820881+00:00"} {"global_step": 11045, "acc_step": 0, "speed/wps": 12961.390060886572, "speed/FLOPS": 203576441722607.44, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7227104902267456, "optim/lr": 0.0029114952970311653, "optim/total_tokens": 2895380480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313436, "loss/out": 4.109442710876465, "created_at": "2025-01-15T04:33:56.881035+00:00"} {"global_step": 11046, "acc_step": 0, "speed/wps": 12961.920792240253, "speed/FLOPS": 203584777587818.3, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5222874879837036, "optim/lr": 0.0029114662993902454, "optim/total_tokens": 2895642624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427332, "loss/out": 4.132504940032959, "created_at": "2025-01-15T04:34:01.941446+00:00"} {"global_step": 11047, "acc_step": 0, "speed/wps": 12955.240046215658, "speed/FLOPS": 203479847291194.97, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.561394989490509, "optim/lr": 0.002911437297144168, "optim/total_tokens": 2895904768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471976, "loss/out": 4.216514587402344, "created_at": "2025-01-15T04:34:07.006009+00:00"} {"global_step": 11048, "acc_step": 0, "speed/wps": 12956.537106813616, "speed/FLOPS": 203500219410233.94, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8269159197807312, "optim/lr": 0.002911408290293029, "optim/total_tokens": 2896166912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 4.120172023773193, "created_at": "2025-01-15T04:34:12.071782+00:00"} {"global_step": 11049, "acc_step": 0, "speed/wps": 12959.82039648101, "speed/FLOPS": 203551787986173.2, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6362676024436951, "optim/lr": 0.0029113792788369223, "optim/total_tokens": 2896429056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462126, "loss/out": 4.162327289581299, "created_at": "2025-01-15T04:34:17.132879+00:00"} {"global_step": 11050, "acc_step": 0, "speed/wps": 12959.126243789531, "speed/FLOPS": 203540885364287.72, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.42514628171920776, "optim/lr": 0.0029113502627759425, "optim/total_tokens": 2896691200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434230, "loss/out": 4.0699920654296875, "created_at": "2025-01-15T04:34:22.197099+00:00"} {"global_step": 11051, "acc_step": 0, "speed/wps": 12955.994650668372, "speed/FLOPS": 203491699391060.03, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0352532863616943, "optim/lr": 0.002911321242110185, "optim/total_tokens": 2896953344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480711, "loss/out": 3.994837522506714, "created_at": "2025-01-15T04:34:27.265757+00:00"} {"global_step": 11052, "acc_step": 0, "speed/wps": 12963.83867717658, "speed/FLOPS": 203614900606193.4, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.37333163619041443, "optim/lr": 0.0029112922168397434, "optim/total_tokens": 2897215488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324713, "loss/out": 4.105818748474121, "created_at": "2025-01-15T04:34:32.327556+00:00"} {"global_step": 11053, "acc_step": 0, "speed/wps": 12961.27752268907, "speed/FLOPS": 203574674155571.97, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7632246613502502, "optim/lr": 0.0029112631869647135, "optim/total_tokens": 2897477632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365215, "loss/out": 4.097407817840576, "created_at": "2025-01-15T04:34:37.388673+00:00"} {"global_step": 11054, "acc_step": 0, "speed/wps": 12957.538032658324, "speed/FLOPS": 203515940326041.3, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0845303535461426, "optim/lr": 0.0029112341524851894, "optim/total_tokens": 2897739776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440628, "loss/out": 4.087684154510498, "created_at": "2025-01-15T04:34:42.451331+00:00"} {"global_step": 11055, "acc_step": 0, "speed/wps": 12963.412194787254, "speed/FLOPS": 203608202114220.8, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.001, "optim/grad_norm": 0.8257366418838501, "optim/lr": 0.0029112051134012663, "optim/total_tokens": 2898001920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478044, "loss/out": 4.157246112823486, "created_at": "2025-01-15T04:34:47.511968+00:00"} {"global_step": 11056, "acc_step": 0, "speed/wps": 12963.935585239762, "speed/FLOPS": 203616422680491.3, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0810965299606323, "optim/lr": 0.002911176069713038, "optim/total_tokens": 2898264064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437438, "loss/out": 4.177830696105957, "created_at": "2025-01-15T04:34:52.573259+00:00"} {"global_step": 11057, "acc_step": 0, "speed/wps": 12957.33171863403, "speed/FLOPS": 203512699880784.62, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9728911519050598, "optim/lr": 0.0029111470214206008, "optim/total_tokens": 2898526208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503465, "loss/out": 4.259879112243652, "created_at": "2025-01-15T04:34:57.668791+00:00"} {"global_step": 11058, "acc_step": 0, "speed/wps": 12965.93749451385, "speed/FLOPS": 203647865416552.47, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7916021943092346, "optim/lr": 0.002911117968524048, "optim/total_tokens": 2898788352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415290, "loss/out": 4.137618541717529, "created_at": "2025-01-15T04:35:02.727082+00:00"} {"global_step": 11059, "acc_step": 0, "speed/wps": 12954.04376372435, "speed/FLOPS": 203461058030804.6, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5382637977600098, "optim/lr": 0.002911088911023474, "optim/total_tokens": 2899050496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461622, "loss/out": 4.05517578125, "created_at": "2025-01-15T04:35:07.791019+00:00"} {"global_step": 11060, "acc_step": 0, "speed/wps": 12955.067036335384, "speed/FLOPS": 203477129933284.78, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6140408515930176, "optim/lr": 0.002911059848918976, "optim/total_tokens": 2899312640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464799, "loss/out": 4.079524993896484, "created_at": "2025-01-15T04:35:12.853921+00:00"} {"global_step": 11061, "acc_step": 0, "speed/wps": 12957.319345393611, "speed/FLOPS": 203512505542041.22, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.555580735206604, "optim/lr": 0.002911030782210647, "optim/total_tokens": 2899574784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430506, "loss/out": 4.103939056396484, "created_at": "2025-01-15T04:35:17.915673+00:00"} {"global_step": 11062, "acc_step": 0, "speed/wps": 12958.86507151741, "speed/FLOPS": 203536783294865.75, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5653654932975769, "optim/lr": 0.0029110017108985816, "optim/total_tokens": 2899836928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433464, "loss/out": 4.074550151824951, "created_at": "2025-01-15T04:35:22.978555+00:00"} {"global_step": 11063, "acc_step": 0, "speed/wps": 12962.42464249686, "speed/FLOPS": 203592691248461.12, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7126030921936035, "optim/lr": 0.0029109726349828757, "optim/total_tokens": 2900099072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 4.090787887573242, "created_at": "2025-01-15T04:35:28.042432+00:00"} {"global_step": 11064, "acc_step": 0, "speed/wps": 12961.709408421048, "speed/FLOPS": 203581457514465.88, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8398763537406921, "optim/lr": 0.0029109435544636235, "optim/total_tokens": 2900361216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 4.092259407043457, "created_at": "2025-01-15T04:35:33.102316+00:00"} {"global_step": 11065, "acc_step": 0, "speed/wps": 12955.030397128747, "speed/FLOPS": 203476554464197.0, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.5275915265083313, "optim/lr": 0.0029109144693409204, "optim/total_tokens": 2900623360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297882, "loss/out": 4.165510177612305, "created_at": "2025-01-15T04:35:38.165341+00:00"} {"global_step": 11066, "acc_step": 0, "speed/wps": 12964.0901859283, "speed/FLOPS": 203618850896748.4, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7347895503044128, "optim/lr": 0.002910885379614861, "optim/total_tokens": 2900885504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439571, "loss/out": 4.035367012023926, "created_at": "2025-01-15T04:35:43.229116+00:00"} {"global_step": 11067, "acc_step": 0, "speed/wps": 12956.006200798889, "speed/FLOPS": 203491880801731.34, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6640552282333374, "optim/lr": 0.00291085628528554, "optim/total_tokens": 2901147648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 4.205419540405273, "created_at": "2025-01-15T04:35:48.296567+00:00"} {"global_step": 11068, "acc_step": 0, "speed/wps": 12962.937630047416, "speed/FLOPS": 203600748422861.38, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.44720497727394104, "optim/lr": 0.0029108271863530523, "optim/total_tokens": 2901409792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 4.188467979431152, "created_at": "2025-01-15T04:35:53.356620+00:00"} {"global_step": 11069, "acc_step": 0, "speed/wps": 12960.362270751166, "speed/FLOPS": 203560298866199.25, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3371992111206055, "optim/lr": 0.0029107980828174933, "optim/total_tokens": 2901671936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 4.110683441162109, "created_at": "2025-01-15T04:35:58.417365+00:00"} {"global_step": 11070, "acc_step": 0, "speed/wps": 12956.35422910534, "speed/FLOPS": 203497347064525.16, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8290621042251587, "optim/lr": 0.0029107689746789575, "optim/total_tokens": 2901934080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442256, "loss/out": 4.1564040184021, "created_at": "2025-01-15T04:36:03.483129+00:00"} {"global_step": 11071, "acc_step": 0, "speed/wps": 12965.894085340049, "speed/FLOPS": 203647183615827.47, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.28693917393684387, "optim/lr": 0.0029107398619375403, "optim/total_tokens": 2902196224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305807, "loss/out": 4.00241756439209, "created_at": "2025-01-15T04:36:08.548871+00:00"} {"global_step": 11072, "acc_step": 0, "speed/wps": 12963.41571728723, "speed/FLOPS": 203608257439923.34, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5348954200744629, "optim/lr": 0.0029107107445933366, "optim/total_tokens": 2902458368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441442, "loss/out": 4.065837860107422, "created_at": "2025-01-15T04:36:13.609278+00:00"} {"global_step": 11073, "acc_step": 0, "speed/wps": 12967.188341104309, "speed/FLOPS": 203667511681098.97, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.042417049407959, "optim/lr": 0.0029106816226464414, "optim/total_tokens": 2902720512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466392, "loss/out": 4.014649868011475, "created_at": "2025-01-15T04:36:18.667819+00:00"} {"global_step": 11074, "acc_step": 0, "speed/wps": 12958.048576629757, "speed/FLOPS": 203523959120673.62, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.001, "optim/grad_norm": 0.7047826051712036, "optim/lr": 0.0029106524960969492, "optim/total_tokens": 2902982656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467703, "loss/out": 4.048750877380371, "created_at": "2025-01-15T04:36:23.728807+00:00"} {"global_step": 11075, "acc_step": 0, "speed/wps": 12959.786501456145, "speed/FLOPS": 203551255618231.0, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6147114038467407, "optim/lr": 0.0029106233649449555, "optim/total_tokens": 2903244800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457095, "loss/out": 4.109227180480957, "created_at": "2025-01-15T04:36:28.801053+00:00"} {"global_step": 11076, "acc_step": 0, "speed/wps": 12952.866778506479, "speed/FLOPS": 203442571860610.75, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1286226511001587, "optim/lr": 0.0029105942291905557, "optim/total_tokens": 2903506944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462150, "loss/out": 3.9995875358581543, "created_at": "2025-01-15T04:36:33.867450+00:00"} {"global_step": 11077, "acc_step": 0, "speed/wps": 12964.834721643985, "speed/FLOPS": 203630544853262.53, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8268032073974609, "optim/lr": 0.002910565088833844, "optim/total_tokens": 2903769088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479095, "loss/out": 4.069180488586426, "created_at": "2025-01-15T04:36:38.930984+00:00"} {"global_step": 11078, "acc_step": 0, "speed/wps": 12947.501015768901, "speed/FLOPS": 203358295183486.4, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4615646302700043, "optim/lr": 0.002910535943874916, "optim/total_tokens": 2904031232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302343, "loss/out": 4.091463088989258, "created_at": "2025-01-15T04:36:43.996297+00:00"} {"global_step": 11079, "acc_step": 0, "speed/wps": 12959.648674812172, "speed/FLOPS": 203549090861393.75, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7991697788238525, "optim/lr": 0.0029105067943138666, "optim/total_tokens": 2904293376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 391480, "loss/out": 4.163036823272705, "created_at": "2025-01-15T04:36:49.057950+00:00"} {"global_step": 11080, "acc_step": 0, "speed/wps": 12964.373999122656, "speed/FLOPS": 203623308572965.97, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9085459113121033, "optim/lr": 0.002910477640150791, "optim/total_tokens": 2904555520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435329, "loss/out": 4.270195960998535, "created_at": "2025-01-15T04:36:54.124422+00:00"} {"global_step": 11081, "acc_step": 0, "speed/wps": 12963.120075153278, "speed/FLOPS": 203603613973954.78, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8982506990432739, "optim/lr": 0.002910448481385784, "optim/total_tokens": 2904817664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 4.093540191650391, "created_at": "2025-01-15T04:36:59.184337+00:00"} {"global_step": 11082, "acc_step": 0, "speed/wps": 12956.197353440213, "speed/FLOPS": 203494883116634.6, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5756716728210449, "optim/lr": 0.002910419318018942, "optim/total_tokens": 2905079808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 4.077663898468018, "created_at": "2025-01-15T04:37:04.248735+00:00"} {"global_step": 11083, "acc_step": 0, "speed/wps": 12959.871133136789, "speed/FLOPS": 203552584875071.62, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0626652240753174, "optim/lr": 0.002910390150050358, "optim/total_tokens": 2905341952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475946, "loss/out": 4.075166702270508, "created_at": "2025-01-15T04:37:09.310505+00:00"} {"global_step": 11084, "acc_step": 0, "speed/wps": 12961.995072837259, "speed/FLOPS": 203585944266666.53, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1504578590393066, "optim/lr": 0.0029103609774801292, "optim/total_tokens": 2905604096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308786, "loss/out": 4.084111213684082, "created_at": "2025-01-15T04:37:14.371514+00:00"} {"global_step": 11085, "acc_step": 0, "speed/wps": 12968.264596659841, "speed/FLOPS": 203684415753528.72, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8440144658088684, "optim/lr": 0.00291033180030835, "optim/total_tokens": 2905866240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298121, "loss/out": 4.192824363708496, "created_at": "2025-01-15T04:37:19.430063+00:00"} {"global_step": 11086, "acc_step": 0, "speed/wps": 12963.467869615914, "speed/FLOPS": 203609076563910.16, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0267283916473389, "optim/lr": 0.002910302618535115, "optim/total_tokens": 2906128384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 409449, "loss/out": 4.073923110961914, "created_at": "2025-01-15T04:37:24.499567+00:00"} {"global_step": 11087, "acc_step": 0, "speed/wps": 12964.397896002245, "speed/FLOPS": 203623683906297.66, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3051754236221313, "optim/lr": 0.00291027343216052, "optim/total_tokens": 2906390528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475987, "loss/out": 4.221614837646484, "created_at": "2025-01-15T04:37:29.559937+00:00"} {"global_step": 11088, "acc_step": 0, "speed/wps": 12964.857946617773, "speed/FLOPS": 203630909633390.3, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8379783034324646, "optim/lr": 0.0029102442411846605, "optim/total_tokens": 2906652672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484902, "loss/out": 4.167837142944336, "created_at": "2025-01-15T04:37:34.618858+00:00"} {"global_step": 11089, "acc_step": 0, "speed/wps": 12964.623292933988, "speed/FLOPS": 203627224074838.53, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5351037979125977, "optim/lr": 0.0029102150456076317, "optim/total_tokens": 2906914816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484711, "loss/out": 4.124032020568848, "created_at": "2025-01-15T04:37:39.683128+00:00"} {"global_step": 11090, "acc_step": 0, "speed/wps": 12958.946221206927, "speed/FLOPS": 203538057862250.97, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9085234999656677, "optim/lr": 0.002910185845429528, "optim/total_tokens": 2907176960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434816, "loss/out": 4.022605895996094, "created_at": "2025-01-15T04:37:44.748949+00:00"} {"global_step": 11091, "acc_step": 0, "speed/wps": 12966.373828997366, "speed/FLOPS": 203654718649202.5, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.058138370513916, "optim/lr": 0.002910156640650446, "optim/total_tokens": 2907439104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477807, "loss/out": 4.181164264678955, "created_at": "2025-01-15T04:37:49.810763+00:00"} {"global_step": 11092, "acc_step": 0, "speed/wps": 12964.043884867211, "speed/FLOPS": 203618123674959.7, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0486952066421509, "optim/lr": 0.0029101274312704796, "optim/total_tokens": 2907701248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461622, "loss/out": 3.9581985473632812, "created_at": "2025-01-15T04:37:54.869885+00:00"} {"global_step": 11093, "acc_step": 0, "speed/wps": 12971.191127451963, "speed/FLOPS": 203730380941093.72, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.238092064857483, "optim/lr": 0.0029100982172897247, "optim/total_tokens": 2907963392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 344264, "loss/out": 4.0344672203063965, "created_at": "2025-01-15T04:37:59.926439+00:00"} {"global_step": 11094, "acc_step": 0, "speed/wps": 12963.76177701659, "speed/FLOPS": 203613692783508.88, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7755478024482727, "optim/lr": 0.0029100689987082767, "optim/total_tokens": 2908225536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324035, "loss/out": 4.02443265914917, "created_at": "2025-01-15T04:38:04.985409+00:00"} {"global_step": 11095, "acc_step": 0, "speed/wps": 12963.367314722951, "speed/FLOPS": 203607497211139.28, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0433588027954102, "optim/lr": 0.0029100397755262315, "optim/total_tokens": 2908487680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 406246, "loss/out": 4.207873344421387, "created_at": "2025-01-15T04:38:10.049765+00:00"} {"global_step": 11096, "acc_step": 0, "speed/wps": 12965.543745483079, "speed/FLOPS": 203641681046956.34, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.465934693813324, "optim/lr": 0.002910010547743683, "optim/total_tokens": 2908749824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459765, "loss/out": 4.183969020843506, "created_at": "2025-01-15T04:38:15.112262+00:00"} {"global_step": 11097, "acc_step": 0, "speed/wps": 12968.783988377689, "speed/FLOPS": 203692573514176.53, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.001, "optim/grad_norm": 1.1605218648910522, "optim/lr": 0.002909981315360728, "optim/total_tokens": 2909011968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470641, "loss/out": 4.118924617767334, "created_at": "2025-01-15T04:38:20.169377+00:00"} {"global_step": 11098, "acc_step": 0, "speed/wps": 12963.185813102125, "speed/FLOPS": 203604646478774.84, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4080601632595062, "optim/lr": 0.002909952078377461, "optim/total_tokens": 2909274112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442208, "loss/out": 4.1975626945495605, "created_at": "2025-01-15T04:38:25.228457+00:00"} {"global_step": 11099, "acc_step": 0, "speed/wps": 12963.288473861883, "speed/FLOPS": 203606258907077.5, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5358650088310242, "optim/lr": 0.0029099228367939775, "optim/total_tokens": 2909536256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 4.014413833618164, "created_at": "2025-01-15T04:38:30.288129+00:00"} {"global_step": 11100, "acc_step": 0, "speed/wps": 12962.018010674285, "speed/FLOPS": 203586304536916.9, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7383224368095398, "optim/lr": 0.0029098935906103737, "optim/total_tokens": 2909798400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448890, "loss/out": 4.17177152633667, "created_at": "2025-01-15T04:38:35.357572+00:00"} {"global_step": 11101, "acc_step": 0, "speed/wps": 12966.778449919897, "speed/FLOPS": 203661073776798.28, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6678758263587952, "optim/lr": 0.002909864339826744, "optim/total_tokens": 2910060544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 294941, "loss/out": 4.144183158874512, "created_at": "2025-01-15T04:38:40.425028+00:00"} {"global_step": 11102, "acc_step": 0, "speed/wps": 12968.028660574995, "speed/FLOPS": 203680710053106.06, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5104323625564575, "optim/lr": 0.0029098350844431837, "optim/total_tokens": 2910322688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319464, "loss/out": 4.143503665924072, "created_at": "2025-01-15T04:38:45.485150+00:00"} {"global_step": 11103, "acc_step": 0, "speed/wps": 12965.153670415912, "speed/FLOPS": 203635554382007.2, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7712006568908691, "optim/lr": 0.0029098058244597896, "optim/total_tokens": 2910584832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368718, "loss/out": 4.247349739074707, "created_at": "2025-01-15T04:38:50.544936+00:00"} {"global_step": 11104, "acc_step": 0, "speed/wps": 12960.125132674812, "speed/FLOPS": 203556574286848.22, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2573999166488647, "optim/lr": 0.0029097765598766557, "optim/total_tokens": 2910846976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431076, "loss/out": 4.159379005432129, "created_at": "2025-01-15T04:38:55.611211+00:00"} {"global_step": 11105, "acc_step": 0, "speed/wps": 12958.75234090683, "speed/FLOPS": 203535012705718.53, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0261991024017334, "optim/lr": 0.0029097472906938785, "optim/total_tokens": 2911109120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495799, "loss/out": 4.2378692626953125, "created_at": "2025-01-15T04:39:00.672980+00:00"} {"global_step": 11106, "acc_step": 0, "speed/wps": 12968.681161685541, "speed/FLOPS": 203690958479677.66, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6960923671722412, "optim/lr": 0.002909718016911553, "optim/total_tokens": 2911371264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443016, "loss/out": 4.011599540710449, "created_at": "2025-01-15T04:39:05.733301+00:00"} {"global_step": 11107, "acc_step": 0, "speed/wps": 12958.587093089141, "speed/FLOPS": 203532417261667.4, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3607193231582642, "optim/lr": 0.0029096887385297748, "optim/total_tokens": 2911633408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489191, "loss/out": 4.060916900634766, "created_at": "2025-01-15T04:39:10.798844+00:00"} {"global_step": 11108, "acc_step": 0, "speed/wps": 12960.652497249677, "speed/FLOPS": 203564857272171.38, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1190565824508667, "optim/lr": 0.0029096594555486394, "optim/total_tokens": 2911895552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436381, "loss/out": 4.072032451629639, "created_at": "2025-01-15T04:39:15.858926+00:00"} {"global_step": 11109, "acc_step": 0, "speed/wps": 12967.042564904725, "speed/FLOPS": 203665222065567.0, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6382733583450317, "optim/lr": 0.0029096301679682425, "optim/total_tokens": 2912157696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 297001, "loss/out": 3.968310832977295, "created_at": "2025-01-15T04:39:20.921610+00:00"} {"global_step": 11110, "acc_step": 0, "speed/wps": 12968.15777514732, "speed/FLOPS": 203682737974886.44, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.08544921875, "optim/lr": 0.0029096008757886795, "optim/total_tokens": 2912419840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 415534, "loss/out": 4.107330799102783, "created_at": "2025-01-15T04:39:25.979579+00:00"} {"global_step": 11111, "acc_step": 0, "speed/wps": 12972.325934047543, "speed/FLOPS": 203748204638064.97, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5919068455696106, "optim/lr": 0.0029095715790100458, "optim/total_tokens": 2912681984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441147, "loss/out": 4.240414142608643, "created_at": "2025-01-15T04:39:31.039891+00:00"} {"global_step": 11112, "acc_step": 0, "speed/wps": 12969.485374346696, "speed/FLOPS": 203703589744625.1, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5319972038269043, "optim/lr": 0.002909542277632438, "optim/total_tokens": 2912944128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488897, "loss/out": 4.123454570770264, "created_at": "2025-01-15T04:39:36.096788+00:00"} {"global_step": 11113, "acc_step": 0, "speed/wps": 12963.595245526009, "speed/FLOPS": 203611077177611.03, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1175552606582642, "optim/lr": 0.00290951297165595, "optim/total_tokens": 2913206272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441158, "loss/out": 4.009819507598877, "created_at": "2025-01-15T04:39:41.160974+00:00"} {"global_step": 11114, "acc_step": 0, "speed/wps": 12953.585553277415, "speed/FLOPS": 203453861206088.3, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4561758041381836, "optim/lr": 0.002909483661080679, "optim/total_tokens": 2913468416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 4.047235488891602, "created_at": "2025-01-15T04:39:46.224299+00:00"} {"global_step": 11115, "acc_step": 0, "speed/wps": 12964.745656324925, "speed/FLOPS": 203629145960038.12, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.0322359800338745, "optim/lr": 0.002909454345906719, "optim/total_tokens": 2913730560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 3.9774014949798584, "created_at": "2025-01-15T04:39:51.284280+00:00"} {"global_step": 11116, "acc_step": 0, "speed/wps": 12964.27952072805, "speed/FLOPS": 203621824659951.75, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2270779609680176, "optim/lr": 0.0029094250261341675, "optim/total_tokens": 2913992704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470897, "loss/out": 4.161104202270508, "created_at": "2025-01-15T04:39:56.348406+00:00"} {"global_step": 11117, "acc_step": 0, "speed/wps": 12968.303600322006, "speed/FLOPS": 203685028359639.66, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1243932247161865, "optim/lr": 0.0029093957017631194, "optim/total_tokens": 2914254848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437195, "loss/out": 4.206720352172852, "created_at": "2025-01-15T04:40:01.412560+00:00"} {"global_step": 11118, "acc_step": 0, "speed/wps": 12964.712406732717, "speed/FLOPS": 203628623729502.75, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5709994435310364, "optim/lr": 0.00290936637279367, "optim/total_tokens": 2914516992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 365740, "loss/out": 4.064056873321533, "created_at": "2025-01-15T04:40:06.474991+00:00"} {"global_step": 11119, "acc_step": 0, "speed/wps": 12969.049805633527, "speed/FLOPS": 203696748539450.66, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2613791227340698, "optim/lr": 0.0029093370392259147, "optim/total_tokens": 2914779136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441679, "loss/out": 4.120852947235107, "created_at": "2025-01-15T04:40:11.564842+00:00"} {"global_step": 11120, "acc_step": 0, "speed/wps": 12967.78769787537, "speed/FLOPS": 203676925402791.0, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0436742305755615, "optim/lr": 0.00290930770105995, "optim/total_tokens": 2915041280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 4.143397808074951, "created_at": "2025-01-15T04:40:16.622392+00:00"} {"global_step": 11121, "acc_step": 0, "speed/wps": 12969.091510195563, "speed/FLOPS": 203697403566905.5, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1390513181686401, "optim/lr": 0.0029092783582958717, "optim/total_tokens": 2915303424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 4.119232654571533, "created_at": "2025-01-15T04:40:21.685112+00:00"} {"global_step": 11122, "acc_step": 0, "speed/wps": 12959.475544776498, "speed/FLOPS": 203546371616277.9, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5160713195800781, "optim/lr": 0.002909249010933775, "optim/total_tokens": 2915565568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317586, "loss/out": 3.9989147186279297, "created_at": "2025-01-15T04:40:26.749825+00:00"} {"global_step": 11123, "acc_step": 0, "speed/wps": 12952.831742030903, "speed/FLOPS": 203442021564616.66, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7900717854499817, "optim/lr": 0.0029092196589737556, "optim/total_tokens": 2915827712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447029, "loss/out": 4.257717609405518, "created_at": "2025-01-15T04:40:31.816596+00:00"} {"global_step": 11124, "acc_step": 0, "speed/wps": 12956.311936090164, "speed/FLOPS": 203496682794605.12, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5981146097183228, "optim/lr": 0.0029091903024159095, "optim/total_tokens": 2916089856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486517, "loss/out": 4.22406530380249, "created_at": "2025-01-15T04:40:36.878391+00:00"} {"global_step": 11125, "acc_step": 0, "speed/wps": 12958.148662462256, "speed/FLOPS": 203525531106205.34, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.897965669631958, "optim/lr": 0.0029091609412603327, "optim/total_tokens": 2916352000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437432, "loss/out": 4.0088090896606445, "created_at": "2025-01-15T04:40:41.944527+00:00"} {"global_step": 11126, "acc_step": 0, "speed/wps": 12964.535391354428, "speed/FLOPS": 203625843459742.1, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0806258916854858, "optim/lr": 0.0029091315755071213, "optim/total_tokens": 2916614144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 4.095677375793457, "created_at": "2025-01-15T04:40:47.003575+00:00"} {"global_step": 11127, "acc_step": 0, "speed/wps": 12956.946756345975, "speed/FLOPS": 203506653519057.1, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6817286014556885, "optim/lr": 0.0029091022051563704, "optim/total_tokens": 2916876288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435338, "loss/out": 4.183075904846191, "created_at": "2025-01-15T04:40:52.066221+00:00"} {"global_step": 11128, "acc_step": 0, "speed/wps": 12957.720041864424, "speed/FLOPS": 203518799030730.0, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4783056974411011, "optim/lr": 0.002909072830208175, "optim/total_tokens": 2917138432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435852, "loss/out": 4.114342212677002, "created_at": "2025-01-15T04:40:57.132019+00:00"} {"global_step": 11129, "acc_step": 0, "speed/wps": 12971.727807635894, "speed/FLOPS": 203738810240858.22, "speed/curr_iter_time": 1.2728, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1118316650390625, "optim/lr": 0.0029090434506626327, "optim/total_tokens": 2917400576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431842, "loss/out": 4.204893112182617, "created_at": "2025-01-15T04:41:02.193371+00:00"} {"global_step": 11130, "acc_step": 0, "speed/wps": 12958.08238382082, "speed/FLOPS": 203524490109065.22, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4656845033168793, "optim/lr": 0.002909014066519839, "optim/total_tokens": 2917662720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325620, "loss/out": 4.01594352722168, "created_at": "2025-01-15T04:41:07.255619+00:00"} {"global_step": 11131, "acc_step": 0, "speed/wps": 12958.334358897026, "speed/FLOPS": 203528447723887.06, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6937260031700134, "optim/lr": 0.002908984677779889, "optim/total_tokens": 2917924864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465870, "loss/out": 4.039951801300049, "created_at": "2025-01-15T04:41:12.321630+00:00"} {"global_step": 11132, "acc_step": 0, "speed/wps": 12958.628782888545, "speed/FLOPS": 203533072057254.84, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6055793166160583, "optim/lr": 0.0029089552844428785, "optim/total_tokens": 2918187008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 4.098143577575684, "created_at": "2025-01-15T04:41:17.383046+00:00"} {"global_step": 11133, "acc_step": 0, "speed/wps": 12955.227758063775, "speed/FLOPS": 203479654288883.97, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1442203521728516, "optim/lr": 0.0029089258865089046, "optim/total_tokens": 2918449152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428403, "loss/out": 4.0425190925598145, "created_at": "2025-01-15T04:41:22.450946+00:00"} {"global_step": 11134, "acc_step": 0, "speed/wps": 12968.42299791729, "speed/FLOPS": 203686903662943.1, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7781906723976135, "optim/lr": 0.0029088964839780623, "optim/total_tokens": 2918711296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 4.158464431762695, "created_at": "2025-01-15T04:41:27.514802+00:00"} {"global_step": 11135, "acc_step": 0, "speed/wps": 12964.345788736808, "speed/FLOPS": 203622865490091.1, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.33079683780670166, "optim/lr": 0.0029088670768504473, "optim/total_tokens": 2918973440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325907, "loss/out": 4.067365646362305, "created_at": "2025-01-15T04:41:32.574314+00:00"} {"global_step": 11136, "acc_step": 0, "speed/wps": 12962.408871288035, "speed/FLOPS": 203592443539954.62, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9994949698448181, "optim/lr": 0.0029088376651261565, "optim/total_tokens": 2919235584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353982, "loss/out": 4.000898361206055, "created_at": "2025-01-15T04:41:37.634832+00:00"} {"global_step": 11137, "acc_step": 0, "speed/wps": 12962.567695441254, "speed/FLOPS": 203594938091528.2, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.465043306350708, "optim/lr": 0.002908808248805285, "optim/total_tokens": 2919497728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443307, "loss/out": 4.083524227142334, "created_at": "2025-01-15T04:41:42.694203+00:00"} {"global_step": 11138, "acc_step": 0, "speed/wps": 12962.245586016652, "speed/FLOPS": 203589878920390.06, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.8584137558937073, "optim/lr": 0.002908778827887929, "optim/total_tokens": 2919759872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 4.099849700927734, "created_at": "2025-01-15T04:41:47.756793+00:00"} {"global_step": 11139, "acc_step": 0, "speed/wps": 12953.672846218185, "speed/FLOPS": 203455232261676.2, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1303236484527588, "optim/lr": 0.0029087494023741856, "optim/total_tokens": 2920022016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446218, "loss/out": 4.08371639251709, "created_at": "2025-01-15T04:41:52.819675+00:00"} {"global_step": 11140, "acc_step": 0, "speed/wps": 12965.53912053494, "speed/FLOPS": 203641608405790.5, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2992011308670044, "optim/lr": 0.002908719972264149, "optim/total_tokens": 2920284160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453911, "loss/out": 4.102370738983154, "created_at": "2025-01-15T04:41:57.881250+00:00"} {"global_step": 11141, "acc_step": 0, "speed/wps": 12962.72241155454, "speed/FLOPS": 203597368120689.56, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9667930603027344, "optim/lr": 0.0029086905375579164, "optim/total_tokens": 2920546304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 4.128418445587158, "created_at": "2025-01-15T04:42:02.946881+00:00"} {"global_step": 11142, "acc_step": 0, "speed/wps": 12962.778646391062, "speed/FLOPS": 203598251366071.88, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3017079830169678, "optim/lr": 0.002908661098255583, "optim/total_tokens": 2920808448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 4.052811622619629, "created_at": "2025-01-15T04:42:08.006988+00:00"} {"global_step": 11143, "acc_step": 0, "speed/wps": 12959.80614723059, "speed/FLOPS": 203551564182114.8, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6516854763031006, "optim/lr": 0.002908631654357245, "optim/total_tokens": 2921070592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433464, "loss/out": 4.181659698486328, "created_at": "2025-01-15T04:42:13.069448+00:00"} {"global_step": 11144, "acc_step": 0, "speed/wps": 12956.804295035035, "speed/FLOPS": 203504415968406.44, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7061843872070312, "optim/lr": 0.0029086022058629995, "optim/total_tokens": 2921332736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 289311, "loss/out": 4.21993350982666, "created_at": "2025-01-15T04:42:18.136394+00:00"} {"global_step": 11145, "acc_step": 0, "speed/wps": 12959.805968220107, "speed/FLOPS": 203551561370509.2, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6755102276802063, "optim/lr": 0.002908572752772942, "optim/total_tokens": 2921594880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450462, "loss/out": 4.085005760192871, "created_at": "2025-01-15T04:42:23.197270+00:00"} {"global_step": 11146, "acc_step": 0, "speed/wps": 12958.438868575715, "speed/FLOPS": 203530089192002.22, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6827390193939209, "optim/lr": 0.002908543295087168, "optim/total_tokens": 2921857024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487905, "loss/out": 4.149578094482422, "created_at": "2025-01-15T04:42:28.263528+00:00"} {"global_step": 11147, "acc_step": 0, "speed/wps": 12951.73970755817, "speed/FLOPS": 203424869662609.78, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6342544555664062, "optim/lr": 0.0029085138328057737, "optim/total_tokens": 2922119168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443545, "loss/out": 4.08124303817749, "created_at": "2025-01-15T04:42:33.329824+00:00"} {"global_step": 11148, "acc_step": 0, "speed/wps": 12957.360313650477, "speed/FLOPS": 203513149004811.6, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0186277627944946, "optim/lr": 0.0029084843659288565, "optim/total_tokens": 2922381312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461605, "loss/out": 4.080667495727539, "created_at": "2025-01-15T04:42:38.393067+00:00"} {"global_step": 11149, "acc_step": 0, "speed/wps": 12961.901808576044, "speed/FLOPS": 203584479423285.6, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5933600664138794, "optim/lr": 0.002908454894456511, "optim/total_tokens": 2922643456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334608, "loss/out": 4.155371189117432, "created_at": "2025-01-15T04:42:43.458979+00:00"} {"global_step": 11150, "acc_step": 0, "speed/wps": 12961.926159534596, "speed/FLOPS": 203584861888551.4, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4640662372112274, "optim/lr": 0.002908425418388834, "optim/total_tokens": 2922905600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 387505, "loss/out": 4.1499924659729, "created_at": "2025-01-15T04:42:48.522794+00:00"} {"global_step": 11151, "acc_step": 0, "speed/wps": 12960.761353143067, "speed/FLOPS": 203566567003560.72, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6892580986022949, "optim/lr": 0.002908395937725922, "optim/total_tokens": 2923167744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462435, "loss/out": 4.145242691040039, "created_at": "2025-01-15T04:42:53.583972+00:00"} {"global_step": 11152, "acc_step": 0, "speed/wps": 12967.398565488891, "speed/FLOPS": 203670813543937.4, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.8518179655075073, "optim/lr": 0.0029083664524678704, "optim/total_tokens": 2923429888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474897, "loss/out": 4.122828483581543, "created_at": "2025-01-15T04:42:58.643802+00:00"} {"global_step": 11153, "acc_step": 0, "speed/wps": 12960.470197980554, "speed/FLOPS": 203561994011644.34, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5985216498374939, "optim/lr": 0.0029083369626147766, "optim/total_tokens": 2923692032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427880, "loss/out": 4.134974479675293, "created_at": "2025-01-15T04:43:03.704176+00:00"} {"global_step": 11154, "acc_step": 0, "speed/wps": 12961.700216523346, "speed/FLOPS": 203581313143081.5, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.4475729763507843, "optim/lr": 0.002908307468166735, "optim/total_tokens": 2923954176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475696, "loss/out": 4.031549453735352, "created_at": "2025-01-15T04:43:08.768716+00:00"} {"global_step": 11155, "acc_step": 0, "speed/wps": 12962.658036996063, "speed/FLOPS": 203596357029788.75, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6012637615203857, "optim/lr": 0.0029082779691238436, "optim/total_tokens": 2924216320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 4.0534772872924805, "created_at": "2025-01-15T04:43:13.833027+00:00"} {"global_step": 11156, "acc_step": 0, "speed/wps": 12961.557540842678, "speed/FLOPS": 203579072225459.66, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.001, "optim/grad_norm": 1.316293716430664, "optim/lr": 0.0029082484654861978, "optim/total_tokens": 2924478464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460578, "loss/out": 4.251510143280029, "created_at": "2025-01-15T04:43:18.893133+00:00"} {"global_step": 11157, "acc_step": 0, "speed/wps": 12967.377950408814, "speed/FLOPS": 203670489755776.72, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0048326253890991, "optim/lr": 0.0029082189572538944, "optim/total_tokens": 2924740608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443307, "loss/out": 4.159514427185059, "created_at": "2025-01-15T04:43:23.950829+00:00"} {"global_step": 11158, "acc_step": 0, "speed/wps": 12959.0997482983, "speed/FLOPS": 203540469216189.34, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5447742938995361, "optim/lr": 0.002908189444427029, "optim/total_tokens": 2925002752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313142, "loss/out": 4.041443824768066, "created_at": "2025-01-15T04:43:29.016067+00:00"} {"global_step": 11159, "acc_step": 0, "speed/wps": 12957.606818185099, "speed/FLOPS": 203517020697260.53, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.081221103668213, "optim/lr": 0.0029081599270056975, "optim/total_tokens": 2925264896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437195, "loss/out": 4.0337815284729, "created_at": "2025-01-15T04:43:34.083060+00:00"} {"global_step": 11160, "acc_step": 0, "speed/wps": 12954.625328520327, "speed/FLOPS": 203470192305079.2, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8410831093788147, "optim/lr": 0.002908130404989998, "optim/total_tokens": 2925527040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473043, "loss/out": 4.168604850769043, "created_at": "2025-01-15T04:43:39.145816+00:00"} {"global_step": 11161, "acc_step": 0, "speed/wps": 12956.307267629974, "speed/FLOPS": 203496609470022.72, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0124393701553345, "optim/lr": 0.002908100878380025, "optim/total_tokens": 2925789184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 426829, "loss/out": 4.112085342407227, "created_at": "2025-01-15T04:43:44.207801+00:00"} {"global_step": 11162, "acc_step": 0, "speed/wps": 12956.73977407502, "speed/FLOPS": 203503402578068.3, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8617632389068604, "optim/lr": 0.002908071347175875, "optim/total_tokens": 2926051328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 4.103847980499268, "created_at": "2025-01-15T04:43:49.271405+00:00"} {"global_step": 11163, "acc_step": 0, "speed/wps": 12957.357925359334, "speed/FLOPS": 203513111493417.34, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.0237808227539062, "optim/lr": 0.002908041811377646, "optim/total_tokens": 2926313472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314942, "loss/out": 4.089738845825195, "created_at": "2025-01-15T04:43:54.336497+00:00"} {"global_step": 11164, "acc_step": 0, "speed/wps": 12961.496107228182, "speed/FLOPS": 203578107326125.38, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5249778032302856, "optim/lr": 0.002908012270985433, "optim/total_tokens": 2926575616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355857, "loss/out": 3.972886323928833, "created_at": "2025-01-15T04:43:59.399709+00:00"} {"global_step": 11165, "acc_step": 0, "speed/wps": 12963.965854976597, "speed/FLOPS": 203616898108303.28, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8209323287010193, "optim/lr": 0.002907982725999332, "optim/total_tokens": 2926837760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 4.0931854248046875, "created_at": "2025-01-15T04:44:04.461900+00:00"} {"global_step": 11166, "acc_step": 0, "speed/wps": 12956.849279967717, "speed/FLOPS": 203505122518589.97, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0495448112487793, "optim/lr": 0.0029079531764194407, "optim/total_tokens": 2927099904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486761, "loss/out": 4.00596809387207, "created_at": "2025-01-15T04:44:09.524563+00:00"} {"global_step": 11167, "acc_step": 0, "speed/wps": 12959.394849081953, "speed/FLOPS": 203545104179507.56, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4526524543762207, "optim/lr": 0.0029079236222458544, "optim/total_tokens": 2927362048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430506, "loss/out": 4.046546936035156, "created_at": "2025-01-15T04:44:14.586592+00:00"} {"global_step": 11168, "acc_step": 0, "speed/wps": 12958.879574050257, "speed/FLOPS": 203537011077072.7, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.45639586448669434, "optim/lr": 0.00290789406347867, "optim/total_tokens": 2927624192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436381, "loss/out": 4.03392219543457, "created_at": "2025-01-15T04:44:19.648685+00:00"} {"global_step": 11169, "acc_step": 0, "speed/wps": 12953.898027758607, "speed/FLOPS": 203458769047200.38, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.030221700668335, "optim/lr": 0.0029078645001179844, "optim/total_tokens": 2927886336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441157, "loss/out": 3.9476678371429443, "created_at": "2025-01-15T04:44:24.711664+00:00"} {"global_step": 11170, "acc_step": 0, "speed/wps": 12963.232055496695, "speed/FLOPS": 203605372779125.28, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5950084924697876, "optim/lr": 0.002907834932163893, "optim/total_tokens": 2928148480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480190, "loss/out": 4.139819145202637, "created_at": "2025-01-15T04:44:29.775573+00:00"} {"global_step": 11171, "acc_step": 0, "speed/wps": 12962.999772230485, "speed/FLOPS": 203601724451238.78, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0866645574569702, "optim/lr": 0.0029078053596164927, "optim/total_tokens": 2928410624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315181, "loss/out": 4.012084484100342, "created_at": "2025-01-15T04:44:34.836383+00:00"} {"global_step": 11172, "acc_step": 0, "speed/wps": 12954.566216694086, "speed/FLOPS": 203469263872619.88, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.8072351217269897, "optim/lr": 0.002907775782475881, "optim/total_tokens": 2928672768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401696, "loss/out": 4.041454315185547, "created_at": "2025-01-15T04:44:39.908938+00:00"} {"global_step": 11173, "acc_step": 0, "speed/wps": 12955.89862937686, "speed/FLOPS": 203490191244729.75, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8509032726287842, "optim/lr": 0.002907746200742153, "optim/total_tokens": 2928934912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471217, "loss/out": 3.9875001907348633, "created_at": "2025-01-15T04:44:44.974548+00:00"} {"global_step": 11174, "acc_step": 0, "speed/wps": 12953.745775641339, "speed/FLOPS": 203456377718482.34, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1417710781097412, "optim/lr": 0.0029077166144154056, "optim/total_tokens": 2929197056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 428118, "loss/out": 4.035694122314453, "created_at": "2025-01-15T04:44:50.037498+00:00"} {"global_step": 11175, "acc_step": 0, "speed/wps": 12956.90769452365, "speed/FLOPS": 203506039999460.9, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0297743082046509, "optim/lr": 0.0029076870234957353, "optim/total_tokens": 2929459200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472230, "loss/out": 4.037837028503418, "created_at": "2025-01-15T04:44:55.099553+00:00"} {"global_step": 11176, "acc_step": 0, "speed/wps": 12959.075952088806, "speed/FLOPS": 203540095464019.8, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7629207372665405, "optim/lr": 0.0029076574279832392, "optim/total_tokens": 2929721344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467734, "loss/out": 4.191296100616455, "created_at": "2025-01-15T04:45:00.160770+00:00"} {"global_step": 11177, "acc_step": 0, "speed/wps": 12959.382703406061, "speed/FLOPS": 203544913414977.78, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8122868537902832, "optim/lr": 0.0029076278278780134, "optim/total_tokens": 2929983488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317879, "loss/out": 4.077953338623047, "created_at": "2025-01-15T04:45:05.228689+00:00"} {"global_step": 11178, "acc_step": 0, "speed/wps": 12957.782518188764, "speed/FLOPS": 203519780307255.22, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.1360362768173218, "optim/lr": 0.0029075982231801544, "optim/total_tokens": 2930245632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449942, "loss/out": 4.050471305847168, "created_at": "2025-01-15T04:45:10.294703+00:00"} {"global_step": 11179, "acc_step": 0, "speed/wps": 12951.473834226652, "speed/FLOPS": 203420693756590.03, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.9121460914611816, "optim/lr": 0.0029075686138897586, "optim/total_tokens": 2930507776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 486288, "loss/out": 4.1042656898498535, "created_at": "2025-01-15T04:45:15.362924+00:00"} {"global_step": 11180, "acc_step": 0, "speed/wps": 12942.761476226962, "speed/FLOPS": 203283854202170.44, "speed/curr_iter_time": 1.2767, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6033966541290283, "optim/lr": 0.002907539000006924, "optim/total_tokens": 2930769920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478372, "loss/out": 3.9535534381866455, "created_at": "2025-01-15T04:45:20.434648+00:00"} {"global_step": 11181, "acc_step": 0, "speed/wps": 12951.234818809111, "speed/FLOPS": 203416939691015.94, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5055193305015564, "optim/lr": 0.0029075093815317454, "optim/total_tokens": 2931032064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431618, "loss/out": 4.089370250701904, "created_at": "2025-01-15T04:45:25.499688+00:00"} {"global_step": 11182, "acc_step": 0, "speed/wps": 12961.087473437336, "speed/FLOPS": 203571689170918.47, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2735533714294434, "optim/lr": 0.00290747975846432, "optim/total_tokens": 2931294208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441442, "loss/out": 4.227602958679199, "created_at": "2025-01-15T04:45:30.559692+00:00"} {"global_step": 11183, "acc_step": 0, "speed/wps": 12964.369709397275, "speed/FLOPS": 203623241196934.16, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0532058477401733, "optim/lr": 0.002907450130804745, "optim/total_tokens": 2931556352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326727, "loss/out": 4.061568737030029, "created_at": "2025-01-15T04:45:35.624347+00:00"} {"global_step": 11184, "acc_step": 0, "speed/wps": 12959.366400216853, "speed/FLOPS": 203544657350988.34, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.636945903301239, "optim/lr": 0.0029074204985531167, "optim/total_tokens": 2931818496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366319, "loss/out": 4.112362384796143, "created_at": "2025-01-15T04:45:40.685194+00:00"} {"global_step": 11185, "acc_step": 0, "speed/wps": 12956.135817957116, "speed/FLOPS": 203493916617313.94, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0091668367385864, "optim/lr": 0.002907390861709532, "optim/total_tokens": 2932080640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444357, "loss/out": 4.176591396331787, "created_at": "2025-01-15T04:45:45.747834+00:00"} {"global_step": 11186, "acc_step": 0, "speed/wps": 12959.7558717337, "speed/FLOPS": 203550774536350.38, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5951741933822632, "optim/lr": 0.002907361220274087, "optim/total_tokens": 2932342784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482328, "loss/out": 4.026157379150391, "created_at": "2025-01-15T04:45:50.813514+00:00"} {"global_step": 11187, "acc_step": 0, "speed/wps": 12962.388350865753, "speed/FLOPS": 203592121238524.66, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8840853571891785, "optim/lr": 0.002907331574246879, "optim/total_tokens": 2932604928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433464, "loss/out": 4.114075660705566, "created_at": "2025-01-15T04:45:55.873265+00:00"} {"global_step": 11188, "acc_step": 0, "speed/wps": 12959.524149793086, "speed/FLOPS": 203547135024853.7, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.031894564628601, "optim/lr": 0.0029073019236280045, "optim/total_tokens": 2932867072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 4.1235151290893555, "created_at": "2025-01-15T04:46:00.941560+00:00"} {"global_step": 11189, "acc_step": 0, "speed/wps": 12964.997300596644, "speed/FLOPS": 203633098379120.94, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5933462381362915, "optim/lr": 0.00290727226841756, "optim/total_tokens": 2933129216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443306, "loss/out": 4.097802639007568, "created_at": "2025-01-15T04:46:06.000771+00:00"} {"global_step": 11190, "acc_step": 0, "speed/wps": 12960.058215606334, "speed/FLOPS": 203555523262334.1, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0353801250457764, "optim/lr": 0.0029072426086156427, "optim/total_tokens": 2933391360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325239, "loss/out": 4.027655601501465, "created_at": "2025-01-15T04:46:11.066313+00:00"} {"global_step": 11191, "acc_step": 0, "speed/wps": 12959.639817594587, "speed/FLOPS": 203548951746620.5, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1093862056732178, "optim/lr": 0.0029072129442223498, "optim/total_tokens": 2933653504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440105, "loss/out": 4.184810638427734, "created_at": "2025-01-15T04:46:16.137368+00:00"} {"global_step": 11192, "acc_step": 0, "speed/wps": 12961.561444327288, "speed/FLOPS": 203579133535049.4, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2211227416992188, "optim/lr": 0.0029071832752377764, "optim/total_tokens": 2933915648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452325, "loss/out": 3.9973888397216797, "created_at": "2025-01-15T04:46:21.200448+00:00"} {"global_step": 11193, "acc_step": 0, "speed/wps": 12958.925093647556, "speed/FLOPS": 203537726024898.75, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.039020299911499, "optim/lr": 0.002907153601662021, "optim/total_tokens": 2934177792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474361, "loss/out": 4.061261177062988, "created_at": "2025-01-15T04:46:26.270110+00:00"} {"global_step": 11194, "acc_step": 0, "speed/wps": 12960.132630007394, "speed/FLOPS": 203556692042758.84, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.1560263633728027, "optim/lr": 0.00290712392349518, "optim/total_tokens": 2934439936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441971, "loss/out": 4.042078971862793, "created_at": "2025-01-15T04:46:31.333594+00:00"} {"global_step": 11195, "acc_step": 0, "speed/wps": 12962.324056032689, "speed/FLOPS": 203591111399821.03, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.40543216466903687, "optim/lr": 0.002907094240737349, "optim/total_tokens": 2934702080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429977, "loss/out": 4.145280361175537, "created_at": "2025-01-15T04:46:36.398625+00:00"} {"global_step": 11196, "acc_step": 0, "speed/wps": 12958.44097329521, "speed/FLOPS": 203530122249514.1, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5179694294929504, "optim/lr": 0.002907064553388627, "optim/total_tokens": 2934964224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315714, "loss/out": 4.041561126708984, "created_at": "2025-01-15T04:46:41.463027+00:00"} {"global_step": 11197, "acc_step": 0, "speed/wps": 12962.581977031063, "speed/FLOPS": 203595162403521.1, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.001, "optim/grad_norm": 1.0648452043533325, "optim/lr": 0.0029070348614491093, "optim/total_tokens": 2935226368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 407300, "loss/out": 4.158026695251465, "created_at": "2025-01-15T04:46:46.526615+00:00"} {"global_step": 11198, "acc_step": 0, "speed/wps": 12961.43138912569, "speed/FLOPS": 203577090839393.75, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8846462368965149, "optim/lr": 0.002907005164918893, "optim/total_tokens": 2935488512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465870, "loss/out": 4.003640174865723, "created_at": "2025-01-15T04:46:51.592836+00:00"} {"global_step": 11199, "acc_step": 0, "speed/wps": 12964.138591502444, "speed/FLOPS": 203619611172806.56, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9411673545837402, "optim/lr": 0.0029069754637980755, "optim/total_tokens": 2935750656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451275, "loss/out": 4.212880611419678, "created_at": "2025-01-15T04:46:56.656506+00:00"} {"global_step": 11200, "acc_step": 0, "speed/wps": 12962.413907086955, "speed/FLOPS": 203592522634096.56, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8057692050933838, "optim/lr": 0.0029069457580867533, "optim/total_tokens": 2936012800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 4.136715412139893, "created_at": "2025-01-15T04:47:01.717851+00:00"} {"global_step": 11201, "acc_step": 0, "speed/wps": 12966.228023327027, "speed/FLOPS": 203652428570791.84, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8314110040664673, "optim/lr": 0.002906916047785023, "optim/total_tokens": 2936274944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431320, "loss/out": 4.1447343826293945, "created_at": "2025-01-15T04:47:06.777151+00:00"} {"global_step": 11202, "acc_step": 0, "speed/wps": 12962.917566437874, "speed/FLOPS": 203600433296303.22, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9486885070800781, "optim/lr": 0.0029068863328929827, "optim/total_tokens": 2936537088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 291959, "loss/out": 4.047307014465332, "created_at": "2025-01-15T04:47:11.841540+00:00"} {"global_step": 11203, "acc_step": 0, "speed/wps": 12956.925086584268, "speed/FLOPS": 203506313165672.1, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.2826573848724365, "optim/lr": 0.0029068566134107282, "optim/total_tokens": 2936799232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437195, "loss/out": 4.1493940353393555, "created_at": "2025-01-15T04:47:16.908486+00:00"} {"global_step": 11204, "acc_step": 0, "speed/wps": 12961.499726571574, "speed/FLOPS": 203578164172886.84, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5681238770484924, "optim/lr": 0.002906826889338357, "optim/total_tokens": 2937061376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 4.062081336975098, "created_at": "2025-01-15T04:47:21.974537+00:00"} {"global_step": 11205, "acc_step": 0, "speed/wps": 12962.729990692595, "speed/FLOPS": 203597487161467.53, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0174784660339355, "optim/lr": 0.0029067971606759656, "optim/total_tokens": 2937323520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457665, "loss/out": 4.204070568084717, "created_at": "2025-01-15T04:47:27.034576+00:00"} {"global_step": 11206, "acc_step": 0, "speed/wps": 12960.304877704713, "speed/FLOPS": 203559397429538.4, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6740001440048218, "optim/lr": 0.002906767427423652, "optim/total_tokens": 2937585664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 4.058980941772461, "created_at": "2025-01-15T04:47:32.097160+00:00"} {"global_step": 11207, "acc_step": 0, "speed/wps": 12960.722834070499, "speed/FLOPS": 203565962008595.28, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8025404810905457, "optim/lr": 0.002906737689581512, "optim/total_tokens": 2937847808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301637, "loss/out": 4.0722150802612305, "created_at": "2025-01-15T04:47:37.162647+00:00"} {"global_step": 11208, "acc_step": 0, "speed/wps": 12958.913927308016, "speed/FLOPS": 203537550642192.12, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0046719312667847, "optim/lr": 0.0029067079471496436, "optim/total_tokens": 2938109952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 298593, "loss/out": 3.9575610160827637, "created_at": "2025-01-15T04:47:42.223398+00:00"} {"global_step": 11209, "acc_step": 0, "speed/wps": 12964.734682985194, "speed/FLOPS": 203628973608658.7, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1333379745483398, "optim/lr": 0.002906678200128143, "optim/total_tokens": 2938372096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306381, "loss/out": 4.1029253005981445, "created_at": "2025-01-15T04:47:47.285705+00:00"} {"global_step": 11210, "acc_step": 0, "speed/wps": 12957.926645044456, "speed/FLOPS": 203522044017574.9, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1293389797210693, "optim/lr": 0.0029066484485171075, "optim/total_tokens": 2938634240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 407007, "loss/out": 4.132354736328125, "created_at": "2025-01-15T04:47:52.351402+00:00"} {"global_step": 11211, "acc_step": 0, "speed/wps": 12970.556479012579, "speed/FLOPS": 203720412915254.84, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6635825037956238, "optim/lr": 0.0029066186923166347, "optim/total_tokens": 2938896384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439054, "loss/out": 4.072678565979004, "created_at": "2025-01-15T04:47:57.409808+00:00"} {"global_step": 11212, "acc_step": 0, "speed/wps": 12959.712403696936, "speed/FLOPS": 203550091811101.66, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.098288655281067, "optim/lr": 0.0029065889315268213, "optim/total_tokens": 2939158528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 4.07442569732666, "created_at": "2025-01-15T04:48:02.476344+00:00"} {"global_step": 11213, "acc_step": 0, "speed/wps": 12961.310599486784, "speed/FLOPS": 203575193672132.38, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6933828592300415, "optim/lr": 0.0029065591661477648, "optim/total_tokens": 2939420672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430791, "loss/out": 4.0631513595581055, "created_at": "2025-01-15T04:48:07.537404+00:00"} {"global_step": 11214, "acc_step": 0, "speed/wps": 12964.1844012256, "speed/FLOPS": 203620330677457.9, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.9364594221115112, "optim/lr": 0.0029065293961795615, "optim/total_tokens": 2939682816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 4.043383598327637, "created_at": "2025-01-15T04:48:12.597491+00:00"} {"global_step": 11215, "acc_step": 0, "speed/wps": 12960.122175276665, "speed/FLOPS": 203556527836846.44, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3113056421279907, "optim/lr": 0.002906499621622309, "optim/total_tokens": 2939944960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434515, "loss/out": 4.185530185699463, "created_at": "2025-01-15T04:48:17.661694+00:00"} {"global_step": 11216, "acc_step": 0, "speed/wps": 12955.72249147097, "speed/FLOPS": 203487424756878.9, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.932220458984375, "optim/lr": 0.0029064698424761046, "optim/total_tokens": 2940207104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302403, "loss/out": 4.104267120361328, "created_at": "2025-01-15T04:48:22.728726+00:00"} {"global_step": 11217, "acc_step": 0, "speed/wps": 12966.441403835363, "speed/FLOPS": 203655780004890.8, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1477872133255005, "optim/lr": 0.002906440058741045, "optim/total_tokens": 2940469248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449657, "loss/out": 4.029137134552002, "created_at": "2025-01-15T04:48:27.788654+00:00"} {"global_step": 11218, "acc_step": 0, "speed/wps": 12965.27102968397, "speed/FLOPS": 203637397670580.56, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.036747694015503, "optim/lr": 0.002906410270417228, "optim/total_tokens": 2940731392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466392, "loss/out": 4.197723388671875, "created_at": "2025-01-15T04:48:32.847073+00:00"} {"global_step": 11219, "acc_step": 0, "speed/wps": 12958.058763151183, "speed/FLOPS": 203524119113990.34, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.001, "optim/grad_norm": 0.8461589217185974, "optim/lr": 0.00290638047750475, "optim/total_tokens": 2940993536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 4.063875198364258, "created_at": "2025-01-15T04:48:37.908603+00:00"} {"global_step": 11220, "acc_step": 0, "speed/wps": 12961.669902453375, "speed/FLOPS": 203580837018956.7, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2901073694229126, "optim/lr": 0.002906350680003709, "optim/total_tokens": 2941255680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433481, "loss/out": 3.957364559173584, "created_at": "2025-01-15T04:48:42.968853+00:00"} {"global_step": 11221, "acc_step": 0, "speed/wps": 12962.773681945731, "speed/FLOPS": 203598173392636.12, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.187788963317871, "optim/lr": 0.002906320877914202, "optim/total_tokens": 2941517824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444644, "loss/out": 4.077150344848633, "created_at": "2025-01-15T04:48:48.028078+00:00"} {"global_step": 11222, "acc_step": 0, "speed/wps": 12962.57218932718, "speed/FLOPS": 203595008674181.9, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.773921012878418, "optim/lr": 0.002906291071236326, "optim/total_tokens": 2941779968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 4.105896949768066, "created_at": "2025-01-15T04:48:53.093449+00:00"} {"global_step": 11223, "acc_step": 0, "speed/wps": 12961.490272354305, "speed/FLOPS": 203578015681413.2, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2169005870819092, "optim/lr": 0.002906261259970178, "optim/total_tokens": 2942042112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480995, "loss/out": 4.023276329040527, "created_at": "2025-01-15T04:48:58.157386+00:00"} {"global_step": 11224, "acc_step": 0, "speed/wps": 12954.753477542876, "speed/FLOPS": 203472205061573.5, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6535075306892395, "optim/lr": 0.0029062314441158556, "optim/total_tokens": 2942304256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327786, "loss/out": 4.047978401184082, "created_at": "2025-01-15T04:49:03.220162+00:00"} {"global_step": 11225, "acc_step": 0, "speed/wps": 12964.288023631636, "speed/FLOPS": 203621958209736.78, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5969703197479248, "optim/lr": 0.0029062016236734564, "optim/total_tokens": 2942566400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 312776, "loss/out": 4.087705612182617, "created_at": "2025-01-15T04:49:08.280221+00:00"} {"global_step": 11226, "acc_step": 0, "speed/wps": 12960.362729057359, "speed/FLOPS": 203560306064527.78, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0174542665481567, "optim/lr": 0.0029061717986430774, "optim/total_tokens": 2942828544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 398201, "loss/out": 4.088520050048828, "created_at": "2025-01-15T04:49:13.342198+00:00"} {"global_step": 11227, "acc_step": 0, "speed/wps": 12959.747269755555, "speed/FLOPS": 203550639430463.25, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5074268579483032, "optim/lr": 0.0029061419690248162, "optim/total_tokens": 2943090688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 4.096128463745117, "created_at": "2025-01-15T04:49:18.402757+00:00"} {"global_step": 11228, "acc_step": 0, "speed/wps": 12950.67073106517, "speed/FLOPS": 203408079917859.56, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9308992028236389, "optim/lr": 0.002906112134818769, "optim/total_tokens": 2943352832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 3.990468740463257, "created_at": "2025-01-15T04:49:23.466951+00:00"} {"global_step": 11229, "acc_step": 0, "speed/wps": 12956.463089146915, "speed/FLOPS": 203499056861067.47, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0118463039398193, "optim/lr": 0.0029060822960250347, "optim/total_tokens": 2943614976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453660, "loss/out": 3.924558162689209, "created_at": "2025-01-15T04:49:28.531695+00:00"} {"global_step": 11230, "acc_step": 0, "speed/wps": 12962.27525093293, "speed/FLOPS": 203590344848664.06, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6544676423072815, "optim/lr": 0.0029060524526437095, "optim/total_tokens": 2943877120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458952, "loss/out": 4.045254707336426, "created_at": "2025-01-15T04:49:33.595380+00:00"} {"global_step": 11231, "acc_step": 0, "speed/wps": 12967.446092495058, "speed/FLOPS": 203671560020878.22, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0183639526367188, "optim/lr": 0.002906022604674891, "optim/total_tokens": 2944139264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466155, "loss/out": 3.9914848804473877, "created_at": "2025-01-15T04:49:38.653935+00:00"} {"global_step": 11232, "acc_step": 0, "speed/wps": 12959.805168640003, "speed/FLOPS": 203551548812004.9, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0753575563430786, "optim/lr": 0.0029059927521186773, "optim/total_tokens": 2944401408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325000, "loss/out": 4.113066673278809, "created_at": "2025-01-15T04:49:43.717690+00:00"} {"global_step": 11233, "acc_step": 0, "speed/wps": 12961.774334627657, "speed/FLOPS": 203582477269758.62, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0058742761611938, "optim/lr": 0.002905962894975165, "optim/total_tokens": 2944663552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372786, "loss/out": 4.173713207244873, "created_at": "2025-01-15T04:49:48.782124+00:00"} {"global_step": 11234, "acc_step": 0, "speed/wps": 12956.554168665343, "speed/FLOPS": 203500487390060.75, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5500620007514954, "optim/lr": 0.0029059330332444518, "optim/total_tokens": 2944925696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432941, "loss/out": 4.108024597167969, "created_at": "2025-01-15T04:49:53.849751+00:00"} {"global_step": 11235, "acc_step": 0, "speed/wps": 12958.576721987809, "speed/FLOPS": 203532254369269.6, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6925747990608215, "optim/lr": 0.0029059031669266353, "optim/total_tokens": 2945187840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 477279, "loss/out": 3.998106002807617, "created_at": "2025-01-15T04:49:58.911344+00:00"} {"global_step": 11236, "acc_step": 0, "speed/wps": 12965.045892100306, "speed/FLOPS": 203633861575457.66, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2099841833114624, "optim/lr": 0.0029058732960218128, "optim/total_tokens": 2945449984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494987, "loss/out": 4.075347900390625, "created_at": "2025-01-15T04:50:03.974171+00:00"} {"global_step": 11237, "acc_step": 0, "speed/wps": 12964.55478639952, "speed/FLOPS": 203626148085576.88, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1535600423812866, "optim/lr": 0.0029058434205300816, "optim/total_tokens": 2945712128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441442, "loss/out": 4.0265793800354, "created_at": "2025-01-15T04:50:09.037894+00:00"} {"global_step": 11238, "acc_step": 0, "speed/wps": 12955.142750545896, "speed/FLOPS": 203478319129003.1, "speed/curr_iter_time": 1.2772, "speed/data_load_time": 0.001, "optim/grad_norm": 0.5257080793380737, "optim/lr": 0.002905813540451539, "optim/total_tokens": 2945974272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492324, "loss/out": 3.9329493045806885, "created_at": "2025-01-15T04:50:14.106276+00:00"} {"global_step": 11239, "acc_step": 0, "speed/wps": 12954.574067713289, "speed/FLOPS": 203469387183664.9, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0739389657974243, "optim/lr": 0.002905783655786283, "optim/total_tokens": 2946236416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 334369, "loss/out": 4.078521728515625, "created_at": "2025-01-15T04:50:19.170693+00:00"} {"global_step": 11240, "acc_step": 0, "speed/wps": 12962.81961097086, "speed/FLOPS": 203598894771088.88, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6846587061882019, "optim/lr": 0.0029057537665344107, "optim/total_tokens": 2946498560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331435, "loss/out": 4.036810874938965, "created_at": "2025-01-15T04:50:24.230210+00:00"} {"global_step": 11241, "acc_step": 0, "speed/wps": 12951.913603594025, "speed/FLOPS": 203427600938811.06, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0102850198745728, "optim/lr": 0.0029057238726960203, "optim/total_tokens": 2946760704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445976, "loss/out": 4.142113208770752, "created_at": "2025-01-15T04:50:29.316646+00:00"} {"global_step": 11242, "acc_step": 0, "speed/wps": 12959.59460086405, "speed/FLOPS": 203548241555733.1, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4876675307750702, "optim/lr": 0.002905693974271208, "optim/total_tokens": 2947022848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427880, "loss/out": 4.058871269226074, "created_at": "2025-01-15T04:50:34.382529+00:00"} {"global_step": 11243, "acc_step": 0, "speed/wps": 12959.66076370454, "speed/FLOPS": 203549280734060.28, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1599825620651245, "optim/lr": 0.0029056640712600734, "optim/total_tokens": 2947284992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 4.053186893463135, "created_at": "2025-01-15T04:50:39.444290+00:00"} {"global_step": 11244, "acc_step": 0, "speed/wps": 12964.836900101427, "speed/FLOPS": 203630579068930.16, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4954845607280731, "optim/lr": 0.002905634163662712, "optim/total_tokens": 2947547136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435582, "loss/out": 3.9471354484558105, "created_at": "2025-01-15T04:50:44.510779+00:00"} {"global_step": 11245, "acc_step": 0, "speed/wps": 12960.367939905573, "speed/FLOPS": 203560387908059.44, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.891536295413971, "optim/lr": 0.002905604251479222, "optim/total_tokens": 2947809280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 362577, "loss/out": 4.040505409240723, "created_at": "2025-01-15T04:50:49.572950+00:00"} {"global_step": 11246, "acc_step": 0, "speed/wps": 12963.724009124002, "speed/FLOPS": 203613099586857.25, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0340206623077393, "optim/lr": 0.002905574334709702, "optim/total_tokens": 2948071424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455247, "loss/out": 4.1117095947265625, "created_at": "2025-01-15T04:50:54.636881+00:00"} {"global_step": 11247, "acc_step": 0, "speed/wps": 12967.707794776334, "speed/FLOPS": 203675670414822.28, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8553744554519653, "optim/lr": 0.0029055444133542483, "optim/total_tokens": 2948333568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488142, "loss/out": 4.079819679260254, "created_at": "2025-01-15T04:50:59.696131+00:00"} {"global_step": 11248, "acc_step": 0, "speed/wps": 12965.712897583802, "speed/FLOPS": 203644337813137.56, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.3149383068084717, "optim/lr": 0.002905514487412959, "optim/total_tokens": 2948595712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481474, "loss/out": 4.012873649597168, "created_at": "2025-01-15T04:51:04.754265+00:00"} {"global_step": 11249, "acc_step": 0, "speed/wps": 12957.275867455264, "speed/FLOPS": 203511822661274.53, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0286865234375, "optim/lr": 0.002905484556885932, "optim/total_tokens": 2948857856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463459, "loss/out": 4.021052837371826, "created_at": "2025-01-15T04:51:09.816348+00:00"} {"global_step": 11250, "acc_step": 0, "speed/wps": 12957.638797774569, "speed/FLOPS": 203517522980657.75, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7750735282897949, "optim/lr": 0.002905454621773265, "optim/total_tokens": 2949120000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424963, "loss/out": 4.0425214767456055, "created_at": "2025-01-15T04:51:14.880674+00:00"} {"global_step": 11251, "acc_step": 0, "speed/wps": 12959.35286795657, "speed/FLOPS": 203544444808245.3, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2044607400894165, "optim/lr": 0.0029054246820750555, "optim/total_tokens": 2949382144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 4.057607173919678, "created_at": "2025-01-15T04:51:19.955117+00:00"} {"global_step": 11252, "acc_step": 0, "speed/wps": 12956.465706123514, "speed/FLOPS": 203499097964281.06, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9226540923118591, "optim/lr": 0.0029053947377914004, "optim/total_tokens": 2949644288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356143, "loss/out": 4.108019828796387, "created_at": "2025-01-15T04:51:25.026977+00:00"} {"global_step": 11253, "acc_step": 0, "speed/wps": 12963.59978788851, "speed/FLOPS": 203611148521655.97, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9341456890106201, "optim/lr": 0.002905364788922398, "optim/total_tokens": 2949906432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447839, "loss/out": 3.9399802684783936, "created_at": "2025-01-15T04:51:30.088434+00:00"} {"global_step": 11254, "acc_step": 0, "speed/wps": 12961.793776325887, "speed/FLOPS": 203582782628345.03, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.4269967079162598, "optim/lr": 0.0029053348354681472, "optim/total_tokens": 2950168576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 4.097914218902588, "created_at": "2025-01-15T04:51:35.154643+00:00"} {"global_step": 11255, "acc_step": 0, "speed/wps": 12962.76756892544, "speed/FLOPS": 203598077379252.25, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8731500506401062, "optim/lr": 0.002905304877428744, "optim/total_tokens": 2950430720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424393, "loss/out": 4.130742073059082, "created_at": "2025-01-15T04:51:40.214369+00:00"} {"global_step": 11256, "acc_step": 0, "speed/wps": 12961.880360994144, "speed/FLOPS": 203584142559539.8, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8277580738067627, "optim/lr": 0.002905274914804287, "optim/total_tokens": 2950692864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485239, "loss/out": 4.052567481994629, "created_at": "2025-01-15T04:51:45.278117+00:00"} {"global_step": 11257, "acc_step": 0, "speed/wps": 12961.382636108094, "speed/FLOPS": 203576325106258.72, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6639831066131592, "optim/lr": 0.002905244947594873, "optim/total_tokens": 2950955008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435852, "loss/out": 4.146582126617432, "created_at": "2025-01-15T04:51:50.338309+00:00"} {"global_step": 11258, "acc_step": 0, "speed/wps": 12959.963489811535, "speed/FLOPS": 203554035463559.9, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5688026547431946, "optim/lr": 0.0029052149758006015, "optim/total_tokens": 2951217152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300225, "loss/out": 3.9686496257781982, "created_at": "2025-01-15T04:51:55.402320+00:00"} {"global_step": 11259, "acc_step": 0, "speed/wps": 12963.736236974584, "speed/FLOPS": 203613291642053.47, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.3272571563720703, "optim/lr": 0.002905184999421569, "optim/total_tokens": 2951479296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420893, "loss/out": 4.135461807250977, "created_at": "2025-01-15T04:52:00.465734+00:00"} {"global_step": 11260, "acc_step": 0, "speed/wps": 12961.846252250767, "speed/FLOPS": 203583606834854.75, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.8962822556495667, "optim/lr": 0.0029051550184578732, "optim/total_tokens": 2951741440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457380, "loss/out": 4.078234672546387, "created_at": "2025-01-15T04:52:05.531421+00:00"} {"global_step": 11261, "acc_step": 0, "speed/wps": 12961.72684441243, "speed/FLOPS": 203581731370670.1, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2039895057678223, "optim/lr": 0.0029051250329096123, "optim/total_tokens": 2952003584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 470369, "loss/out": 4.185051918029785, "created_at": "2025-01-15T04:52:10.592551+00:00"} {"global_step": 11262, "acc_step": 0, "speed/wps": 12958.618406947742, "speed/FLOPS": 203532909088846.5, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7310119271278381, "optim/lr": 0.002905095042776885, "optim/total_tokens": 2952265728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464799, "loss/out": 4.078229904174805, "created_at": "2025-01-15T04:52:15.653941+00:00"} {"global_step": 11263, "acc_step": 0, "speed/wps": 12958.894976315027, "speed/FLOPS": 203537252990805.84, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2399028539657593, "optim/lr": 0.0029050650480597876, "optim/total_tokens": 2952527872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430506, "loss/out": 4.051347732543945, "created_at": "2025-01-15T04:52:20.715612+00:00"} {"global_step": 11264, "acc_step": 0, "speed/wps": 12963.00963221969, "speed/FLOPS": 203601879315918.28, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0348149538040161, "optim/lr": 0.0029050350487584186, "optim/total_tokens": 2952790016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494704, "loss/out": 4.085470199584961, "created_at": "2025-01-15T04:52:25.779655+00:00"} {"global_step": 11265, "acc_step": 0, "speed/wps": 12965.725566371324, "speed/FLOPS": 203644536793854.3, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0069257020950317, "optim/lr": 0.0029050050448728762, "optim/total_tokens": 2953052160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435338, "loss/out": 4.005176544189453, "created_at": "2025-01-15T04:52:30.839226+00:00"} {"global_step": 11266, "acc_step": 0, "speed/wps": 12965.41121192603, "speed/FLOPS": 203639599425322.7, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0352351665496826, "optim/lr": 0.002904975036403258, "optim/total_tokens": 2953314304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 282898, "loss/out": 4.02706241607666, "created_at": "2025-01-15T04:52:35.899976+00:00"} {"global_step": 11267, "acc_step": 0, "speed/wps": 12957.00322021727, "speed/FLOPS": 203507540361745.25, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.8309926986694336, "optim/lr": 0.0029049450233496624, "optim/total_tokens": 2953576448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 409163, "loss/out": 4.054277420043945, "created_at": "2025-01-15T04:52:40.961556+00:00"} {"global_step": 11268, "acc_step": 0, "speed/wps": 12963.691794102318, "speed/FLOPS": 203612593605673.62, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0324424505233765, "optim/lr": 0.0029049150057121865, "optim/total_tokens": 2953838592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452081, "loss/out": 4.130216121673584, "created_at": "2025-01-15T04:52:46.020562+00:00"} {"global_step": 11269, "acc_step": 0, "speed/wps": 12962.648762565834, "speed/FLOPS": 203596211362117.6, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5821780562400818, "optim/lr": 0.002904884983490929, "optim/total_tokens": 2954100736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 4.098094463348389, "created_at": "2025-01-15T04:52:51.085095+00:00"} {"global_step": 11270, "acc_step": 0, "speed/wps": 12959.608926024766, "speed/FLOPS": 203548466552067.0, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7480976581573486, "optim/lr": 0.0029048549566859866, "optim/total_tokens": 2954362880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460815, "loss/out": 4.047883033752441, "created_at": "2025-01-15T04:52:56.151236+00:00"} {"global_step": 11271, "acc_step": 0, "speed/wps": 12963.92764883144, "speed/FLOPS": 203616298028292.84, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1556614637374878, "optim/lr": 0.0029048249252974583, "optim/total_tokens": 2954625024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457095, "loss/out": 4.127009868621826, "created_at": "2025-01-15T04:53:01.216598+00:00"} {"global_step": 11272, "acc_step": 0, "speed/wps": 12960.318820129452, "speed/FLOPS": 203559616414477.8, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2666882276535034, "optim/lr": 0.0029047948893254423, "optim/total_tokens": 2954887168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305274, "loss/out": 4.000246047973633, "created_at": "2025-01-15T04:53:06.276854+00:00"} {"global_step": 11273, "acc_step": 0, "speed/wps": 12957.925979318616, "speed/FLOPS": 203522033561435.78, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1081771850585938, "optim/lr": 0.0029047648487700362, "optim/total_tokens": 2955149312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368425, "loss/out": 4.069062232971191, "created_at": "2025-01-15T04:53:11.341951+00:00"} {"global_step": 11274, "acc_step": 0, "speed/wps": 12956.037530810512, "speed/FLOPS": 203492372882615.84, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0436135530471802, "optim/lr": 0.002904734803631338, "optim/total_tokens": 2955411456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437717, "loss/out": 4.066947937011719, "created_at": "2025-01-15T04:53:16.404952+00:00"} {"global_step": 11275, "acc_step": 0, "speed/wps": 12963.232642961588, "speed/FLOPS": 203605382006068.66, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9845547676086426, "optim/lr": 0.002904704753909446, "optim/total_tokens": 2955673600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 438773, "loss/out": 4.01453971862793, "created_at": "2025-01-15T04:53:21.464118+00:00"} {"global_step": 11276, "acc_step": 0, "speed/wps": 12959.127219891034, "speed/FLOPS": 203540900695303.12, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8639311194419861, "optim/lr": 0.002904674699604457, "optim/total_tokens": 2955935744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462696, "loss/out": 4.039161205291748, "created_at": "2025-01-15T04:53:26.527199+00:00"} {"global_step": 11277, "acc_step": 0, "speed/wps": 12955.272195394105, "speed/FLOPS": 203480352238220.1, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.025650143623352, "optim/lr": 0.002904644640716471, "optim/total_tokens": 2956197888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 4.104280948638916, "created_at": "2025-01-15T04:53:31.589728+00:00"} {"global_step": 11278, "acc_step": 0, "speed/wps": 12960.097012384327, "speed/FLOPS": 203556132619044.84, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8408915400505066, "optim/lr": 0.002904614577245585, "optim/total_tokens": 2956460032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440390, "loss/out": 4.026843070983887, "created_at": "2025-01-15T04:53:36.650183+00:00"} {"global_step": 11279, "acc_step": 0, "speed/wps": 12959.140952151236, "speed/FLOPS": 203541116379319.72, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.001, "optim/grad_norm": 0.7109752297401428, "optim/lr": 0.002904584509191897, "optim/total_tokens": 2956722176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 4.1125359535217285, "created_at": "2025-01-15T04:53:41.712912+00:00"} {"global_step": 11280, "acc_step": 0, "speed/wps": 12957.77537671163, "speed/FLOPS": 203519668140543.4, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0460779666900635, "optim/lr": 0.0029045544365555055, "optim/total_tokens": 2956984320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446220, "loss/out": 4.140064239501953, "created_at": "2025-01-15T04:53:46.774557+00:00"} {"global_step": 11281, "acc_step": 0, "speed/wps": 12960.492512172596, "speed/FLOPS": 203562344486693.25, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0141196250915527, "optim/lr": 0.002904524359336508, "optim/total_tokens": 2957246464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301637, "loss/out": 4.041719913482666, "created_at": "2025-01-15T04:53:51.847034+00:00"} {"global_step": 11282, "acc_step": 0, "speed/wps": 12960.337462589816, "speed/FLOPS": 203559909219935.25, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0300294160842896, "optim/lr": 0.0029044942775350033, "optim/total_tokens": 2957508608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440399, "loss/out": 4.222064018249512, "created_at": "2025-01-15T04:53:56.912275+00:00"} {"global_step": 11283, "acc_step": 0, "speed/wps": 12962.634712450159, "speed/FLOPS": 203595990685744.06, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6674551367759705, "optim/lr": 0.00290446419115109, "optim/total_tokens": 2957770752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 4.09316349029541, "created_at": "2025-01-15T04:54:01.975157+00:00"} {"global_step": 11284, "acc_step": 0, "speed/wps": 12962.438536951697, "speed/FLOPS": 203592909479967.3, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9161348342895508, "optim/lr": 0.0029044341001848648, "optim/total_tokens": 2958032896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435044, "loss/out": 4.099326133728027, "created_at": "2025-01-15T04:54:07.037332+00:00"} {"global_step": 11285, "acc_step": 0, "speed/wps": 12965.821480144845, "speed/FLOPS": 203646043251466.8, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4651021659374237, "optim/lr": 0.002904404004636427, "optim/total_tokens": 2958295040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 3.97153639793396, "created_at": "2025-01-15T04:54:12.098178+00:00"} {"global_step": 11286, "acc_step": 0, "speed/wps": 12958.384620979112, "speed/FLOPS": 203529237158953.56, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.44003361463546753, "optim/lr": 0.002904373904505874, "optim/total_tokens": 2958557184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 432365, "loss/out": 4.2014312744140625, "created_at": "2025-01-15T04:54:17.164882+00:00"} {"global_step": 11287, "acc_step": 0, "speed/wps": 12965.467183427048, "speed/FLOPS": 203640478534659.1, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8018873333930969, "optim/lr": 0.0029043437997933047, "optim/total_tokens": 2958819328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317586, "loss/out": 3.9946517944335938, "created_at": "2025-01-15T04:54:22.227217+00:00"} {"global_step": 11288, "acc_step": 0, "speed/wps": 12964.466190688605, "speed/FLOPS": 203624756568193.7, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0574430227279663, "optim/lr": 0.002904313690498817, "optim/total_tokens": 2959081472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355742, "loss/out": 4.128146171569824, "created_at": "2025-01-15T04:54:27.290977+00:00"} {"global_step": 11289, "acc_step": 0, "speed/wps": 12960.72121319982, "speed/FLOPS": 203565936550593.97, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7836428880691528, "optim/lr": 0.0029042835766225094, "optim/total_tokens": 2959343616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446503, "loss/out": 4.111227035522461, "created_at": "2025-01-15T04:54:32.354965+00:00"} {"global_step": 11290, "acc_step": 0, "speed/wps": 12960.031021719227, "speed/FLOPS": 203555096144968.78, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0336644649505615, "optim/lr": 0.0029042534581644797, "optim/total_tokens": 2959605760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 4.077532768249512, "created_at": "2025-01-15T04:54:37.416551+00:00"} {"global_step": 11291, "acc_step": 0, "speed/wps": 12965.576612332501, "speed/FLOPS": 203642197265990.3, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.01179039478302, "optim/lr": 0.0029042233351248264, "optim/total_tokens": 2959867904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472752, "loss/out": 4.0745439529418945, "created_at": "2025-01-15T04:54:42.476075+00:00"} {"global_step": 11292, "acc_step": 0, "speed/wps": 12958.678431565713, "speed/FLOPS": 203533851857954.2, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0314117670059204, "optim/lr": 0.0029041932075036486, "optim/total_tokens": 2960130048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418281, "loss/out": 4.122658729553223, "created_at": "2025-01-15T04:54:47.541154+00:00"} {"global_step": 11293, "acc_step": 0, "speed/wps": 12960.405614247993, "speed/FLOPS": 203560979635377.12, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5661777853965759, "optim/lr": 0.002904163075301043, "optim/total_tokens": 2960392192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490228, "loss/out": 4.1323723793029785, "created_at": "2025-01-15T04:54:52.602908+00:00"} {"global_step": 11294, "acc_step": 0, "speed/wps": 12958.704667602875, "speed/FLOPS": 203534263930969.9, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7695846557617188, "optim/lr": 0.002904132938517109, "optim/total_tokens": 2960654336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436679, "loss/out": 4.129547119140625, "created_at": "2025-01-15T04:54:57.669097+00:00"} {"global_step": 11295, "acc_step": 0, "speed/wps": 12965.336765561879, "speed/FLOPS": 203638430142873.7, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9023157954216003, "optim/lr": 0.0029041027971519443, "optim/total_tokens": 2960916480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300997, "loss/out": 4.102077960968018, "created_at": "2025-01-15T04:55:02.731852+00:00"} {"global_step": 11296, "acc_step": 0, "speed/wps": 12963.509588437551, "speed/FLOPS": 203609731815331.62, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7169055342674255, "optim/lr": 0.0029040726512056477, "optim/total_tokens": 2961178624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463795, "loss/out": 4.12905216217041, "created_at": "2025-01-15T04:55:07.792037+00:00"} {"global_step": 11297, "acc_step": 0, "speed/wps": 12958.55494893354, "speed/FLOPS": 203531912393533.0, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5492441058158875, "optim/lr": 0.0029040425006783177, "optim/total_tokens": 2961440768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461385, "loss/out": 4.152253150939941, "created_at": "2025-01-15T04:55:12.854121+00:00"} {"global_step": 11298, "acc_step": 0, "speed/wps": 12961.32777656959, "speed/FLOPS": 203575463461821.6, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7319162487983704, "optim/lr": 0.0029040123455700523, "optim/total_tokens": 2961702912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466938, "loss/out": 4.006809234619141, "created_at": "2025-01-15T04:55:17.914475+00:00"} {"global_step": 11299, "acc_step": 0, "speed/wps": 12963.092252306935, "speed/FLOPS": 203603176977920.38, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5267865657806396, "optim/lr": 0.0029039821858809502, "optim/total_tokens": 2961965056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418803, "loss/out": 4.105650901794434, "created_at": "2025-01-15T04:55:22.973862+00:00"} {"global_step": 11300, "acc_step": 0, "speed/wps": 12957.35086308498, "speed/FLOPS": 203513000570694.0, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7975691556930542, "optim/lr": 0.0029039520216111096, "optim/total_tokens": 2962227200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491567, "loss/out": 3.985476493835449, "created_at": "2025-01-15T04:55:28.035838+00:00"} {"global_step": 11301, "acc_step": 0, "speed/wps": 12961.923280104784, "speed/FLOPS": 203584816663149.44, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.8845890760421753, "optim/lr": 0.0029039218527606285, "optim/total_tokens": 2962489344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 4.066080570220947, "created_at": "2025-01-15T04:55:33.100665+00:00"} {"global_step": 11302, "acc_step": 0, "speed/wps": 12965.260404221988, "speed/FLOPS": 203637230783099.22, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7375388741493225, "optim/lr": 0.0029038916793296065, "optim/total_tokens": 2962751488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 4.008221626281738, "created_at": "2025-01-15T04:55:38.163987+00:00"} {"global_step": 11303, "acc_step": 0, "speed/wps": 12964.927301953505, "speed/FLOPS": 203631998954242.22, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7249080538749695, "optim/lr": 0.00290386150131814, "optim/total_tokens": 2963013632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429977, "loss/out": 4.100640296936035, "created_at": "2025-01-15T04:55:43.223884+00:00"} {"global_step": 11304, "acc_step": 0, "speed/wps": 12957.777151939425, "speed/FLOPS": 203519696022935.34, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9035711884498596, "optim/lr": 0.00290383131872633, "optim/total_tokens": 2963275776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 339985, "loss/out": 4.044527053833008, "created_at": "2025-01-15T04:55:48.289632+00:00"} {"global_step": 11305, "acc_step": 0, "speed/wps": 12961.2392536058, "speed/FLOPS": 203574073087032.0, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0422028303146362, "optim/lr": 0.0029038011315542732, "optim/total_tokens": 2963537920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449696, "loss/out": 4.169498443603516, "created_at": "2025-01-15T04:55:53.355940+00:00"} {"global_step": 11306, "acc_step": 0, "speed/wps": 12956.792636125143, "speed/FLOPS": 203504232849205.53, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0935765504837036, "optim/lr": 0.0029037709398020686, "optim/total_tokens": 2963800064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 4.171456336975098, "created_at": "2025-01-15T04:55:58.417792+00:00"} {"global_step": 11307, "acc_step": 0, "speed/wps": 12963.715452030568, "speed/FLOPS": 203612965185946.2, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5285466313362122, "optim/lr": 0.002903740743469815, "optim/total_tokens": 2964062208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 4.1536054611206055, "created_at": "2025-01-15T04:56:03.480426+00:00"} {"global_step": 11308, "acc_step": 0, "speed/wps": 12964.450619924048, "speed/FLOPS": 203624512007939.84, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0239461660385132, "optim/lr": 0.0029037105425576103, "optim/total_tokens": 2964324352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442778, "loss/out": 4.047885894775391, "created_at": "2025-01-15T04:56:08.544968+00:00"} {"global_step": 11309, "acc_step": 0, "speed/wps": 12961.070881873155, "speed/FLOPS": 203571428577603.78, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0584596395492554, "optim/lr": 0.0029036803370655535, "optim/total_tokens": 2964586496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479377, "loss/out": 4.032601356506348, "created_at": "2025-01-15T04:56:13.606938+00:00"} {"global_step": 11310, "acc_step": 0, "speed/wps": 12956.836394720403, "speed/FLOPS": 203504920138074.72, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0898326635360718, "optim/lr": 0.002903650126993743, "optim/total_tokens": 2964848640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439820, "loss/out": 3.966993808746338, "created_at": "2025-01-15T04:56:18.674966+00:00"} {"global_step": 11311, "acc_step": 0, "speed/wps": 12962.650252586685, "speed/FLOPS": 203596234764942.6, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0752514600753784, "optim/lr": 0.002903619912342278, "optim/total_tokens": 2965110784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 363102, "loss/out": 4.001962184906006, "created_at": "2025-01-15T04:56:23.738536+00:00"} {"global_step": 11312, "acc_step": 0, "speed/wps": 12959.839011252505, "speed/FLOPS": 203552080356738.88, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8975521922111511, "optim/lr": 0.0029035896931112557, "optim/total_tokens": 2965372928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445695, "loss/out": 3.9971134662628174, "created_at": "2025-01-15T04:56:28.810141+00:00"} {"global_step": 11313, "acc_step": 0, "speed/wps": 12959.25882072164, "speed/FLOPS": 203542967667186.28, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0490525960922241, "optim/lr": 0.002903559469300776, "optim/total_tokens": 2965635072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442778, "loss/out": 4.160852432250977, "created_at": "2025-01-15T04:56:33.881848+00:00"} {"global_step": 11314, "acc_step": 0, "speed/wps": 12961.877713178925, "speed/FLOPS": 203584100971963.3, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4551663398742676, "optim/lr": 0.0029035292409109368, "optim/total_tokens": 2965897216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482285, "loss/out": 3.988438367843628, "created_at": "2025-01-15T04:56:38.941881+00:00"} {"global_step": 11315, "acc_step": 0, "speed/wps": 12956.930895876922, "speed/FLOPS": 203506404408596.03, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 2.8703129291534424, "optim/lr": 0.0029034990079418368, "optim/total_tokens": 2966159360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485185, "loss/out": 4.1100687980651855, "created_at": "2025-01-15T04:56:44.008008+00:00"} {"global_step": 11316, "acc_step": 0, "speed/wps": 12956.749035280365, "speed/FLOPS": 203503548038024.47, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.9819633960723877, "optim/lr": 0.002903468770393575, "optim/total_tokens": 2966421504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435852, "loss/out": 4.080867290496826, "created_at": "2025-01-15T04:56:49.074501+00:00"} {"global_step": 11317, "acc_step": 0, "speed/wps": 12956.464112558788, "speed/FLOPS": 203499072935157.25, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6294395923614502, "optim/lr": 0.0029034385282662497, "optim/total_tokens": 2966683648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324035, "loss/out": 4.123631954193115, "created_at": "2025-01-15T04:56:54.136581+00:00"} {"global_step": 11318, "acc_step": 0, "speed/wps": 12966.39757794972, "speed/FLOPS": 203655091659133.7, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0059101581573486, "optim/lr": 0.0029034082815599595, "optim/total_tokens": 2966945792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460271, "loss/out": 4.012732982635498, "created_at": "2025-01-15T04:56:59.199649+00:00"} {"global_step": 11319, "acc_step": 0, "speed/wps": 12958.96365453805, "speed/FLOPS": 203538331676672.2, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4466562867164612, "optim/lr": 0.002903378030274804, "optim/total_tokens": 2967207936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489140, "loss/out": 3.9946794509887695, "created_at": "2025-01-15T04:57:04.264899+00:00"} {"global_step": 11320, "acc_step": 0, "speed/wps": 12951.964843348453, "speed/FLOPS": 203428405729565.16, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.001, "optim/grad_norm": 0.8638623952865601, "optim/lr": 0.00290334777441088, "optim/total_tokens": 2967470080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450746, "loss/out": 4.082441806793213, "created_at": "2025-01-15T04:57:09.333496+00:00"} {"global_step": 11321, "acc_step": 0, "speed/wps": 12962.036242576009, "speed/FLOPS": 203586590893986.72, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.0971828699111938, "optim/lr": 0.002903317513968288, "optim/total_tokens": 2967732224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480954, "loss/out": 4.0803680419921875, "created_at": "2025-01-15T04:57:14.401018+00:00"} {"global_step": 11322, "acc_step": 0, "speed/wps": 12956.306287182197, "speed/FLOPS": 203496594070743.1, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7243994474411011, "optim/lr": 0.002903287248947126, "optim/total_tokens": 2967994368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326870, "loss/out": 4.143011093139648, "created_at": "2025-01-15T04:57:19.463171+00:00"} {"global_step": 11323, "acc_step": 0, "speed/wps": 12959.741720483262, "speed/FLOPS": 203550552271516.38, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5256423354148865, "optim/lr": 0.002903256979347493, "optim/total_tokens": 2968256512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 368286, "loss/out": 4.19028902053833, "created_at": "2025-01-15T04:57:24.523731+00:00"} {"global_step": 11324, "acc_step": 0, "speed/wps": 12962.003801955634, "speed/FLOPS": 203586081369465.66, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1344573497772217, "optim/lr": 0.0029032267051694875, "optim/total_tokens": 2968518656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431076, "loss/out": 4.053946495056152, "created_at": "2025-01-15T04:57:29.588368+00:00"} {"global_step": 11325, "acc_step": 0, "speed/wps": 12959.538496278425, "speed/FLOPS": 203547360356120.1, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.8305386304855347, "optim/lr": 0.002903196426413209, "optim/total_tokens": 2968780800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469036, "loss/out": 3.960496187210083, "created_at": "2025-01-15T04:57:34.651609+00:00"} {"global_step": 11326, "acc_step": 0, "speed/wps": 12955.42170941665, "speed/FLOPS": 203482700561397.97, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.553533673286438, "optim/lr": 0.002903166143078755, "optim/total_tokens": 2969042944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463248, "loss/out": 4.025176048278809, "created_at": "2025-01-15T04:57:39.720415+00:00"} {"global_step": 11327, "acc_step": 0, "speed/wps": 12962.565586993192, "speed/FLOPS": 203594904975454.06, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7968286871910095, "optim/lr": 0.002903135855166225, "optim/total_tokens": 2969305088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452325, "loss/out": 4.050344467163086, "created_at": "2025-01-15T04:57:44.785188+00:00"} {"global_step": 11328, "acc_step": 0, "speed/wps": 12960.326883404703, "speed/FLOPS": 203559743059295.72, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8577247262001038, "optim/lr": 0.0029031055626757177, "optim/total_tokens": 2969567232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301104, "loss/out": 4.1380228996276855, "created_at": "2025-01-15T04:57:49.853088+00:00"} {"global_step": 11329, "acc_step": 0, "speed/wps": 12962.23461213322, "speed/FLOPS": 203589706560471.06, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0352598428726196, "optim/lr": 0.0029030752656073323, "optim/total_tokens": 2969829376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 353101, "loss/out": 4.062863349914551, "created_at": "2025-01-15T04:57:54.917858+00:00"} {"global_step": 11330, "acc_step": 0, "speed/wps": 12958.545401171641, "speed/FLOPS": 203531762432812.38, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.47361159324646, "optim/lr": 0.0029030449639611674, "optim/total_tokens": 2970091520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 3.9781882762908936, "created_at": "2025-01-15T04:57:59.979913+00:00"} {"global_step": 11331, "acc_step": 0, "speed/wps": 12958.957542724222, "speed/FLOPS": 203538235682237.5, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5983029007911682, "optim/lr": 0.0029030146577373223, "optim/total_tokens": 2970353664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491276, "loss/out": 3.9430432319641113, "created_at": "2025-01-15T04:58:05.041257+00:00"} {"global_step": 11332, "acc_step": 0, "speed/wps": 12957.443854290505, "speed/FLOPS": 203514461125360.1, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2964614629745483, "optim/lr": 0.0029029843469358945, "optim/total_tokens": 2970615808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435567, "loss/out": 4.16920804977417, "created_at": "2025-01-15T04:58:10.105779+00:00"} {"global_step": 11333, "acc_step": 0, "speed/wps": 12959.43361324835, "speed/FLOPS": 203545713024008.38, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0008, "optim/grad_norm": 3.3131322860717773, "optim/lr": 0.002902954031556984, "optim/total_tokens": 2970877952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 4.034138202667236, "created_at": "2025-01-15T04:58:15.166552+00:00"} {"global_step": 11334, "acc_step": 0, "speed/wps": 12966.546345373523, "speed/FLOPS": 203657428255952.4, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.0912519693374634, "optim/lr": 0.0029029237116006902, "optim/total_tokens": 2971140096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 4.052275657653809, "created_at": "2025-01-15T04:58:20.231056+00:00"} {"global_step": 11335, "acc_step": 0, "speed/wps": 12958.038433075142, "speed/FLOPS": 203523799802209.7, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8685958981513977, "optim/lr": 0.002902893387067111, "optim/total_tokens": 2971402240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338751, "loss/out": 4.088756084442139, "created_at": "2025-01-15T04:58:25.292830+00:00"} {"global_step": 11336, "acc_step": 0, "speed/wps": 12962.255292045225, "speed/FLOPS": 203590031366906.28, "speed/curr_iter_time": 1.2731, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8507897853851318, "optim/lr": 0.0029028630579563458, "optim/total_tokens": 2971664384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 343977, "loss/out": 4.046849727630615, "created_at": "2025-01-15T04:58:30.353388+00:00"} {"global_step": 11337, "acc_step": 0, "speed/wps": 12960.326730636812, "speed/FLOPS": 203559740659866.1, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5440220236778259, "optim/lr": 0.0029028327242684934, "optim/total_tokens": 2971926528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451275, "loss/out": 4.178963661193848, "created_at": "2025-01-15T04:58:35.418148+00:00"} {"global_step": 11338, "acc_step": 0, "speed/wps": 12965.755842078668, "speed/FLOPS": 203645012315441.34, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8025857210159302, "optim/lr": 0.0029028023860036526, "optim/total_tokens": 2972188672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 4.13502311706543, "created_at": "2025-01-15T04:58:40.478457+00:00"} {"global_step": 11339, "acc_step": 0, "speed/wps": 12968.368594997182, "speed/FLOPS": 203686049190325.34, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6265766024589539, "optim/lr": 0.0029027720431619233, "optim/total_tokens": 2972450816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456045, "loss/out": 4.084290027618408, "created_at": "2025-01-15T04:58:45.540419+00:00"} {"global_step": 11340, "acc_step": 0, "speed/wps": 12948.412881896675, "speed/FLOPS": 203372617294058.47, "speed/curr_iter_time": 1.2788, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5603998303413391, "optim/lr": 0.0029027416957434033, "optim/total_tokens": 2972712960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461622, "loss/out": 3.9559810161590576, "created_at": "2025-01-15T04:58:50.605504+00:00"} {"global_step": 11341, "acc_step": 0, "speed/wps": 12951.521168292156, "speed/FLOPS": 203421437203132.6, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.4186055362224579, "optim/lr": 0.002902711343748192, "optim/total_tokens": 2972975104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 4.035147190093994, "created_at": "2025-01-15T04:58:55.669319+00:00"} {"global_step": 11342, "acc_step": 0, "speed/wps": 12967.470927878008, "speed/FLOPS": 203671950094694.84, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.5911425352096558, "optim/lr": 0.0029026809871763887, "optim/total_tokens": 2973237248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 4.004898548126221, "created_at": "2025-01-15T04:59:00.731553+00:00"} {"global_step": 11343, "acc_step": 0, "speed/wps": 12957.445216660357, "speed/FLOPS": 203514482523250.72, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4736248552799225, "optim/lr": 0.0029026506260280923, "optim/total_tokens": 2973499392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436666, "loss/out": 3.954012632369995, "created_at": "2025-01-15T04:59:05.797860+00:00"} {"global_step": 11344, "acc_step": 0, "speed/wps": 12961.891226836906, "speed/FLOPS": 203584313222531.62, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.42021647095680237, "optim/lr": 0.002902620260303402, "optim/total_tokens": 2973761536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 324468, "loss/out": 4.060162544250488, "created_at": "2025-01-15T04:59:10.857753+00:00"} {"global_step": 11345, "acc_step": 0, "speed/wps": 12960.937957523784, "speed/FLOPS": 203569340818039.72, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.286221504211426, "optim/lr": 0.0029025898900024165, "optim/total_tokens": 2974023680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465633, "loss/out": 4.006823539733887, "created_at": "2025-01-15T04:59:15.918021+00:00"} {"global_step": 11346, "acc_step": 0, "speed/wps": 12957.479149544459, "speed/FLOPS": 203515015485824.2, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7899985909461975, "optim/lr": 0.0029025595151252353, "optim/total_tokens": 2974285824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457665, "loss/out": 4.029566764831543, "created_at": "2025-01-15T04:59:20.982184+00:00"} {"global_step": 11347, "acc_step": 0, "speed/wps": 12963.76224033977, "speed/FLOPS": 203613700060636.1, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5049197673797607, "optim/lr": 0.0029025291356719572, "optim/total_tokens": 2974547968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 472515, "loss/out": 3.9998619556427, "created_at": "2025-01-15T04:59:26.041156+00:00"} {"global_step": 11348, "acc_step": 0, "speed/wps": 12952.28836593874, "speed/FLOPS": 203433487096412.25, "speed/curr_iter_time": 1.2777, "speed/data_load_time": 0.001, "optim/grad_norm": 1.0825726985931396, "optim/lr": 0.0029024987516426816, "optim/total_tokens": 2974810112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446747, "loss/out": 4.065868377685547, "created_at": "2025-01-15T04:59:31.108054+00:00"} {"global_step": 11349, "acc_step": 0, "speed/wps": 12967.672885871663, "speed/FLOPS": 203675122122504.56, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6502658128738403, "optim/lr": 0.0029024683630375072, "optim/total_tokens": 2975072256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480954, "loss/out": 4.046566963195801, "created_at": "2025-01-15T04:59:36.166101+00:00"} {"global_step": 11350, "acc_step": 0, "speed/wps": 12959.87562276638, "speed/FLOPS": 203552655390873.78, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2614028453826904, "optim/lr": 0.002902437969856534, "optim/total_tokens": 2975334400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444878, "loss/out": 4.1039838790893555, "created_at": "2025-01-15T04:59:41.231643+00:00"} {"global_step": 11351, "acc_step": 0, "speed/wps": 12953.583113918263, "speed/FLOPS": 203453822892600.8, "speed/curr_iter_time": 1.2768, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9832552671432495, "optim/lr": 0.0029024075720998598, "optim/total_tokens": 2975596544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290611, "loss/out": 4.131786346435547, "created_at": "2025-01-15T04:59:46.296618+00:00"} {"global_step": 11352, "acc_step": 0, "speed/wps": 12953.915032583101, "speed/FLOPS": 203459036131336.16, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7458628416061401, "optim/lr": 0.002902377169767585, "optim/total_tokens": 2975858688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 419875, "loss/out": 4.1318206787109375, "created_at": "2025-01-15T04:59:51.359943+00:00"} {"global_step": 11353, "acc_step": 0, "speed/wps": 12952.795113082542, "speed/FLOPS": 203441446256649.2, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7913044691085815, "optim/lr": 0.002902346762859808, "optim/total_tokens": 2976120832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449420, "loss/out": 3.9969944953918457, "created_at": "2025-01-15T04:59:56.426490+00:00"} {"global_step": 11354, "acc_step": 0, "speed/wps": 12957.030575561692, "speed/FLOPS": 203507970015019.6, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7133147120475769, "optim/lr": 0.002902316351376629, "optim/total_tokens": 2976382976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490003, "loss/out": 4.085008144378662, "created_at": "2025-01-15T05:00:01.488682+00:00"} {"global_step": 11355, "acc_step": 0, "speed/wps": 12963.846252843434, "speed/FLOPS": 203615019592451.4, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5541220307350159, "optim/lr": 0.002902285935318146, "optim/total_tokens": 2976645120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485760, "loss/out": 4.16298246383667, "created_at": "2025-01-15T05:00:06.548303+00:00"} {"global_step": 11356, "acc_step": 0, "speed/wps": 12962.907646889998, "speed/FLOPS": 203600277496172.97, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5502488613128662, "optim/lr": 0.002902255514684459, "optim/total_tokens": 2976907264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480954, "loss/out": 4.0276899337768555, "created_at": "2025-01-15T05:00:11.609579+00:00"} {"global_step": 11357, "acc_step": 0, "speed/wps": 12951.733936293242, "speed/FLOPS": 203424779016963.5, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6602634191513062, "optim/lr": 0.0029022250894756674, "optim/total_tokens": 2977169408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 4.075494289398193, "created_at": "2025-01-15T05:00:16.673430+00:00"} {"global_step": 11358, "acc_step": 0, "speed/wps": 12959.976326381246, "speed/FLOPS": 203554237079526.5, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5431414246559143, "optim/lr": 0.0029021946596918696, "optim/total_tokens": 2977431552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 4.001798629760742, "created_at": "2025-01-15T05:00:21.737291+00:00"} {"global_step": 11359, "acc_step": 0, "speed/wps": 12955.938887347793, "speed/FLOPS": 203490823551483.5, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8888225555419922, "optim/lr": 0.002902164225333166, "optim/total_tokens": 2977693696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 4.047994613647461, "created_at": "2025-01-15T05:00:26.799427+00:00"} {"global_step": 11360, "acc_step": 0, "speed/wps": 12954.513263105939, "speed/FLOPS": 203468432163752.2, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0791419744491577, "optim/lr": 0.0029021337863996545, "optim/total_tokens": 2977955840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 288203, "loss/out": 4.00446891784668, "created_at": "2025-01-15T05:00:31.864785+00:00"} {"global_step": 11361, "acc_step": 0, "speed/wps": 12951.272556871783, "speed/FLOPS": 203417532419147.75, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.001, "optim/grad_norm": 0.6031758189201355, "optim/lr": 0.0029021033428914362, "optim/total_tokens": 2978217984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431829, "loss/out": 4.157771587371826, "created_at": "2025-01-15T05:00:36.931806+00:00"} {"global_step": 11362, "acc_step": 0, "speed/wps": 12956.324481549927, "speed/FLOPS": 203496879838290.0, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.49322831630706787, "optim/lr": 0.0029020728948086086, "optim/total_tokens": 2978480128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453131, "loss/out": 4.086716175079346, "created_at": "2025-01-15T05:00:41.998660+00:00"} {"global_step": 11363, "acc_step": 0, "speed/wps": 12962.715498643345, "speed/FLOPS": 203597259543919.9, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7116097211837769, "optim/lr": 0.0029020424421512723, "optim/total_tokens": 2978742272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459765, "loss/out": 4.025266170501709, "created_at": "2025-01-15T05:00:47.062569+00:00"} {"global_step": 11364, "acc_step": 0, "speed/wps": 12964.43663994595, "speed/FLOPS": 203624292433173.3, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0766417980194092, "optim/lr": 0.0029020119849195265, "optim/total_tokens": 2979004416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460287, "loss/out": 4.148294448852539, "created_at": "2025-01-15T05:00:52.121861+00:00"} {"global_step": 11365, "acc_step": 0, "speed/wps": 12953.033339663663, "speed/FLOPS": 203445187932463.34, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5151420831680298, "optim/lr": 0.0029019815231134706, "optim/total_tokens": 2979266560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461605, "loss/out": 3.9150896072387695, "created_at": "2025-01-15T05:00:57.192155+00:00"} {"global_step": 11366, "acc_step": 0, "speed/wps": 12952.85176966301, "speed/FLOPS": 203442336126099.75, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5800260901451111, "optim/lr": 0.0029019510567332026, "optim/total_tokens": 2979528704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328073, "loss/out": 3.93082594871521, "created_at": "2025-01-15T05:01:02.256495+00:00"} {"global_step": 11367, "acc_step": 0, "speed/wps": 12955.65410759297, "speed/FLOPS": 203486350694105.62, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.001, "optim/grad_norm": 0.7783485054969788, "optim/lr": 0.002901920585778824, "optim/total_tokens": 2979790848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 317586, "loss/out": 4.156974792480469, "created_at": "2025-01-15T05:01:07.320324+00:00"} {"global_step": 11368, "acc_step": 0, "speed/wps": 12962.017079502053, "speed/FLOPS": 203586289911577.44, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5726044774055481, "optim/lr": 0.0029018901102504328, "optim/total_tokens": 2980052992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330717, "loss/out": 3.958967447280884, "created_at": "2025-01-15T05:01:12.384580+00:00"} {"global_step": 11369, "acc_step": 0, "speed/wps": 12955.707840959243, "speed/FLOPS": 203487194650460.3, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4205830991268158, "optim/lr": 0.002901859630148129, "optim/total_tokens": 2980315136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446505, "loss/out": 4.126700401306152, "created_at": "2025-01-15T05:01:17.447071+00:00"} {"global_step": 11370, "acc_step": 0, "speed/wps": 12952.4821339259, "speed/FLOPS": 203436530488915.7, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.431802362203598, "optim/lr": 0.002901829145472012, "optim/total_tokens": 2980577280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 4.029189586639404, "created_at": "2025-01-15T05:01:22.516550+00:00"} {"global_step": 11371, "acc_step": 0, "speed/wps": 12961.378964307605, "speed/FLOPS": 203576267435586.47, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1599005460739136, "optim/lr": 0.002901798656222181, "optim/total_tokens": 2980839424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439871, "loss/out": 4.065458297729492, "created_at": "2025-01-15T05:01:27.580906+00:00"} {"global_step": 11372, "acc_step": 0, "speed/wps": 12956.018640708466, "speed/FLOPS": 203492076187605.47, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1741666793823242, "optim/lr": 0.002901768162398736, "optim/total_tokens": 2981101568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496101, "loss/out": 4.2279815673828125, "created_at": "2025-01-15T05:01:32.648502+00:00"} {"global_step": 11373, "acc_step": 0, "speed/wps": 12954.749451755693, "speed/FLOPS": 203472141831053.8, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.2420904636383057, "optim/lr": 0.0029017376640017756, "optim/total_tokens": 2981363712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445933, "loss/out": 4.0325493812561035, "created_at": "2025-01-15T05:01:37.710961+00:00"} {"global_step": 11374, "acc_step": 0, "speed/wps": 12957.265522311349, "speed/FLOPS": 203511660176573.62, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0011, "optim/grad_norm": 1.1101353168487549, "optim/lr": 0.0029017071610314, "optim/total_tokens": 2981625856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332731, "loss/out": 4.197174072265625, "created_at": "2025-01-15T05:01:42.772412+00:00"} {"global_step": 11375, "acc_step": 0, "speed/wps": 12957.436052283032, "speed/FLOPS": 203514338584111.66, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.030299186706543, "optim/lr": 0.002901676653487709, "optim/total_tokens": 2981888000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434531, "loss/out": 4.00463342666626, "created_at": "2025-01-15T05:01:47.841036+00:00"} {"global_step": 11376, "acc_step": 0, "speed/wps": 12955.60881610452, "speed/FLOPS": 203485639329042.5, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1702224016189575, "optim/lr": 0.0029016461413708017, "optim/total_tokens": 2982150144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476995, "loss/out": 4.003551483154297, "created_at": "2025-01-15T05:01:52.907545+00:00"} {"global_step": 11377, "acc_step": 0, "speed/wps": 12959.797490289231, "speed/FLOPS": 203551428212954.88, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1032812595367432, "optim/lr": 0.002901615624680777, "optim/total_tokens": 2982412288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456567, "loss/out": 4.004251956939697, "created_at": "2025-01-15T05:01:57.968033+00:00"} {"global_step": 11378, "acc_step": 0, "speed/wps": 12952.538504056893, "speed/FLOPS": 203437415859283.62, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.017781138420105, "optim/lr": 0.002901585103417736, "optim/total_tokens": 2982674432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462963, "loss/out": 3.998664617538452, "created_at": "2025-01-15T05:02:03.034586+00:00"} {"global_step": 11379, "acc_step": 0, "speed/wps": 12963.170680037316, "speed/FLOPS": 203604408793197.0, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.5029411315917969, "optim/lr": 0.0029015545775817764, "optim/total_tokens": 2982936576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329272, "loss/out": 4.043740749359131, "created_at": "2025-01-15T05:02:08.094331+00:00"} {"global_step": 11380, "acc_step": 0, "speed/wps": 12964.621078719623, "speed/FLOPS": 203627189297559.28, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7511724233627319, "optim/lr": 0.002901524047173, "optim/total_tokens": 2983198720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 303224, "loss/out": 4.046229362487793, "created_at": "2025-01-15T05:02:13.154299+00:00"} {"global_step": 11381, "acc_step": 0, "speed/wps": 12953.758905123148, "speed/FLOPS": 203456583935035.97, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.063422441482544, "optim/lr": 0.0029014935121915043, "optim/total_tokens": 2983460864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 344264, "loss/out": 4.104648113250732, "created_at": "2025-01-15T05:02:18.221613+00:00"} {"global_step": 11382, "acc_step": 0, "speed/wps": 12958.622981618766, "speed/FLOPS": 203532980940341.2, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0316828489303589, "optim/lr": 0.0029014629726373904, "optim/total_tokens": 2983723008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 420893, "loss/out": 4.034982204437256, "created_at": "2025-01-15T05:02:23.284756+00:00"} {"global_step": 11383, "acc_step": 0, "speed/wps": 12951.811537709656, "speed/FLOPS": 203425997853843.84, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.687514305114746, "optim/lr": 0.002901432428510757, "optim/total_tokens": 2983985152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465057, "loss/out": 4.091800212860107, "created_at": "2025-01-15T05:02:28.353538+00:00"} {"global_step": 11384, "acc_step": 0, "speed/wps": 12962.441982513674, "speed/FLOPS": 203592963597252.88, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7519968152046204, "optim/lr": 0.0029014018798117045, "optim/total_tokens": 2984247296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 459237, "loss/out": 4.003401756286621, "created_at": "2025-01-15T05:02:33.414329+00:00"} {"global_step": 11385, "acc_step": 0, "speed/wps": 12956.167532655703, "speed/FLOPS": 203494414740235.8, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0359363555908203, "optim/lr": 0.002901371326540332, "optim/total_tokens": 2984509440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454182, "loss/out": 4.181413650512695, "created_at": "2025-01-15T05:02:38.478404+00:00"} {"global_step": 11386, "acc_step": 0, "speed/wps": 12956.642837580328, "speed/FLOPS": 203501880057214.47, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8038222193717957, "optim/lr": 0.0029013407686967396, "optim/total_tokens": 2984771584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476758, "loss/out": 4.065608024597168, "created_at": "2025-01-15T05:02:43.545342+00:00"} {"global_step": 11387, "acc_step": 0, "speed/wps": 12952.761080747929, "speed/FLOPS": 203440911732068.75, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8911297917366028, "optim/lr": 0.002901310206281026, "optim/total_tokens": 2985033728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 332731, "loss/out": 4.119570255279541, "created_at": "2025-01-15T05:02:48.613940+00:00"} {"global_step": 11388, "acc_step": 0, "speed/wps": 12958.990607681939, "speed/FLOPS": 203538755012834.44, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2860099077224731, "optim/lr": 0.0029012796392932924, "optim/total_tokens": 2985295872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355325, "loss/out": 4.08707332611084, "created_at": "2025-01-15T05:02:53.677409+00:00"} {"global_step": 11389, "acc_step": 0, "speed/wps": 12962.492097381028, "speed/FLOPS": 203593750720109.7, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0605756044387817, "optim/lr": 0.002901249067733638, "optim/total_tokens": 2985558016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437955, "loss/out": 4.215546607971191, "created_at": "2025-01-15T05:02:58.743012+00:00"} {"global_step": 11390, "acc_step": 0, "speed/wps": 12960.206195507382, "speed/FLOPS": 203557847490025.75, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2245984077453613, "optim/lr": 0.0029012184916021623, "optim/total_tokens": 2985820160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482002, "loss/out": 4.052175521850586, "created_at": "2025-01-15T05:03:03.803335+00:00"} {"global_step": 11391, "acc_step": 0, "speed/wps": 12958.68554062878, "speed/FLOPS": 203533963515558.53, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.828603208065033, "optim/lr": 0.0029011879108989647, "optim/total_tokens": 2986082304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434807, "loss/out": 3.933093547821045, "created_at": "2025-01-15T05:03:08.868378+00:00"} {"global_step": 11392, "acc_step": 0, "speed/wps": 12953.334630776866, "speed/FLOPS": 203449920123413.56, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0420104265213013, "optim/lr": 0.002901157325624146, "optim/total_tokens": 2986344448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 4.0763020515441895, "created_at": "2025-01-15T05:03:13.935375+00:00"} {"global_step": 11393, "acc_step": 0, "speed/wps": 12952.935490604279, "speed/FLOPS": 203443651078527.0, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8402422070503235, "optim/lr": 0.002901126735777805, "optim/total_tokens": 2986606592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330329, "loss/out": 4.0962066650390625, "created_at": "2025-01-15T05:03:18.998453+00:00"} {"global_step": 11394, "acc_step": 0, "speed/wps": 12958.206210607064, "speed/FLOPS": 203526434978899.0, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9048054218292236, "optim/lr": 0.002901096141360042, "optim/total_tokens": 2986868736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 296822, "loss/out": 3.8730156421661377, "created_at": "2025-01-15T05:03:24.060523+00:00"} {"global_step": 11395, "acc_step": 0, "speed/wps": 12964.912140935561, "speed/FLOPS": 203631760829621.97, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9425844550132751, "optim/lr": 0.0029010655423709567, "optim/total_tokens": 2987130880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 290611, "loss/out": 4.146113872528076, "created_at": "2025-01-15T05:03:29.123716+00:00"} {"global_step": 11396, "acc_step": 0, "speed/wps": 12955.18056862883, "speed/FLOPS": 203478913113963.06, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6791093349456787, "optim/lr": 0.0029010349388106493, "optim/total_tokens": 2987393024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458442, "loss/out": 3.993252992630005, "created_at": "2025-01-15T05:03:34.188783+00:00"} {"global_step": 11397, "acc_step": 0, "speed/wps": 12951.861105555774, "speed/FLOPS": 203426776384980.47, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.02397620677948, "optim/lr": 0.002901004330679219, "optim/total_tokens": 2987655168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492149, "loss/out": 4.006966590881348, "created_at": "2025-01-15T05:03:39.254443+00:00"} {"global_step": 11398, "acc_step": 0, "speed/wps": 12953.357099237244, "speed/FLOPS": 203450273021458.7, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.237101435661316, "optim/lr": 0.0029009737179767664, "optim/total_tokens": 2987917312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434230, "loss/out": 4.312933921813965, "created_at": "2025-01-15T05:03:44.322877+00:00"} {"global_step": 11399, "acc_step": 0, "speed/wps": 12947.978806032977, "speed/FLOPS": 203365799536136.34, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0009, "optim/grad_norm": 2.202425718307495, "optim/lr": 0.00290094310070339, "optim/total_tokens": 2988179456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449940, "loss/out": 4.023387432098389, "created_at": "2025-01-15T05:03:49.389161+00:00"} {"global_step": 11400, "acc_step": 0, "speed/wps": 12950.374228043162, "speed/FLOPS": 203403422930461.06, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0798406600952148, "optim/lr": 0.0029009124788591917, "optim/total_tokens": 2988441600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 309025, "loss/out": 4.155696392059326, "created_at": "2025-01-15T05:03:54.454960+00:00"} {"global_step": 11401, "acc_step": 0, "speed/wps": 12949.325243449151, "speed/FLOPS": 203386947185956.25, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0510298013687134, "optim/lr": 0.0029008818524442702, "optim/total_tokens": 2988703744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433167, "loss/out": 4.109410762786865, "created_at": "2025-01-15T05:03:59.524727+00:00"} {"global_step": 11402, "acc_step": 0, "speed/wps": 12956.314908448277, "speed/FLOPS": 203496729479573.75, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.001, "optim/grad_norm": 0.9746978878974915, "optim/lr": 0.0029008512214587254, "optim/total_tokens": 2988965888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484426, "loss/out": 4.052346229553223, "created_at": "2025-01-15T05:04:04.587654+00:00"} {"global_step": 11403, "acc_step": 0, "speed/wps": 12961.329058585796, "speed/FLOPS": 203575483597647.8, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8236727714538574, "optim/lr": 0.0029008205859026577, "optim/total_tokens": 2989228032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465633, "loss/out": 4.038829803466797, "created_at": "2025-01-15T05:04:09.655388+00:00"} {"global_step": 11404, "acc_step": 0, "speed/wps": 12953.843231559445, "speed/FLOPS": 203457908397594.7, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.8523650169372559, "optim/lr": 0.002900789945776166, "optim/total_tokens": 2989490176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484189, "loss/out": 3.9433674812316895, "created_at": "2025-01-15T05:04:14.718278+00:00"} {"global_step": 11405, "acc_step": 0, "speed/wps": 12964.195934913381, "speed/FLOPS": 203620511829873.44, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.5998976230621338, "optim/lr": 0.002900759301079352, "optim/total_tokens": 2989752320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440105, "loss/out": 4.135979652404785, "created_at": "2025-01-15T05:04:19.782334+00:00"} {"global_step": 11406, "acc_step": 0, "speed/wps": 12962.310465064857, "speed/FLOPS": 203590897934995.28, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.752080500125885, "optim/lr": 0.002900728651812314, "optim/total_tokens": 2990014464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300290, "loss/out": 4.022970676422119, "created_at": "2025-01-15T05:04:24.844604+00:00"} {"global_step": 11407, "acc_step": 0, "speed/wps": 12960.564515829748, "speed/FLOPS": 203563475403072.25, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3220808506011963, "optim/lr": 0.0029006979979751533, "optim/total_tokens": 2990276608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356011, "loss/out": 3.9504683017730713, "created_at": "2025-01-15T05:04:29.907542+00:00"} {"global_step": 11408, "acc_step": 0, "speed/wps": 12954.91942657917, "speed/FLOPS": 203474811519226.38, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8674336671829224, "optim/lr": 0.0029006673395679692, "optim/total_tokens": 2990538752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456868, "loss/out": 4.226620674133301, "created_at": "2025-01-15T05:04:34.970050+00:00"} {"global_step": 11409, "acc_step": 0, "speed/wps": 12961.839290135373, "speed/FLOPS": 203583497485265.53, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.008914828300476, "optim/lr": 0.0029006366765908617, "optim/total_tokens": 2990800896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454710, "loss/out": 3.985309600830078, "created_at": "2025-01-15T05:04:40.030050+00:00"} {"global_step": 11410, "acc_step": 0, "speed/wps": 12965.54700636017, "speed/FLOPS": 203641732263511.97, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7587065100669861, "optim/lr": 0.002900606009043931, "optim/total_tokens": 2991063040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465349, "loss/out": 4.025361061096191, "created_at": "2025-01-15T05:04:45.089991+00:00"} {"global_step": 11411, "acc_step": 0, "speed/wps": 12956.01661787974, "speed/FLOPS": 203492044416300.7, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0422425270080566, "optim/lr": 0.002900575336927277, "optim/total_tokens": 2991325184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455517, "loss/out": 3.998905658721924, "created_at": "2025-01-15T05:04:50.156030+00:00"} {"global_step": 11412, "acc_step": 0, "speed/wps": 12961.553119283742, "speed/FLOPS": 203579002778800.72, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7890178561210632, "optim/lr": 0.002900544660241001, "optim/total_tokens": 2991587328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464007, "loss/out": 4.0325164794921875, "created_at": "2025-01-15T05:04:55.219831+00:00"} {"global_step": 11413, "acc_step": 0, "speed/wps": 12956.230432939965, "speed/FLOPS": 203495402675634.2, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5437406897544861, "optim/lr": 0.0029005139789852006, "optim/total_tokens": 2991849472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448890, "loss/out": 4.011490821838379, "created_at": "2025-01-15T05:05:00.281873+00:00"} {"global_step": 11414, "acc_step": 0, "speed/wps": 12954.106636234961, "speed/FLOPS": 203462045529979.6, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1406466960906982, "optim/lr": 0.0029004832931599786, "optim/total_tokens": 2992111616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327499, "loss/out": 4.048981666564941, "created_at": "2025-01-15T05:05:05.346859+00:00"} {"global_step": 11415, "acc_step": 0, "speed/wps": 12957.390366588104, "speed/FLOPS": 203513621027494.1, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6147736310958862, "optim/lr": 0.002900452602765433, "optim/total_tokens": 2992373760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 360123, "loss/out": 3.9851884841918945, "created_at": "2025-01-15T05:05:10.434349+00:00"} {"global_step": 11416, "acc_step": 0, "speed/wps": 12959.7355339453, "speed/FLOPS": 203550455103436.5, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9766415357589722, "optim/lr": 0.0029004219078016653, "optim/total_tokens": 2992635904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443022, "loss/out": 4.066046237945557, "created_at": "2025-01-15T05:05:15.500030+00:00"} {"global_step": 11417, "acc_step": 0, "speed/wps": 12962.051267936396, "speed/FLOPS": 203586826887918.62, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.117409586906433, "optim/lr": 0.002900391208268775, "optim/total_tokens": 2992898048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479899, "loss/out": 3.992964267730713, "created_at": "2025-01-15T05:05:20.560933+00:00"} {"global_step": 11418, "acc_step": 0, "speed/wps": 12956.302157852673, "speed/FLOPS": 203496529213948.62, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.45804861187934875, "optim/lr": 0.0029003605041668626, "optim/total_tokens": 2993160192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446742, "loss/out": 4.06264591217041, "created_at": "2025-01-15T05:05:25.622994+00:00"} {"global_step": 11419, "acc_step": 0, "speed/wps": 12957.501066913108, "speed/FLOPS": 203515359728216.44, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6885249614715576, "optim/lr": 0.0029003297954960277, "optim/total_tokens": 2993422336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314655, "loss/out": 4.021171569824219, "created_at": "2025-01-15T05:05:30.686191+00:00"} {"global_step": 11420, "acc_step": 0, "speed/wps": 12957.542523057542, "speed/FLOPS": 203516010853931.5, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2945448160171509, "optim/lr": 0.0029002990822563713, "optim/total_tokens": 2993684480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 399773, "loss/out": 3.9587810039520264, "created_at": "2025-01-15T05:05:35.747972+00:00"} {"global_step": 11421, "acc_step": 0, "speed/wps": 12953.757421914825, "speed/FLOPS": 203456560639211.1, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.9728864431381226, "optim/lr": 0.0029002683644479927, "optim/total_tokens": 2993946624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453666, "loss/out": 4.052166938781738, "created_at": "2025-01-15T05:05:40.813221+00:00"} {"global_step": 11422, "acc_step": 0, "speed/wps": 12958.899598910775, "speed/FLOPS": 203537325595024.16, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7539063096046448, "optim/lr": 0.002900237642070993, "optim/total_tokens": 2994208768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 460050, "loss/out": 4.1011576652526855, "created_at": "2025-01-15T05:05:45.880046+00:00"} {"global_step": 11423, "acc_step": 0, "speed/wps": 12953.421381448854, "speed/FLOPS": 203451282661925.03, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7336515784263611, "optim/lr": 0.002900206915125472, "optim/total_tokens": 2994470912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439820, "loss/out": 4.025017261505127, "created_at": "2025-01-15T05:05:50.952158+00:00"} {"global_step": 11424, "acc_step": 0, "speed/wps": 12956.923786353156, "speed/FLOPS": 203506292743755.78, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.1222689151763916, "optim/lr": 0.0029001761836115296, "optim/total_tokens": 2994733056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 489945, "loss/out": 4.084935188293457, "created_at": "2025-01-15T05:05:56.019348+00:00"} {"global_step": 11425, "acc_step": 0, "speed/wps": 12958.576798351125, "speed/FLOPS": 203532255568660.4, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7549536228179932, "optim/lr": 0.0029001454475292665, "optim/total_tokens": 2994995200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 431605, "loss/out": 3.9982855319976807, "created_at": "2025-01-15T05:06:01.080696+00:00"} {"global_step": 11426, "acc_step": 0, "speed/wps": 12956.772388749536, "speed/FLOPS": 203503914836348.8, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1494661569595337, "optim/lr": 0.0029001147068787834, "optim/total_tokens": 2995257344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292246, "loss/out": 4.065334320068359, "created_at": "2025-01-15T05:06:06.146556+00:00"} {"global_step": 11427, "acc_step": 0, "speed/wps": 12957.639902496478, "speed/FLOPS": 203517540331833.34, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.056771993637085, "optim/lr": 0.0029000839616601807, "optim/total_tokens": 2995519488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443306, "loss/out": 3.9830849170684814, "created_at": "2025-01-15T05:06:11.210568+00:00"} {"global_step": 11428, "acc_step": 0, "speed/wps": 12954.091371739627, "speed/FLOPS": 203461805780104.78, "speed/curr_iter_time": 1.2763, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0326648950576782, "optim/lr": 0.002900053211873557, "optim/total_tokens": 2995781632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 4.1118245124816895, "created_at": "2025-01-15T05:06:16.277284+00:00"} {"global_step": 11429, "acc_step": 0, "speed/wps": 12960.390848113915, "speed/FLOPS": 203560747712950.62, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.073800802230835, "optim/lr": 0.0029000224575190144, "optim/total_tokens": 2996043776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 4.031219959259033, "created_at": "2025-01-15T05:06:21.338752+00:00"} {"global_step": 11430, "acc_step": 0, "speed/wps": 12954.643568250276, "speed/FLOPS": 203470478785102.03, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9177780151367188, "optim/lr": 0.0028999916985966525, "optim/total_tokens": 2996305920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474085, "loss/out": 4.01711368560791, "created_at": "2025-01-15T05:06:26.402569+00:00"} {"global_step": 11431, "acc_step": 0, "speed/wps": 12954.26033246304, "speed/FLOPS": 203464459540442.34, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7495729923248291, "optim/lr": 0.0028999609351065716, "optim/total_tokens": 2996568064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330329, "loss/out": 4.119412422180176, "created_at": "2025-01-15T05:06:31.466196+00:00"} {"global_step": 11432, "acc_step": 0, "speed/wps": 12955.689545755858, "speed/FLOPS": 203486907299150.9, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5708848237991333, "optim/lr": 0.0028999301670488723, "optim/total_tokens": 2996830208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354800, "loss/out": 4.0954132080078125, "created_at": "2025-01-15T05:06:36.533412+00:00"} {"global_step": 11433, "acc_step": 0, "speed/wps": 12954.143682773407, "speed/FLOPS": 203462627396773.2, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7544553875923157, "optim/lr": 0.002899899394423655, "optim/total_tokens": 2997092352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 425730, "loss/out": 3.9505810737609863, "created_at": "2025-01-15T05:06:41.599838+00:00"} {"global_step": 11434, "acc_step": 0, "speed/wps": 12963.640000592952, "speed/FLOPS": 203611780117437.28, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9243071675300598, "optim/lr": 0.00289986861723102, "optim/total_tokens": 2997354496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474361, "loss/out": 4.092345237731934, "created_at": "2025-01-15T05:06:46.665430+00:00"} {"global_step": 11435, "acc_step": 0, "speed/wps": 12956.730221830725, "speed/FLOPS": 203503252546945.47, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6611314415931702, "optim/lr": 0.0028998378354710684, "optim/total_tokens": 2997616640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 4.0970306396484375, "created_at": "2025-01-15T05:06:51.728868+00:00"} {"global_step": 11436, "acc_step": 0, "speed/wps": 12951.426748414113, "speed/FLOPS": 203419954209201.4, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0291540622711182, "optim/lr": 0.002899807049143899, "optim/total_tokens": 2997878784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437201, "loss/out": 4.0540971755981445, "created_at": "2025-01-15T05:06:56.801148+00:00"} {"global_step": 11437, "acc_step": 0, "speed/wps": 12964.076975785798, "speed/FLOPS": 203618643413307.78, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1277035474777222, "optim/lr": 0.0028997762582496133, "optim/total_tokens": 2998140928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493419, "loss/out": 3.999897003173828, "created_at": "2025-01-15T05:07:01.860006+00:00"} {"global_step": 11438, "acc_step": 0, "speed/wps": 12961.136398507142, "speed/FLOPS": 203572457606369.84, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8610352277755737, "optim/lr": 0.002899745462788313, "optim/total_tokens": 2998403072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453945, "loss/out": 4.06298303604126, "created_at": "2025-01-15T05:07:06.920829+00:00"} {"global_step": 11439, "acc_step": 0, "speed/wps": 12956.40321864106, "speed/FLOPS": 203498116512502.34, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2460397481918335, "optim/lr": 0.002899714662760096, "optim/total_tokens": 2998665216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 304859, "loss/out": 4.073932647705078, "created_at": "2025-01-15T05:07:11.986357+00:00"} {"global_step": 11440, "acc_step": 0, "speed/wps": 12955.39687246204, "speed/FLOPS": 203482310462896.25, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0347907543182373, "optim/lr": 0.002899683858165064, "optim/total_tokens": 2998927360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441679, "loss/out": 4.082104682922363, "created_at": "2025-01-15T05:07:17.054143+00:00"} {"global_step": 11441, "acc_step": 0, "speed/wps": 12960.071491554056, "speed/FLOPS": 203555731779336.12, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0209723711013794, "optim/lr": 0.0028996530490033178, "optim/total_tokens": 2999189504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483614, "loss/out": 4.130901336669922, "created_at": "2025-01-15T05:07:22.114876+00:00"} {"global_step": 11442, "acc_step": 0, "speed/wps": 12959.098177951022, "speed/FLOPS": 203540444551727.44, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2802162170410156, "optim/lr": 0.002899622235274958, "optim/total_tokens": 2999451648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452093, "loss/out": 4.128498554229736, "created_at": "2025-01-15T05:07:27.177885+00:00"} {"global_step": 11443, "acc_step": 0, "speed/wps": 12957.183059984753, "speed/FLOPS": 203510364992419.3, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.001, "optim/grad_norm": 1.1709798574447632, "optim/lr": 0.0028995914169800845, "optim/total_tokens": 2999713792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487621, "loss/out": 4.053783416748047, "created_at": "2025-01-15T05:07:32.243787+00:00"} {"global_step": 11444, "acc_step": 0, "speed/wps": 12958.815187538814, "speed/FLOPS": 203535999798434.12, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6538816094398499, "optim/lr": 0.0028995605941187987, "optim/total_tokens": 2999975936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444641, "loss/out": 3.975048542022705, "created_at": "2025-01-15T05:07:37.310618+00:00"} {"global_step": 11445, "acc_step": 0, "speed/wps": 12957.366032668913, "speed/FLOPS": 203513238829854.88, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1700446605682373, "optim/lr": 0.0028995297666912, "optim/total_tokens": 3000238080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 356674, "loss/out": 3.986571788787842, "created_at": "2025-01-15T05:07:42.375919+00:00"} {"global_step": 11446, "acc_step": 0, "speed/wps": 12951.856828874828, "speed/FLOPS": 203426709213829.44, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1779533624649048, "optim/lr": 0.00289949893469739, "optim/total_tokens": 3000500224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439291, "loss/out": 4.067045211791992, "created_at": "2025-01-15T05:07:47.439735+00:00"} {"global_step": 11447, "acc_step": 0, "speed/wps": 12959.38370818601, "speed/FLOPS": 203544929196427.56, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0235658884048462, "optim/lr": 0.0028994680981374687, "optim/total_tokens": 3000762368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482572, "loss/out": 4.152864456176758, "created_at": "2025-01-15T05:07:52.510097+00:00"} {"global_step": 11448, "acc_step": 0, "speed/wps": 12956.047566274172, "speed/FLOPS": 203492530503362.78, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6975364685058594, "optim/lr": 0.0028994372570115374, "optim/total_tokens": 3001024512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 4.1510114669799805, "created_at": "2025-01-15T05:07:57.574531+00:00"} {"global_step": 11449, "acc_step": 0, "speed/wps": 12951.775555786537, "speed/FLOPS": 203425432708296.88, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7579843401908875, "optim/lr": 0.0028994064113196955, "optim/total_tokens": 3001286656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434816, "loss/out": 4.084043025970459, "created_at": "2025-01-15T05:08:02.644190+00:00"} {"global_step": 11450, "acc_step": 0, "speed/wps": 12952.287035645564, "speed/FLOPS": 203433466202329.9, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0139487981796265, "optim/lr": 0.002899375561062045, "optim/total_tokens": 3001548800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 491804, "loss/out": 4.127760887145996, "created_at": "2025-01-15T05:08:07.708033+00:00"} {"global_step": 11451, "acc_step": 0, "speed/wps": 12956.602465495693, "speed/FLOPS": 203501245958148.2, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0934618711471558, "optim/lr": 0.0028993447062386855, "optim/total_tokens": 3001810944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 436903, "loss/out": 4.128995895385742, "created_at": "2025-01-15T05:08:12.774922+00:00"} {"global_step": 11452, "acc_step": 0, "speed/wps": 12959.40521910621, "speed/FLOPS": 203545267054988.4, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7601937651634216, "optim/lr": 0.0028993138468497185, "optim/total_tokens": 3002073088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 372174, "loss/out": 4.115936756134033, "created_at": "2025-01-15T05:08:17.838160+00:00"} {"global_step": 11453, "acc_step": 0, "speed/wps": 12950.933649204384, "speed/FLOPS": 203412209408523.3, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9882753491401672, "optim/lr": 0.0028992829828952443, "optim/total_tokens": 3002335232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443835, "loss/out": 3.9472498893737793, "created_at": "2025-01-15T05:08:22.902057+00:00"} {"global_step": 11454, "acc_step": 0, "speed/wps": 12956.229175787072, "speed/FLOPS": 203495382930320.5, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8620166778564453, "optim/lr": 0.002899252114375363, "optim/total_tokens": 3002597376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475425, "loss/out": 4.101415634155273, "created_at": "2025-01-15T05:08:27.965018+00:00"} {"global_step": 11455, "acc_step": 0, "speed/wps": 12952.353178725452, "speed/FLOPS": 203434505070289.66, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0311968326568604, "optim/lr": 0.0028992212412901766, "optim/total_tokens": 3002859520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448605, "loss/out": 4.0177130699157715, "created_at": "2025-01-15T05:08:33.028958+00:00"} {"global_step": 11456, "acc_step": 0, "speed/wps": 12959.66779985603, "speed/FLOPS": 203549391246488.12, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7670412659645081, "optim/lr": 0.002899190363639785, "optim/total_tokens": 3003121664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462417, "loss/out": 3.9614899158477783, "created_at": "2025-01-15T05:08:38.089386+00:00"} {"global_step": 11457, "acc_step": 0, "speed/wps": 12955.320666324451, "speed/FLOPS": 203481113540787.22, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0322316884994507, "optim/lr": 0.002899159481424289, "optim/total_tokens": 3003383808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 330091, "loss/out": 4.045082092285156, "created_at": "2025-01-15T05:08:43.154133+00:00"} {"global_step": 11458, "acc_step": 0, "speed/wps": 12958.58585934244, "speed/FLOPS": 203532397883980.12, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0643104314804077, "optim/lr": 0.002899128594643789, "optim/total_tokens": 3003645952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 364255, "loss/out": 4.085093021392822, "created_at": "2025-01-15T05:08:48.215036+00:00"} {"global_step": 11459, "acc_step": 0, "speed/wps": 12961.24110140288, "speed/FLOPS": 203574102109224.28, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1595693826675415, "optim/lr": 0.0028990977032983868, "optim/total_tokens": 3003908096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457380, "loss/out": 4.061545372009277, "created_at": "2025-01-15T05:08:53.281644+00:00"} {"global_step": 11460, "acc_step": 0, "speed/wps": 12964.805424806635, "speed/FLOPS": 203630084706173.88, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7928891777992249, "optim/lr": 0.002899066807388182, "optim/total_tokens": 3004170240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487093, "loss/out": 4.127057075500488, "created_at": "2025-01-15T05:08:58.341385+00:00"} {"global_step": 11461, "acc_step": 0, "speed/wps": 12954.821341529821, "speed/FLOPS": 203473270958743.38, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7464126944541931, "optim/lr": 0.0028990359069132766, "optim/total_tokens": 3004432384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 3.986506700515747, "created_at": "2025-01-15T05:09:03.404959+00:00"} {"global_step": 11462, "acc_step": 0, "speed/wps": 12957.598605590492, "speed/FLOPS": 203516891707176.97, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7857308387756348, "optim/lr": 0.0028990050018737707, "optim/total_tokens": 3004694528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481758, "loss/out": 3.9399752616882324, "created_at": "2025-01-15T05:09:08.467340+00:00"} {"global_step": 11463, "acc_step": 0, "speed/wps": 12962.037992707565, "speed/FLOPS": 203586618382207.75, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0372233390808105, "optim/lr": 0.0028989740922697655, "optim/total_tokens": 3004956672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449696, "loss/out": 4.107243537902832, "created_at": "2025-01-15T05:09:13.529012+00:00"} {"global_step": 11464, "acc_step": 0, "speed/wps": 12956.99741801708, "speed/FLOPS": 203507449230218.22, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1637665033340454, "optim/lr": 0.0028989431781013607, "optim/total_tokens": 3005218816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469848, "loss/out": 4.128034591674805, "created_at": "2025-01-15T05:09:18.592266+00:00"} {"global_step": 11465, "acc_step": 0, "speed/wps": 12954.249889589099, "speed/FLOPS": 203464295520757.06, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.9177682995796204, "optim/lr": 0.002898912259368659, "optim/total_tokens": 3005480960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447026, "loss/out": 4.121575355529785, "created_at": "2025-01-15T05:09:23.656820+00:00"} {"global_step": 11466, "acc_step": 0, "speed/wps": 12953.78186866512, "speed/FLOPS": 203456944609017.84, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2075539827346802, "optim/lr": 0.00289888133607176, "optim/total_tokens": 3005743104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 302403, "loss/out": 4.068304061889648, "created_at": "2025-01-15T05:09:28.719872+00:00"} {"global_step": 11467, "acc_step": 0, "speed/wps": 12959.532465070355, "speed/FLOPS": 203547265627709.6, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0370184183120728, "optim/lr": 0.002898850408210765, "optim/total_tokens": 3006005248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 418774, "loss/out": 3.927229404449463, "created_at": "2025-01-15T05:09:33.780820+00:00"} {"global_step": 11468, "acc_step": 0, "speed/wps": 12958.975849544993, "speed/FLOPS": 203538523216013.9, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2080520391464233, "optim/lr": 0.002898819475785775, "optim/total_tokens": 3006267392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448368, "loss/out": 4.076651573181152, "created_at": "2025-01-15T05:09:38.842745+00:00"} {"global_step": 11469, "acc_step": 0, "speed/wps": 12956.025546478744, "speed/FLOPS": 203492184652217.22, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0188541412353516, "optim/lr": 0.00289878853879689, "optim/total_tokens": 3006529536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 476952, "loss/out": 3.9984962940216064, "created_at": "2025-01-15T05:09:43.911127+00:00"} {"global_step": 11470, "acc_step": 0, "speed/wps": 12959.726318592415, "speed/FLOPS": 203550310363656.22, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.356695532798767, "optim/lr": 0.0028987575972442125, "optim/total_tokens": 3006791680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 468018, "loss/out": 4.0047607421875, "created_at": "2025-01-15T05:09:48.977159+00:00"} {"global_step": 11471, "acc_step": 0, "speed/wps": 12955.8175154658, "speed/FLOPS": 203488917239294.38, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0006, "optim/grad_norm": 4.746696949005127, "optim/lr": 0.0028987266511278425, "optim/total_tokens": 3007053824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452325, "loss/out": 4.146044731140137, "created_at": "2025-01-15T05:09:54.043686+00:00"} {"global_step": 11472, "acc_step": 0, "speed/wps": 12954.704142742881, "speed/FLOPS": 203471430190746.47, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0399882793426514, "optim/lr": 0.002898695700447881, "optim/total_tokens": 3007315968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 311268, "loss/out": 4.083449840545654, "created_at": "2025-01-15T05:09:59.106024+00:00"} {"global_step": 11473, "acc_step": 0, "speed/wps": 12960.145543293973, "speed/FLOPS": 203556894863669.28, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6746916770935059, "optim/lr": 0.002898664745204429, "optim/total_tokens": 3007578112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 406769, "loss/out": 4.095477104187012, "created_at": "2025-01-15T05:10:04.166772+00:00"} {"global_step": 11474, "acc_step": 0, "speed/wps": 12959.57976020279, "speed/FLOPS": 203548008462754.84, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2140870094299316, "optim/lr": 0.002898633785397588, "optim/total_tokens": 3007840256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455532, "loss/out": 3.9263405799865723, "created_at": "2025-01-15T05:10:09.229424+00:00"} {"global_step": 11475, "acc_step": 0, "speed/wps": 12951.578431126649, "speed/FLOPS": 203422336594635.7, "speed/curr_iter_time": 1.2773, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.4974491596221924, "optim/lr": 0.002898602821027458, "optim/total_tokens": 3008102400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 403566, "loss/out": 4.207411766052246, "created_at": "2025-01-15T05:10:14.293458+00:00"} {"global_step": 11476, "acc_step": 0, "speed/wps": 12960.10494649009, "speed/FLOPS": 203556257235078.44, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1747801303863525, "optim/lr": 0.0028985718520941413, "optim/total_tokens": 3008364544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437717, "loss/out": 3.9537851810455322, "created_at": "2025-01-15T05:10:19.357735+00:00"} {"global_step": 11477, "acc_step": 0, "speed/wps": 12958.009813590294, "speed/FLOPS": 203523350293872.84, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.7250505685806274, "optim/lr": 0.002898540878597738, "optim/total_tokens": 3008626688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493191, "loss/out": 4.054288387298584, "created_at": "2025-01-15T05:10:24.423553+00:00"} {"global_step": 11478, "acc_step": 0, "speed/wps": 12953.752729065005, "speed/FLOPS": 203456486931556.03, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.239559531211853, "optim/lr": 0.0028985099005383493, "optim/total_tokens": 3008888832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441442, "loss/out": 3.882291316986084, "created_at": "2025-01-15T05:10:29.490211+00:00"} {"global_step": 11479, "acc_step": 0, "speed/wps": 12957.321862508674, "speed/FLOPS": 203512545076792.16, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0474486351013184, "optim/lr": 0.002898478917916077, "optim/total_tokens": 3009150976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463750, "loss/out": 4.03099250793457, "created_at": "2025-01-15T05:10:34.555939+00:00"} {"global_step": 11480, "acc_step": 0, "speed/wps": 12960.116477694211, "speed/FLOPS": 203556438348484.72, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0615521669387817, "optim/lr": 0.0028984479307310206, "optim/total_tokens": 3009413120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 335427, "loss/out": 4.112447738647461, "created_at": "2025-01-15T05:10:39.621875+00:00"} {"global_step": 11481, "acc_step": 0, "speed/wps": 12957.090585470787, "speed/FLOPS": 203508912553104.2, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2035871744155884, "optim/lr": 0.002898416938983283, "optim/total_tokens": 3009675264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 408639, "loss/out": 4.075835227966309, "created_at": "2025-01-15T05:10:44.684644+00:00"} {"global_step": 11482, "acc_step": 0, "speed/wps": 12953.409354267453, "speed/FLOPS": 203451093758514.06, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1901154518127441, "optim/lr": 0.0028983859426729647, "optim/total_tokens": 3009937408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461913, "loss/out": 4.027807235717773, "created_at": "2025-01-15T05:10:49.750454+00:00"} {"global_step": 11483, "acc_step": 0, "speed/wps": 12953.278992880956, "speed/FLOPS": 203449046253803.8, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.2377864122390747, "optim/lr": 0.002898354941800166, "optim/total_tokens": 3010199552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452325, "loss/out": 3.99514102935791, "created_at": "2025-01-15T05:10:54.816075+00:00"} {"global_step": 11484, "acc_step": 0, "speed/wps": 12959.22393342523, "speed/FLOPS": 203542419714256.0, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.001, "optim/grad_norm": 1.3018784523010254, "optim/lr": 0.0028983239363649885, "optim/total_tokens": 3010461696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466683, "loss/out": 3.9261345863342285, "created_at": "2025-01-15T05:10:59.881862+00:00"} {"global_step": 11485, "acc_step": 0, "speed/wps": 12962.65752599467, "speed/FLOPS": 203596349003809.7, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8476906418800354, "optim/lr": 0.002898292926367534, "optim/total_tokens": 3010723840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449409, "loss/out": 4.005050182342529, "created_at": "2025-01-15T05:11:04.946687+00:00"} {"global_step": 11486, "acc_step": 0, "speed/wps": 12950.255653515398, "speed/FLOPS": 203401560554564.06, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 9.957806587219238, "optim/lr": 0.0028982619118079036, "optim/total_tokens": 3010985984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 305105, "loss/out": 3.977092742919922, "created_at": "2025-01-15T05:11:10.015606+00:00"} {"global_step": 11487, "acc_step": 0, "speed/wps": 12956.029885548534, "speed/FLOPS": 203492252803270.78, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.9483500719070435, "optim/lr": 0.002898230892686198, "optim/total_tokens": 3011248128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444881, "loss/out": 4.075425148010254, "created_at": "2025-01-15T05:11:15.081567+00:00"} {"global_step": 11488, "acc_step": 0, "speed/wps": 12949.250674301893, "speed/FLOPS": 203385775975029.8, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3090046644210815, "optim/lr": 0.002898199869002518, "optim/total_tokens": 3011510272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450227, "loss/out": 4.136597156524658, "created_at": "2025-01-15T05:11:20.147799+00:00"} {"global_step": 11489, "acc_step": 0, "speed/wps": 12959.941179827954, "speed/FLOPS": 203553685054610.62, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6019121408462524, "optim/lr": 0.0028981688407569654, "optim/total_tokens": 3011772416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466921, "loss/out": 4.124680519104004, "created_at": "2025-01-15T05:11:25.212604+00:00"} {"global_step": 11490, "acc_step": 0, "speed/wps": 12961.826886796745, "speed/FLOPS": 203583302673788.4, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3568545579910278, "optim/lr": 0.0028981378079496415, "optim/total_tokens": 3012034560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454995, "loss/out": 4.211489677429199, "created_at": "2025-01-15T05:11:30.278934+00:00"} {"global_step": 11491, "acc_step": 0, "speed/wps": 12954.47108704174, "speed/FLOPS": 203467769730707.97, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2124757766723633, "optim/lr": 0.0028981067705806476, "optim/total_tokens": 3012296704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 4.013756275177002, "created_at": "2025-01-15T05:11:35.344949+00:00"} {"global_step": 11492, "acc_step": 0, "speed/wps": 12953.284560464088, "speed/FLOPS": 203449133700347.56, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7226746678352356, "optim/lr": 0.002898075728650085, "optim/total_tokens": 3012558848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 328073, "loss/out": 4.028449058532715, "created_at": "2025-01-15T05:11:40.410028+00:00"} {"global_step": 11493, "acc_step": 0, "speed/wps": 12958.322398950873, "speed/FLOPS": 203528259876498.06, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8908248543739319, "optim/lr": 0.0028980446821580547, "optim/total_tokens": 3012820992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 351344, "loss/out": 3.8927865028381348, "created_at": "2025-01-15T05:11:45.471104+00:00"} {"global_step": 11494, "acc_step": 0, "speed/wps": 12953.27581447269, "speed/FLOPS": 203448996332534.4, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0407088994979858, "optim/lr": 0.002898013631104658, "optim/total_tokens": 3013083136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 433718, "loss/out": 3.9364728927612305, "created_at": "2025-01-15T05:11:50.537785+00:00"} {"global_step": 11495, "acc_step": 0, "speed/wps": 12954.70550215175, "speed/FLOPS": 203471451542130.78, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5921068787574768, "optim/lr": 0.002897982575489996, "optim/total_tokens": 3013345280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487337, "loss/out": 4.1090407371521, "created_at": "2025-01-15T05:11:55.605860+00:00"} {"global_step": 11496, "acc_step": 0, "speed/wps": 12955.778560635477, "speed/FLOPS": 203488305400154.6, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.7923076152801514, "optim/lr": 0.00289795151531417, "optim/total_tokens": 3013607424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453945, "loss/out": 3.9288244247436523, "created_at": "2025-01-15T05:12:00.669217+00:00"} {"global_step": 11497, "acc_step": 0, "speed/wps": 12956.058112234554, "speed/FLOPS": 203492696142162.25, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.7209144830703735, "optim/lr": 0.002897920450577282, "optim/total_tokens": 3013869568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469564, "loss/out": 3.9150891304016113, "created_at": "2025-01-15T05:12:05.738138+00:00"} {"global_step": 11498, "acc_step": 0, "speed/wps": 12959.591999330429, "speed/FLOPS": 203548200695072.72, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0273585319519043, "optim/lr": 0.0028978893812794337, "optim/total_tokens": 3014131712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323942, "loss/out": 4.00632381439209, "created_at": "2025-01-15T05:12:10.802423+00:00"} {"global_step": 11499, "acc_step": 0, "speed/wps": 12956.96634102507, "speed/FLOPS": 203506961123354.47, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.990795910358429, "optim/lr": 0.0028978583074207253, "optim/total_tokens": 3014393856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401411, "loss/out": 3.909674644470215, "created_at": "2025-01-15T05:12:15.865197+00:00"} {"global_step": 11500, "acc_step": 0, "speed/wps": 12962.7111837488, "speed/FLOPS": 203597191772571.78, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.992140531539917, "optim/lr": 0.0028978272290012584, "optim/total_tokens": 3014656000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446505, "loss/out": 4.080747127532959, "created_at": "2025-01-15T05:12:20.924686+00:00"} {"global_step": 11501, "acc_step": 0, "speed/wps": 12950.622263938452, "speed/FLOPS": 203407318675033.66, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.253260850906372, "optim/lr": 0.002897796146021135, "optim/total_tokens": 3014918144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445976, "loss/out": 4.033944606781006, "created_at": "2025-01-15T05:12:25.995633+00:00"} {"global_step": 11502, "acc_step": 0, "speed/wps": 12960.989423877434, "speed/FLOPS": 203570149167846.0, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1487672328948975, "optim/lr": 0.002897765058480456, "optim/total_tokens": 3015180288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483905, "loss/out": 4.053755283355713, "created_at": "2025-01-15T05:12:31.060405+00:00"} {"global_step": 11503, "acc_step": 0, "speed/wps": 12959.251274281361, "speed/FLOPS": 203542849139971.8, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.001, "optim/grad_norm": 1.490186333656311, "optim/lr": 0.002897733966379323, "optim/total_tokens": 3015442432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439820, "loss/out": 4.132365703582764, "created_at": "2025-01-15T05:12:36.126944+00:00"} {"global_step": 11504, "acc_step": 0, "speed/wps": 12967.282372040385, "speed/FLOPS": 203668988566162.1, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.225417971611023, "optim/lr": 0.0028977028697178375, "optim/total_tokens": 3015704576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 496035, "loss/out": 4.0308098793029785, "created_at": "2025-01-15T05:12:41.184833+00:00"} {"global_step": 11505, "acc_step": 0, "speed/wps": 12957.106446330792, "speed/FLOPS": 203509161669705.7, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.015142798423767, "optim/lr": 0.0028976717684961, "optim/total_tokens": 3015966720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449420, "loss/out": 4.059372425079346, "created_at": "2025-01-15T05:12:46.246388+00:00"} {"global_step": 11506, "acc_step": 0, "speed/wps": 12961.290438257261, "speed/FLOPS": 203574877012318.28, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0744736194610596, "optim/lr": 0.0028976406627142137, "optim/total_tokens": 3016228864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 308786, "loss/out": 4.167376518249512, "created_at": "2025-01-15T05:12:51.310347+00:00"} {"global_step": 11507, "acc_step": 0, "speed/wps": 12961.869060632498, "speed/FLOPS": 203583965071831.84, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0868875980377197, "optim/lr": 0.0028976095523722796, "optim/total_tokens": 3016491008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 393700, "loss/out": 4.046452045440674, "created_at": "2025-01-15T05:12:56.369952+00:00"} {"global_step": 11508, "acc_step": 0, "speed/wps": 12958.981009161495, "speed/FLOPS": 203538604254881.12, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0764933824539185, "optim/lr": 0.0028975784374703985, "optim/total_tokens": 3016753152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448606, "loss/out": 4.1019134521484375, "created_at": "2025-01-15T05:13:01.435397+00:00"} {"global_step": 11509, "acc_step": 0, "speed/wps": 12957.487159181688, "speed/FLOPS": 203515141288183.28, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8840617537498474, "optim/lr": 0.0028975473180086722, "optim/total_tokens": 3017015296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465633, "loss/out": 4.133766174316406, "created_at": "2025-01-15T05:13:06.496884+00:00"} {"global_step": 11510, "acc_step": 0, "speed/wps": 12954.783370521938, "speed/FLOPS": 203472674571886.94, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.020068883895874, "optim/lr": 0.0028975161939872023, "optim/total_tokens": 3017277440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 461913, "loss/out": 3.847346305847168, "created_at": "2025-01-15T05:13:11.564857+00:00"} {"global_step": 11511, "acc_step": 0, "speed/wps": 12960.138432628919, "speed/FLOPS": 203556783180903.53, "speed/curr_iter_time": 1.2766, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.316900372505188, "optim/lr": 0.0028974850654060903, "optim/total_tokens": 3017539584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451275, "loss/out": 4.018394470214844, "created_at": "2025-01-15T05:13:16.626061+00:00"} {"global_step": 11512, "acc_step": 0, "speed/wps": 12957.395551187266, "speed/FLOPS": 203513702458748.3, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0411350727081299, "optim/lr": 0.0028974539322654384, "optim/total_tokens": 3017801728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307153, "loss/out": 3.9456005096435547, "created_at": "2025-01-15T05:13:21.688510+00:00"} {"global_step": 11513, "acc_step": 0, "speed/wps": 12956.863269830144, "speed/FLOPS": 203505342248603.47, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2786036729812622, "optim/lr": 0.002897422794565347, "optim/total_tokens": 3018063872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 352401, "loss/out": 3.9094738960266113, "created_at": "2025-01-15T05:13:26.755143+00:00"} {"global_step": 11514, "acc_step": 0, "speed/wps": 12959.278748903142, "speed/FLOPS": 203543280666660.97, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7863873839378357, "optim/lr": 0.0028973916523059185, "optim/total_tokens": 3018326016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 393877, "loss/out": 3.9984500408172607, "created_at": "2025-01-15T05:13:31.818818+00:00"} {"global_step": 11515, "acc_step": 0, "speed/wps": 12954.049148344071, "speed/FLOPS": 203461142603656.53, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9329281449317932, "optim/lr": 0.0028973605054872544, "optim/total_tokens": 3018588160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445691, "loss/out": 4.035899639129639, "created_at": "2025-01-15T05:13:36.884796+00:00"} {"global_step": 11516, "acc_step": 0, "speed/wps": 12955.181009872902, "speed/FLOPS": 203478920044307.56, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0286674499511719, "optim/lr": 0.002897329354109456, "optim/total_tokens": 3018850304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 3.9962711334228516, "created_at": "2025-01-15T05:13:41.947131+00:00"} {"global_step": 11517, "acc_step": 0, "speed/wps": 12955.397463987367, "speed/FLOPS": 203482319753614.34, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.2419748306274414, "optim/lr": 0.002897298198172625, "optim/total_tokens": 3019112448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442493, "loss/out": 4.058623313903809, "created_at": "2025-01-15T05:13:47.013384+00:00"} {"global_step": 11518, "acc_step": 0, "speed/wps": 12944.810893934327, "speed/FLOPS": 203316043123459.66, "speed/curr_iter_time": 1.2783, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0747764110565186, "optim/lr": 0.002897267037676863, "optim/total_tokens": 3019374592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484902, "loss/out": 3.9953222274780273, "created_at": "2025-01-15T05:13:52.083939+00:00"} {"global_step": 11519, "acc_step": 0, "speed/wps": 12954.339733673209, "speed/FLOPS": 203465706645556.72, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8499502539634705, "optim/lr": 0.0028972358726222732, "optim/total_tokens": 3019636736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444881, "loss/out": 4.064611434936523, "created_at": "2025-01-15T05:13:57.147024+00:00"} {"global_step": 11520, "acc_step": 0, "speed/wps": 12957.717756040272, "speed/FLOPS": 203518763128720.62, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6447697877883911, "optim/lr": 0.002897204703008955, "optim/total_tokens": 3019898880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 345847, "loss/out": 4.092446804046631, "created_at": "2025-01-15T05:14:02.208323+00:00"} {"global_step": 11521, "acc_step": 0, "speed/wps": 12956.86472989256, "speed/FLOPS": 203505365180890.06, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0579005479812622, "optim/lr": 0.002897173528837011, "optim/total_tokens": 3020161024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444357, "loss/out": 4.137722015380859, "created_at": "2025-01-15T05:14:07.270847+00:00"} {"global_step": 11522, "acc_step": 0, "speed/wps": 12956.086983111461, "speed/FLOPS": 203493149598956.8, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0613867044448853, "optim/lr": 0.002897142350106543, "optim/total_tokens": 3020423168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 4.165012359619141, "created_at": "2025-01-15T05:14:12.333019+00:00"} {"global_step": 11523, "acc_step": 0, "speed/wps": 12956.403784017199, "speed/FLOPS": 203498125392511.5, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0422465801239014, "optim/lr": 0.0028971111668176525, "optim/total_tokens": 3020685312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446742, "loss/out": 4.089639186859131, "created_at": "2025-01-15T05:14:17.395984+00:00"} {"global_step": 11524, "acc_step": 0, "speed/wps": 12953.417916836219, "speed/FLOPS": 203451228245422.72, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7018312811851501, "optim/lr": 0.0028970799789704424, "optim/total_tokens": 3020947456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474376, "loss/out": 4.0242600440979, "created_at": "2025-01-15T05:14:22.461967+00:00"} {"global_step": 11525, "acc_step": 0, "speed/wps": 12950.671987139614, "speed/FLOPS": 203408099646234.75, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0149569511413574, "optim/lr": 0.0028970487865650125, "optim/total_tokens": 3021209600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315181, "loss/out": 4.079984188079834, "created_at": "2025-01-15T05:14:27.528116+00:00"} {"global_step": 11526, "acc_step": 0, "speed/wps": 12947.680920618364, "speed/FLOPS": 203361120836362.4, "speed/curr_iter_time": 1.2769, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1319804191589355, "optim/lr": 0.0028970175896014657, "optim/total_tokens": 3021471744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357724, "loss/out": 4.042980670928955, "created_at": "2025-01-15T05:14:32.598961+00:00"} {"global_step": 11527, "acc_step": 0, "speed/wps": 12954.141066734739, "speed/FLOPS": 203462586308291.1, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.641699492931366, "optim/lr": 0.002896986388079904, "optim/total_tokens": 3021733888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437432, "loss/out": 4.018034934997559, "created_at": "2025-01-15T05:14:37.663039+00:00"} {"global_step": 11528, "acc_step": 0, "speed/wps": 12957.815348087526, "speed/FLOPS": 203520295945928.28, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8448358774185181, "optim/lr": 0.0028969551820004287, "optim/total_tokens": 3021996032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444067, "loss/out": 3.9539642333984375, "created_at": "2025-01-15T05:14:42.724888+00:00"} {"global_step": 11529, "acc_step": 0, "speed/wps": 12959.103779176261, "speed/FLOPS": 203540532526666.72, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.3765701055526733, "optim/lr": 0.0028969239713631414, "optim/total_tokens": 3022258176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 492608, "loss/out": 4.1148152351379395, "created_at": "2025-01-15T05:14:47.785644+00:00"} {"global_step": 11530, "acc_step": 0, "speed/wps": 12947.875062991952, "speed/FLOPS": 203364170109119.16, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6068272590637207, "optim/lr": 0.002896892756168145, "optim/total_tokens": 3022520320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434230, "loss/out": 4.00960636138916, "created_at": "2025-01-15T05:14:52.852994+00:00"} {"global_step": 11531, "acc_step": 0, "speed/wps": 12955.767528525068, "speed/FLOPS": 203488132125700.9, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0013, "optim/grad_norm": 1.794434905052185, "optim/lr": 0.00289686153641554, "optim/total_tokens": 3022782464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484954, "loss/out": 4.0975518226623535, "created_at": "2025-01-15T05:14:57.915008+00:00"} {"global_step": 11532, "acc_step": 0, "speed/wps": 12958.78035035952, "speed/FLOPS": 203535452632662.03, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5961701273918152, "optim/lr": 0.0028968303121054296, "optim/total_tokens": 3023044608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444641, "loss/out": 4.023623466491699, "created_at": "2025-01-15T05:15:02.978535+00:00"} {"global_step": 11533, "acc_step": 0, "speed/wps": 12952.41788726103, "speed/FLOPS": 203435521406759.75, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9861136078834534, "optim/lr": 0.002896799083237914, "optim/total_tokens": 3023306752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 344385, "loss/out": 4.209495544433594, "created_at": "2025-01-15T05:15:08.042090+00:00"} {"global_step": 11534, "acc_step": 0, "speed/wps": 12960.936236335941, "speed/FLOPS": 203569313784419.5, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9043344259262085, "optim/lr": 0.002896767849813097, "optim/total_tokens": 3023568896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439302, "loss/out": 4.103564739227295, "created_at": "2025-01-15T05:15:13.106388+00:00"} {"global_step": 11535, "acc_step": 0, "speed/wps": 12959.310610254788, "speed/FLOPS": 203543781092970.66, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 3.449568748474121, "optim/lr": 0.0028967366118310793, "optim/total_tokens": 3023831040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 503465, "loss/out": 4.160978317260742, "created_at": "2025-01-15T05:15:18.167900+00:00"} {"global_step": 11536, "acc_step": 0, "speed/wps": 12957.369628232942, "speed/FLOPS": 203513295303128.78, "speed/curr_iter_time": 1.2738, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.726893961429596, "optim/lr": 0.0028967053692919627, "optim/total_tokens": 3024093184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435044, "loss/out": 4.066986560821533, "created_at": "2025-01-15T05:15:23.232414+00:00"} {"global_step": 11537, "acc_step": 0, "speed/wps": 12956.387409599152, "speed/FLOPS": 203497868209775.25, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1267162561416626, "optim/lr": 0.00289667412219585, "optim/total_tokens": 3024355328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484381, "loss/out": 4.034724235534668, "created_at": "2025-01-15T05:15:28.300619+00:00"} {"global_step": 11538, "acc_step": 0, "speed/wps": 12951.479190488952, "speed/FLOPS": 203420777884049.72, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8257303237915039, "optim/lr": 0.0028966428705428423, "optim/total_tokens": 3024617472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 306620, "loss/out": 4.0251145362854, "created_at": "2025-01-15T05:15:33.364175+00:00"} {"global_step": 11539, "acc_step": 0, "speed/wps": 12956.414533329871, "speed/FLOPS": 203498294225238.12, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0712082386016846, "optim/lr": 0.0028966116143330422, "optim/total_tokens": 3024879616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 357724, "loss/out": 4.034178733825684, "created_at": "2025-01-15T05:15:38.430654+00:00"} {"global_step": 11540, "acc_step": 0, "speed/wps": 12957.956061127656, "speed/FLOPS": 203522506037583.5, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.007663607597351, "optim/lr": 0.002896580353566551, "optim/total_tokens": 3025141760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441442, "loss/out": 4.0535383224487305, "created_at": "2025-01-15T05:15:43.494489+00:00"} {"global_step": 11541, "acc_step": 0, "speed/wps": 12959.296101983926, "speed/FLOPS": 203543553220640.3, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0011, "optim/grad_norm": 1.0367659330368042, "optim/lr": 0.0028965490882434716, "optim/total_tokens": 3025403904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439583, "loss/out": 3.9653215408325195, "created_at": "2025-01-15T05:15:48.560501+00:00"} {"global_step": 11542, "acc_step": 0, "speed/wps": 12958.860601679618, "speed/FLOPS": 203536713089921.03, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9058926105499268, "optim/lr": 0.002896517818363905, "optim/total_tokens": 3025666048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 4.055315971374512, "created_at": "2025-01-15T05:15:53.622623+00:00"} {"global_step": 11543, "acc_step": 0, "speed/wps": 12958.359567443225, "speed/FLOPS": 203528843658745.3, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.6907435655593872, "optim/lr": 0.0028964865439279545, "optim/total_tokens": 3025928192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434515, "loss/out": 4.088491439819336, "created_at": "2025-01-15T05:15:58.684603+00:00"} {"global_step": 11544, "acc_step": 0, "speed/wps": 12955.075348279022, "speed/FLOPS": 203477260483781.44, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1175135374069214, "optim/lr": 0.002896455264935721, "optim/total_tokens": 3026190336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 498705, "loss/out": 4.1096086502075195, "created_at": "2025-01-15T05:16:03.748523+00:00"} {"global_step": 11545, "acc_step": 0, "speed/wps": 12952.91674068161, "speed/FLOPS": 203443356585226.4, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9817279577255249, "optim/lr": 0.002896423981387307, "optim/total_tokens": 3026452480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453144, "loss/out": 4.069372177124023, "created_at": "2025-01-15T05:16:08.813793+00:00"} {"global_step": 11546, "acc_step": 0, "speed/wps": 12956.294128210771, "speed/FLOPS": 203496403097388.7, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0854171514511108, "optim/lr": 0.0028963926932828138, "optim/total_tokens": 3026714624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331148, "loss/out": 4.0289740562438965, "created_at": "2025-01-15T05:16:13.905756+00:00"} {"global_step": 11547, "acc_step": 0, "speed/wps": 12960.414011815488, "speed/FLOPS": 203561111530714.1, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0011, "optim/grad_norm": 1.1423357725143433, "optim/lr": 0.002896361400622345, "optim/total_tokens": 3026976768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 457902, "loss/out": 4.12453031539917, "created_at": "2025-01-15T05:16:18.970053+00:00"} {"global_step": 11548, "acc_step": 0, "speed/wps": 12962.858426671406, "speed/FLOPS": 203599504425004.28, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8998175859451294, "optim/lr": 0.002896330103406002, "optim/total_tokens": 3027238912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 469320, "loss/out": 4.005669116973877, "created_at": "2025-01-15T05:16:24.035020+00:00"} {"global_step": 11549, "acc_step": 0, "speed/wps": 12962.190614114848, "speed/FLOPS": 203589015511132.94, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.5270791053771973, "optim/lr": 0.0028962988016338863, "optim/total_tokens": 3027501056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455517, "loss/out": 4.0734734535217285, "created_at": "2025-01-15T05:16:29.098794+00:00"} {"global_step": 11550, "acc_step": 0, "speed/wps": 12954.2880649451, "speed/FLOPS": 203464895117181.5, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6574760675430298, "optim/lr": 0.0028962674953061005, "optim/total_tokens": 3027763200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 320810, "loss/out": 4.021656036376953, "created_at": "2025-01-15T05:16:34.167768+00:00"} {"global_step": 11551, "acc_step": 0, "speed/wps": 12959.90094233057, "speed/FLOPS": 203553053069422.12, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.001, "optim/grad_norm": 0.8047965168952942, "optim/lr": 0.002896236184422747, "optim/total_tokens": 3028025344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434807, "loss/out": 4.126130104064941, "created_at": "2025-01-15T05:16:39.235689+00:00"} {"global_step": 11552, "acc_step": 0, "speed/wps": 12955.235376142768, "speed/FLOPS": 203479773941282.88, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0528652667999268, "optim/lr": 0.0028962048689839272, "optim/total_tokens": 3028287488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455808, "loss/out": 3.9685990810394287, "created_at": "2025-01-15T05:16:44.303322+00:00"} {"global_step": 11553, "acc_step": 0, "speed/wps": 12949.151312063113, "speed/FLOPS": 203384215354531.75, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.0936720371246338, "optim/lr": 0.002896173548989744, "optim/total_tokens": 3028549632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 4.035040378570557, "created_at": "2025-01-15T05:16:49.373414+00:00"} {"global_step": 11554, "acc_step": 0, "speed/wps": 12949.581984921977, "speed/FLOPS": 203390979663586.75, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0803698301315308, "optim/lr": 0.0028961422244403, "optim/total_tokens": 3028811776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 458193, "loss/out": 4.040048599243164, "created_at": "2025-01-15T05:16:54.438779+00:00"} {"global_step": 11555, "acc_step": 0, "speed/wps": 12961.30736222652, "speed/FLOPS": 203575142826511.47, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1796631813049316, "optim/lr": 0.0028961108953356966, "optim/total_tokens": 3029073920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447269, "loss/out": 4.01390266418457, "created_at": "2025-01-15T05:16:59.502313+00:00"} {"global_step": 11556, "acc_step": 0, "speed/wps": 12962.129508771815, "speed/FLOPS": 203588055767753.06, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.055461049079895, "optim/lr": 0.0028960795616760357, "optim/total_tokens": 3029336064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 315474, "loss/out": 4.078447341918945, "created_at": "2025-01-15T05:17:04.562905+00:00"} {"global_step": 11557, "acc_step": 0, "speed/wps": 12956.545728386878, "speed/FLOPS": 203500354823889.3, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.1059072017669678, "optim/lr": 0.0028960482234614204, "optim/total_tokens": 3029598208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466392, "loss/out": 4.014265537261963, "created_at": "2025-01-15T05:17:09.626955+00:00"} {"global_step": 11558, "acc_step": 0, "speed/wps": 12962.222929122388, "speed/FLOPS": 203589523062731.8, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.3690361976623535, "optim/lr": 0.0028960168806919525, "optim/total_tokens": 3029860352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473279, "loss/out": 4.047821044921875, "created_at": "2025-01-15T05:17:14.695620+00:00"} {"global_step": 11559, "acc_step": 0, "speed/wps": 12963.776937732475, "speed/FLOPS": 203613930903384.84, "speed/curr_iter_time": 1.2737, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9013211131095886, "optim/lr": 0.0028959855333677343, "optim/total_tokens": 3030122496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 434807, "loss/out": 4.004070281982422, "created_at": "2025-01-15T05:17:19.757462+00:00"} {"global_step": 11560, "acc_step": 0, "speed/wps": 12957.847407447123, "speed/FLOPS": 203520799482224.94, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.401713252067566, "optim/lr": 0.002895954181488868, "optim/total_tokens": 3030384640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483043, "loss/out": 3.9937615394592285, "created_at": "2025-01-15T05:17:24.824408+00:00"} {"global_step": 11561, "acc_step": 0, "speed/wps": 12958.969637486729, "speed/FLOPS": 203538425647102.62, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.821727991104126, "optim/lr": 0.0028959228250554562, "optim/total_tokens": 3030646784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 331148, "loss/out": 4.108047008514404, "created_at": "2025-01-15T05:17:29.886236+00:00"} {"global_step": 11562, "acc_step": 0, "speed/wps": 12953.196350004217, "speed/FLOPS": 203447748233861.4, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1357961893081665, "optim/lr": 0.002895891464067601, "optim/total_tokens": 3030908928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 382664, "loss/out": 4.020298004150391, "created_at": "2025-01-15T05:17:34.950714+00:00"} {"global_step": 11563, "acc_step": 0, "speed/wps": 12960.779944358923, "speed/FLOPS": 203566859004152.75, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8833646774291992, "optim/lr": 0.002895860098525405, "optim/total_tokens": 3031171072, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450756, "loss/out": 4.160250186920166, "created_at": "2025-01-15T05:17:40.010796+00:00"} {"global_step": 11564, "acc_step": 0, "speed/wps": 12957.910500615584, "speed/FLOPS": 203521790447134.72, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.89058256149292, "optim/lr": 0.0028958287284289695, "optim/total_tokens": 3031433216, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 484665, "loss/out": 4.024605751037598, "created_at": "2025-01-15T05:17:45.073444+00:00"} {"global_step": 11565, "acc_step": 0, "speed/wps": 12960.45070535064, "speed/FLOPS": 203561687853105.88, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9435497522354126, "optim/lr": 0.0028957973537783984, "optim/total_tokens": 3031695360, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 452847, "loss/out": 3.932935953140259, "created_at": "2025-01-15T05:17:50.140248+00:00"} {"global_step": 11566, "acc_step": 0, "speed/wps": 12959.940998426884, "speed/FLOPS": 203553682205457.56, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0012, "optim/grad_norm": 1.1652733087539673, "optim/lr": 0.0028957659745737926, "optim/total_tokens": 3031957504, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482856, "loss/out": 4.121432304382324, "created_at": "2025-01-15T05:17:55.202529+00:00"} {"global_step": 11567, "acc_step": 0, "speed/wps": 12956.466118828215, "speed/FLOPS": 203499104446375.5, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1571227312088013, "optim/lr": 0.0028957345908152557, "optim/total_tokens": 3032219648, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456330, "loss/out": 4.057937145233154, "created_at": "2025-01-15T05:18:00.265403+00:00"} {"global_step": 11568, "acc_step": 0, "speed/wps": 12962.665050136075, "speed/FLOPS": 203596467180789.72, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7648619413375854, "optim/lr": 0.0028957032025028894, "optim/total_tokens": 3032481792, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 424156, "loss/out": 4.100748062133789, "created_at": "2025-01-15T05:18:05.328720+00:00"} {"global_step": 11569, "acc_step": 0, "speed/wps": 12957.867739309853, "speed/FLOPS": 203521118822068.0, "speed/curr_iter_time": 1.273, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6835619211196899, "optim/lr": 0.0028956718096367962, "optim/total_tokens": 3032743936, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 314942, "loss/out": 3.932698965072632, "created_at": "2025-01-15T05:18:10.390147+00:00"} {"global_step": 11570, "acc_step": 0, "speed/wps": 12953.108267115764, "speed/FLOPS": 203446364771059.7, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4892021715641022, "optim/lr": 0.002895640412217079, "optim/total_tokens": 3033006080, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 430776, "loss/out": 4.039484024047852, "created_at": "2025-01-15T05:18:15.457121+00:00"} {"global_step": 11571, "acc_step": 0, "speed/wps": 12957.384616531299, "speed/FLOPS": 203513530714950.56, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.42749491333961487, "optim/lr": 0.002895609010243839, "optim/total_tokens": 3033268224, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471418, "loss/out": 4.007081031799316, "created_at": "2025-01-15T05:18:20.524502+00:00"} {"global_step": 11572, "acc_step": 0, "speed/wps": 12956.57333944609, "speed/FLOPS": 203500788493515.78, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8247581124305725, "optim/lr": 0.00289557760371718, "optim/total_tokens": 3033530368, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441679, "loss/out": 4.035388946533203, "created_at": "2025-01-15T05:18:25.587262+00:00"} {"global_step": 11573, "acc_step": 0, "speed/wps": 12959.703596514382, "speed/FLOPS": 203549953482197.3, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5781872868537903, "optim/lr": 0.002895546192637204, "optim/total_tokens": 3033792512, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488426, "loss/out": 4.105475425720215, "created_at": "2025-01-15T05:18:30.655132+00:00"} {"global_step": 11574, "acc_step": 0, "speed/wps": 12954.27326266735, "speed/FLOPS": 203464662627068.97, "speed/curr_iter_time": 1.2761, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5783004760742188, "optim/lr": 0.0028955147770040133, "optim/total_tokens": 3034054656, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435044, "loss/out": 4.099733352661133, "created_at": "2025-01-15T05:18:35.719972+00:00"} {"global_step": 11575, "acc_step": 0, "speed/wps": 12961.973161751377, "speed/FLOPS": 203585600122953.75, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.001, "optim/grad_norm": 0.44097280502319336, "optim/lr": 0.0028954833568177107, "optim/total_tokens": 3034316800, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 329132, "loss/out": 4.104394435882568, "created_at": "2025-01-15T05:18:40.781002+00:00"} {"global_step": 11576, "acc_step": 0, "speed/wps": 12956.30808824818, "speed/FLOPS": 203496622358959.3, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7983359694480896, "optim/lr": 0.0028954519320783987, "optim/total_tokens": 3034578944, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450227, "loss/out": 4.032730579376221, "created_at": "2025-01-15T05:18:45.845433+00:00"} {"global_step": 11577, "acc_step": 0, "speed/wps": 12963.376126885805, "speed/FLOPS": 203607635618266.06, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6392619013786316, "optim/lr": 0.002895420502786179, "optim/total_tokens": 3034841088, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444644, "loss/out": 4.065844535827637, "created_at": "2025-01-15T05:18:50.910397+00:00"} {"global_step": 11578, "acc_step": 0, "speed/wps": 12961.148392295861, "speed/FLOPS": 203572645985302.88, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.021380066871643, "optim/lr": 0.0028953890689411557, "optim/total_tokens": 3035103232, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463750, "loss/out": 4.044981479644775, "created_at": "2025-01-15T05:18:55.976654+00:00"} {"global_step": 11579, "acc_step": 0, "speed/wps": 12960.096964646056, "speed/FLOPS": 203556131869249.7, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4451074004173279, "optim/lr": 0.00289535763054343, "optim/total_tokens": 3035365376, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453945, "loss/out": 4.105197906494141, "created_at": "2025-01-15T05:19:01.050015+00:00"} {"global_step": 11580, "acc_step": 0, "speed/wps": 12964.388872262649, "speed/FLOPS": 203623542176067.38, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4807188808917999, "optim/lr": 0.0028953261875931053, "optim/total_tokens": 3035627520, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448605, "loss/out": 3.9205424785614014, "created_at": "2025-01-15T05:19:06.113323+00:00"} {"global_step": 11581, "acc_step": 0, "speed/wps": 12961.066723246027, "speed/FLOPS": 203571363260650.22, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.47407934069633484, "optim/lr": 0.0028952947400902834, "optim/total_tokens": 3035889664, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448605, "loss/out": 3.9754583835601807, "created_at": "2025-01-15T05:19:11.173518+00:00"} {"global_step": 11582, "acc_step": 0, "speed/wps": 12957.75353715466, "speed/FLOPS": 203519325120290.5, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0006, "optim/grad_norm": 1.0754590034484863, "optim/lr": 0.002895263288035067, "optim/total_tokens": 3036151808, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435567, "loss/out": 3.925413131713867, "created_at": "2025-01-15T05:19:16.234756+00:00"} {"global_step": 11583, "acc_step": 0, "speed/wps": 12960.54751502672, "speed/FLOPS": 203563208382099.12, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5543355941772461, "optim/lr": 0.00289523183142756, "optim/total_tokens": 3036413952, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293489, "loss/out": 3.9158785343170166, "created_at": "2025-01-15T05:19:21.300245+00:00"} {"global_step": 11584, "acc_step": 0, "speed/wps": 12962.310594002225, "speed/FLOPS": 203590899960133.84, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.44859182834625244, "optim/lr": 0.002895200370267864, "optim/total_tokens": 3036676096, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467205, "loss/out": 4.088061332702637, "created_at": "2025-01-15T05:19:26.363212+00:00"} {"global_step": 11585, "acc_step": 0, "speed/wps": 12959.543145587093, "speed/FLOPS": 203547433379901.53, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5586681962013245, "optim/lr": 0.002895168904556081, "optim/total_tokens": 3036938240, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466683, "loss/out": 3.999152421951294, "created_at": "2025-01-15T05:19:31.428597+00:00"} {"global_step": 11586, "acc_step": 0, "speed/wps": 12960.709146200368, "speed/FLOPS": 203565747021785.78, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6625222563743591, "optim/lr": 0.002895137434292315, "optim/total_tokens": 3037200384, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 435044, "loss/out": 3.9813640117645264, "created_at": "2025-01-15T05:19:36.490448+00:00"} {"global_step": 11587, "acc_step": 0, "speed/wps": 12960.253034872005, "speed/FLOPS": 203558583166611.2, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4235657751560211, "optim/lr": 0.0028951059594766676, "optim/total_tokens": 3037462528, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490992, "loss/out": 4.058318138122559, "created_at": "2025-01-15T05:19:41.554923+00:00"} {"global_step": 11588, "acc_step": 0, "speed/wps": 12959.403347965968, "speed/FLOPS": 203545237666159.72, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3753346800804138, "optim/lr": 0.002895074480109243, "optim/total_tokens": 3037724672, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 449135, "loss/out": 4.016924858093262, "created_at": "2025-01-15T05:19:46.615674+00:00"} {"global_step": 11589, "acc_step": 0, "speed/wps": 12959.467150811779, "speed/FLOPS": 203546239777527.47, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.001, "optim/grad_norm": 0.48714327812194824, "optim/lr": 0.0028950429961901417, "optim/total_tokens": 3037986816, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 338880, "loss/out": 4.085888385772705, "created_at": "2025-01-15T05:19:51.678840+00:00"} {"global_step": 11590, "acc_step": 0, "speed/wps": 12959.16295628544, "speed/FLOPS": 203541461984484.22, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5394766926765442, "optim/lr": 0.0028950115077194683, "optim/total_tokens": 3038248960, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450462, "loss/out": 4.041790962219238, "created_at": "2025-01-15T05:19:56.742277+00:00"} {"global_step": 11591, "acc_step": 0, "speed/wps": 12961.767811894859, "speed/FLOPS": 203582374821276.88, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5357071757316589, "optim/lr": 0.0028949800146973246, "optim/total_tokens": 3038511104, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 482044, "loss/out": 3.9899420738220215, "created_at": "2025-01-15T05:20:01.802130+00:00"} {"global_step": 11592, "acc_step": 0, "speed/wps": 12956.81170979773, "speed/FLOPS": 203504532427443.5, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6462369561195374, "optim/lr": 0.0028949485171238143, "optim/total_tokens": 3038773248, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447026, "loss/out": 4.09864616394043, "created_at": "2025-01-15T05:20:06.867220+00:00"} {"global_step": 11593, "acc_step": 0, "speed/wps": 12957.782210386882, "speed/FLOPS": 203519775472803.7, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.5578219890594482, "optim/lr": 0.002894917014999039, "optim/total_tokens": 3039035392, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 3.992140531539917, "created_at": "2025-01-15T05:20:11.932895+00:00"} {"global_step": 11594, "acc_step": 0, "speed/wps": 12960.734645670544, "speed/FLOPS": 203566147526004.8, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0008, "optim/grad_norm": 2.231508731842041, "optim/lr": 0.002894885508323101, "optim/total_tokens": 3039297536, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488092, "loss/out": 4.0451812744140625, "created_at": "2025-01-15T05:20:16.997312+00:00"} {"global_step": 11595, "acc_step": 0, "speed/wps": 12963.015750244267, "speed/FLOPS": 203601975407901.3, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.6863511204719543, "optim/lr": 0.002894853997096105, "optim/total_tokens": 3039559680, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 454995, "loss/out": 4.178409576416016, "created_at": "2025-01-15T05:20:22.063477+00:00"} {"global_step": 11596, "acc_step": 0, "speed/wps": 12956.581524499303, "speed/FLOPS": 203500917051023.94, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5429850816726685, "optim/lr": 0.002894822481318153, "optim/total_tokens": 3039821824, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292246, "loss/out": 4.051499843597412, "created_at": "2025-01-15T05:20:27.125512+00:00"} {"global_step": 11597, "acc_step": 0, "speed/wps": 12954.028189368948, "speed/FLOPS": 203460813414151.6, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7904726266860962, "optim/lr": 0.002894790960989347, "optim/total_tokens": 3040083968, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444644, "loss/out": 4.050747871398926, "created_at": "2025-01-15T05:20:32.190394+00:00"} {"global_step": 11598, "acc_step": 0, "speed/wps": 12943.765622218461, "speed/FLOPS": 203299625694808.12, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5376467108726501, "optim/lr": 0.002894759436109791, "optim/total_tokens": 3040346112, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 3.968778371810913, "created_at": "2025-01-15T05:20:37.264451+00:00"} {"global_step": 11599, "acc_step": 0, "speed/wps": 12958.85849682381, "speed/FLOPS": 203536680030268.2, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5084638595581055, "optim/lr": 0.002894727906679587, "optim/total_tokens": 3040608256, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 440628, "loss/out": 4.072841167449951, "created_at": "2025-01-15T05:20:42.325704+00:00"} {"global_step": 11600, "acc_step": 0, "speed/wps": 12956.34024272767, "speed/FLOPS": 203497127389244.56, "speed/curr_iter_time": 1.2742, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.40495938062667847, "optim/lr": 0.002894696372698839, "optim/total_tokens": 3040870400, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 490756, "loss/out": 4.014906406402588, "created_at": "2025-01-15T05:20:47.387602+00:00"} {"global_step": 11601, "acc_step": 0, "speed/wps": 12957.509279384074, "speed/FLOPS": 203515488716358.1, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6265839338302612, "optim/lr": 0.002894664834167648, "optim/total_tokens": 3041132544, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439302, "loss/out": 4.15149450302124, "created_at": "2025-01-15T05:20:52.449957+00:00"} {"global_step": 11602, "acc_step": 0, "speed/wps": 12960.469658507029, "speed/FLOPS": 203561985538471.3, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.28384852409362793, "optim/lr": 0.0028946332910861194, "optim/total_tokens": 3041394688, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 307153, "loss/out": 4.116029739379883, "created_at": "2025-01-15T05:20:57.511978+00:00"} {"global_step": 11603, "acc_step": 0, "speed/wps": 12964.943234556655, "speed/FLOPS": 203632249197668.38, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.7765498757362366, "optim/lr": 0.0028946017434543538, "optim/total_tokens": 3041656832, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 396909, "loss/out": 4.001038551330566, "created_at": "2025-01-15T05:21:02.571876+00:00"} {"global_step": 11604, "acc_step": 0, "speed/wps": 12961.89618344361, "speed/FLOPS": 203584391072851.0, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.39781251549720764, "optim/lr": 0.0028945701912724553, "optim/total_tokens": 3041918976, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444067, "loss/out": 3.966487169265747, "created_at": "2025-01-15T05:21:07.634793+00:00"} {"global_step": 11605, "acc_step": 0, "speed/wps": 12954.905956074452, "speed/FLOPS": 203474599946439.38, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6472357511520386, "optim/lr": 0.0028945386345405263, "optim/total_tokens": 3042181120, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456556, "loss/out": 4.098076343536377, "created_at": "2025-01-15T05:21:12.702775+00:00"} {"global_step": 11606, "acc_step": 0, "speed/wps": 12957.958015364391, "speed/FLOPS": 203522536731556.8, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0644314289093018, "optim/lr": 0.0028945070732586706, "optim/total_tokens": 3042443264, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 478328, "loss/out": 4.011335849761963, "created_at": "2025-01-15T05:21:17.764004+00:00"} {"global_step": 11607, "acc_step": 0, "speed/wps": 12957.041488148028, "speed/FLOPS": 203508141412183.47, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.0443915128707886, "optim/lr": 0.0028944755074269903, "optim/total_tokens": 3042705408, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 4.0588884353637695, "created_at": "2025-01-15T05:21:22.828209+00:00"} {"global_step": 11608, "acc_step": 0, "speed/wps": 12957.332851934325, "speed/FLOPS": 203512717680823.06, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0133548974990845, "optim/lr": 0.0028944439370455887, "optim/total_tokens": 3042967552, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485713, "loss/out": 4.043474197387695, "created_at": "2025-01-15T05:21:27.890917+00:00"} {"global_step": 11609, "acc_step": 0, "speed/wps": 12961.819437642318, "speed/FLOPS": 203583185674581.94, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3403420448303223, "optim/lr": 0.0028944123621145695, "optim/total_tokens": 3043229696, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445170, "loss/out": 4.020207405090332, "created_at": "2025-01-15T05:21:32.951587+00:00"} {"global_step": 11610, "acc_step": 0, "speed/wps": 12955.277843400389, "speed/FLOPS": 203480440947919.94, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0014, "optim/grad_norm": 0.5379008054733276, "optim/lr": 0.0028943807826340346, "optim/total_tokens": 3043491840, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 339173, "loss/out": 4.025805473327637, "created_at": "2025-01-15T05:21:38.020625+00:00"} {"global_step": 11611, "acc_step": 0, "speed/wps": 12952.53214558456, "speed/FLOPS": 203437315990738.2, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7755499482154846, "optim/lr": 0.0028943491986040872, "optim/total_tokens": 3043753984, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441442, "loss/out": 3.9921963214874268, "created_at": "2025-01-15T05:21:43.087628+00:00"} {"global_step": 11612, "acc_step": 0, "speed/wps": 12961.228928392786, "speed/FLOPS": 203573910915373.28, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8776954412460327, "optim/lr": 0.0028943176100248314, "optim/total_tokens": 3044016128, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 485997, "loss/out": 4.10076904296875, "created_at": "2025-01-15T05:21:48.148887+00:00"} {"global_step": 11613, "acc_step": 0, "speed/wps": 12959.783162324848, "speed/FLOPS": 203551203172585.53, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8174321055412292, "optim/lr": 0.0028942860168963687, "optim/total_tokens": 3044278272, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 427589, "loss/out": 4.041080474853516, "created_at": "2025-01-15T05:21:53.217534+00:00"} {"global_step": 11614, "acc_step": 0, "speed/wps": 12963.020244440864, "speed/FLOPS": 203602045995434.56, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4886087477207184, "optim/lr": 0.0028942544192188035, "optim/total_tokens": 3044540416, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 3.9665446281433105, "created_at": "2025-01-15T05:21:58.276796+00:00"} {"global_step": 11615, "acc_step": 0, "speed/wps": 12958.881192073872, "speed/FLOPS": 203537036490356.97, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.2953835725784302, "optim/lr": 0.0028942228169922387, "optim/total_tokens": 3044802560, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443016, "loss/out": 4.117856979370117, "created_at": "2025-01-15T05:22:03.338463+00:00"} {"global_step": 11616, "acc_step": 0, "speed/wps": 12959.80812111986, "speed/FLOPS": 203551595184758.2, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.506137490272522, "optim/lr": 0.0028941912102167767, "optim/total_tokens": 3045064704, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 301350, "loss/out": 4.034183502197266, "created_at": "2025-01-15T05:22:08.404144+00:00"} {"global_step": 11617, "acc_step": 0, "speed/wps": 12959.409021054818, "speed/FLOPS": 203545326769815.75, "speed/curr_iter_time": 1.276, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4056442975997925, "optim/lr": 0.002894159598892521, "optim/total_tokens": 3045326848, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437948, "loss/out": 3.930525541305542, "created_at": "2025-01-15T05:22:13.467026+00:00"} {"global_step": 11618, "acc_step": 0, "speed/wps": 12958.38002977241, "speed/FLOPS": 203529165047743.34, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.49034103751182556, "optim/lr": 0.0028941279830195744, "optim/total_tokens": 3045588992, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 464820, "loss/out": 4.084502220153809, "created_at": "2025-01-15T05:22:18.529408+00:00"} {"global_step": 11619, "acc_step": 0, "speed/wps": 12960.229587697304, "speed/FLOPS": 203558214896512.72, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.46155276894569397, "optim/lr": 0.0028940963625980407, "optim/total_tokens": 3045851136, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481232, "loss/out": 4.036633014678955, "created_at": "2025-01-15T05:22:23.592854+00:00"} {"global_step": 11620, "acc_step": 0, "speed/wps": 12955.44839499749, "speed/FLOPS": 203483119695114.88, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.41574421525001526, "optim/lr": 0.0028940647376280225, "optim/total_tokens": 3046113280, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 444641, "loss/out": 3.9655327796936035, "created_at": "2025-01-15T05:22:28.660355+00:00"} {"global_step": 11621, "acc_step": 0, "speed/wps": 12962.873972120204, "speed/FLOPS": 203599748587639.34, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5862637162208557, "optim/lr": 0.0028940331081096235, "optim/total_tokens": 3046375424, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 447839, "loss/out": 4.043416976928711, "created_at": "2025-01-15T05:22:33.728124+00:00"} {"global_step": 11622, "acc_step": 0, "speed/wps": 12954.610010334922, "speed/FLOPS": 203469951711928.0, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6045351624488831, "optim/lr": 0.0028940014740429466, "optim/total_tokens": 3046637568, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 495052, "loss/out": 4.0894341468811035, "created_at": "2025-01-15T05:22:38.794338+00:00"} {"global_step": 11623, "acc_step": 0, "speed/wps": 12957.906522974268, "speed/FLOPS": 203521727972812.0, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5232141017913818, "optim/lr": 0.0028939698354280953, "optim/total_tokens": 3046899712, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443306, "loss/out": 3.969357967376709, "created_at": "2025-01-15T05:22:43.856454+00:00"} {"global_step": 11624, "acc_step": 0, "speed/wps": 12952.742741413516, "speed/FLOPS": 203440623687620.94, "speed/curr_iter_time": 1.2765, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5406038165092468, "optim/lr": 0.002893938192265173, "optim/total_tokens": 3047161856, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326966, "loss/out": 4.098233222961426, "created_at": "2025-01-15T05:22:48.923774+00:00"} {"global_step": 11625, "acc_step": 0, "speed/wps": 12949.84824291415, "speed/FLOPS": 203395161611230.62, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.9430239200592041, "optim/lr": 0.002893906544554282, "optim/total_tokens": 3047424000, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 442778, "loss/out": 4.038548469543457, "created_at": "2025-01-15T05:22:53.988841+00:00"} {"global_step": 11626, "acc_step": 0, "speed/wps": 12964.514102254128, "speed/FLOPS": 203625509085167.62, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0250345468521118, "optim/lr": 0.0028938748922955264, "optim/total_tokens": 3047686144, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483377, "loss/out": 4.001978397369385, "created_at": "2025-01-15T05:22:59.047407+00:00"} {"global_step": 11627, "acc_step": 0, "speed/wps": 12962.727796227955, "speed/FLOPS": 203597452694384.84, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.3106744289398193, "optim/lr": 0.002893843235489008, "optim/total_tokens": 3047948288, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 467466, "loss/out": 3.9251930713653564, "created_at": "2025-01-15T05:23:04.114756+00:00"} {"global_step": 11628, "acc_step": 0, "speed/wps": 12953.538576103023, "speed/FLOPS": 203453123365011.1, "speed/curr_iter_time": 1.2764, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.2138166427612305, "optim/lr": 0.002893811574134832, "optim/total_tokens": 3048210432, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 475662, "loss/out": 4.084625244140625, "created_at": "2025-01-15T05:23:09.181993+00:00"} {"global_step": 11629, "acc_step": 0, "speed/wps": 12955.628463526678, "speed/FLOPS": 203485947918805.88, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1038968563079834, "optim/lr": 0.0028937799082331015, "optim/total_tokens": 3048472576, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453144, "loss/out": 3.9338154792785645, "created_at": "2025-01-15T05:23:14.249787+00:00"} {"global_step": 11630, "acc_step": 0, "speed/wps": 12963.081632803822, "speed/FLOPS": 203603010184031.25, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7926046848297119, "optim/lr": 0.002893748237783919, "optim/total_tokens": 3048734720, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 292246, "loss/out": 4.03373908996582, "created_at": "2025-01-15T05:23:19.311725+00:00"} {"global_step": 11631, "acc_step": 0, "speed/wps": 12958.382225151508, "speed/FLOPS": 203529199529188.84, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5382770895957947, "optim/lr": 0.0028937165627873876, "optim/total_tokens": 3048996864, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448606, "loss/out": 4.126490116119385, "created_at": "2025-01-15T05:23:24.373754+00:00"} {"global_step": 11632, "acc_step": 0, "speed/wps": 12961.784319292114, "speed/FLOPS": 203582634092634.28, "speed/curr_iter_time": 1.2734, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.39516592025756836, "optim/lr": 0.0028936848832436116, "optim/total_tokens": 3049259008, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474645, "loss/out": 3.900852680206299, "created_at": "2025-01-15T05:23:29.434500+00:00"} {"global_step": 11633, "acc_step": 0, "speed/wps": 12960.066655687893, "speed/FLOPS": 203555655825412.88, "speed/curr_iter_time": 1.2746, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1016854047775269, "optim/lr": 0.0028936531991526934, "optim/total_tokens": 3049521152, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 479906, "loss/out": 3.9767935276031494, "created_at": "2025-01-15T05:23:34.495828+00:00"} {"global_step": 11634, "acc_step": 0, "speed/wps": 12950.777231312892, "speed/FLOPS": 203409752650596.72, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.1828492879867554, "optim/lr": 0.002893621510514737, "optim/total_tokens": 3049783296, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 448881, "loss/out": 4.0481953620910645, "created_at": "2025-01-15T05:23:39.561207+00:00"} {"global_step": 11635, "acc_step": 0, "speed/wps": 12948.222564050686, "speed/FLOPS": 203369628090763.06, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0012, "optim/grad_norm": 0.5029914975166321, "optim/lr": 0.0028935898173298464, "optim/total_tokens": 3050045440, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 450225, "loss/out": 4.126755714416504, "created_at": "2025-01-15T05:23:44.630462+00:00"} {"global_step": 11636, "acc_step": 0, "speed/wps": 12960.723516793763, "speed/FLOPS": 203565972731702.3, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.8720329403877258, "optim/lr": 0.0028935581195981236, "optim/total_tokens": 3050307584, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 4.056501865386963, "created_at": "2025-01-15T05:23:49.695739+00:00"} {"global_step": 11637, "acc_step": 0, "speed/wps": 12956.826610919852, "speed/FLOPS": 203504766470042.72, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5727424621582031, "optim/lr": 0.0028935264173196727, "optim/total_tokens": 3050569728, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319909, "loss/out": 4.096817970275879, "created_at": "2025-01-15T05:23:54.758731+00:00"} {"global_step": 11638, "acc_step": 0, "speed/wps": 12955.548278470902, "speed/FLOPS": 203484688502319.1, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5175689458847046, "optim/lr": 0.0028934947104945975, "optim/total_tokens": 3050831872, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 366944, "loss/out": 4.080296993255615, "created_at": "2025-01-15T05:23:59.825707+00:00"} {"global_step": 11639, "acc_step": 0, "speed/wps": 12962.31163743991, "speed/FLOPS": 203590916348756.5, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6892510056495667, "optim/lr": 0.0028934629991230003, "optim/total_tokens": 3051094016, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443022, "loss/out": 4.15316104888916, "created_at": "2025-01-15T05:24:04.887380+00:00"} {"global_step": 11640, "acc_step": 0, "speed/wps": 12964.65344884137, "speed/FLOPS": 203627697714802.53, "speed/curr_iter_time": 1.2732, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0535579919815063, "optim/lr": 0.002893431283204986, "optim/total_tokens": 3051356160, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488670, "loss/out": 4.0007805824279785, "created_at": "2025-01-15T05:24:09.955433+00:00"} {"global_step": 11641, "acc_step": 0, "speed/wps": 12958.830966709193, "speed/FLOPS": 203536247631987.78, "speed/curr_iter_time": 1.2752, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.9732893109321594, "optim/lr": 0.002893399562740657, "optim/total_tokens": 3051618304, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 445404, "loss/out": 3.9933061599731445, "created_at": "2025-01-15T05:24:15.016512+00:00"} {"global_step": 11642, "acc_step": 0, "speed/wps": 12966.249626260855, "speed/FLOPS": 203652767874553.16, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.100157618522644, "optim/lr": 0.002893367837730118, "optim/total_tokens": 3051880448, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 474613, "loss/out": 4.109447956085205, "created_at": "2025-01-15T05:24:20.074575+00:00"} {"global_step": 11643, "acc_step": 0, "speed/wps": 12960.19958603555, "speed/FLOPS": 203557743679188.25, "speed/curr_iter_time": 1.2747, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.172837257385254, "optim/lr": 0.0028933361081734713, "optim/total_tokens": 3052142592, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453660, "loss/out": 4.020435333251953, "created_at": "2025-01-15T05:24:25.135961+00:00"} {"global_step": 11644, "acc_step": 0, "speed/wps": 12963.32545353085, "speed/FLOPS": 203606839723594.0, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.944214940071106, "optim/lr": 0.0028933043740708205, "optim/total_tokens": 3052404736, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 327927, "loss/out": 4.0356926918029785, "created_at": "2025-01-15T05:24:30.195038+00:00"} {"global_step": 11645, "acc_step": 0, "speed/wps": 12956.125515186226, "speed/FLOPS": 203493754798139.62, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.4482819139957428, "optim/lr": 0.0028932726354222696, "optim/total_tokens": 3052666880, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 355618, "loss/out": 3.8689169883728027, "created_at": "2025-01-15T05:24:35.260822+00:00"} {"global_step": 11646, "acc_step": 0, "speed/wps": 12960.534865904401, "speed/FLOPS": 203563009710251.4, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 2.0279386043548584, "optim/lr": 0.002893240892227922, "optim/total_tokens": 3052929024, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446505, "loss/out": 4.082887172698975, "created_at": "2025-01-15T05:24:40.332356+00:00"} {"global_step": 11647, "acc_step": 0, "speed/wps": 12955.206528142011, "speed/FLOPS": 203479320843789.7, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.8155051469802856, "optim/lr": 0.0028932091444878816, "optim/total_tokens": 3053191168, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 480474, "loss/out": 4.074070930480957, "created_at": "2025-01-15T05:24:45.400122+00:00"} {"global_step": 11648, "acc_step": 0, "speed/wps": 12958.952347329026, "speed/FLOPS": 203538154081416.72, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1901170015335083, "optim/lr": 0.002893177392202252, "optim/total_tokens": 3053453312, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 465605, "loss/out": 4.0828680992126465, "created_at": "2025-01-15T05:24:50.461064+00:00"} {"global_step": 11649, "acc_step": 0, "speed/wps": 12963.225982641201, "speed/FLOPS": 203605277396584.75, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.49361851811408997, "optim/lr": 0.0028931456353711357, "optim/total_tokens": 3053715456, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439583, "loss/out": 4.011664390563965, "created_at": "2025-01-15T05:24:55.525458+00:00"} {"global_step": 11650, "acc_step": 0, "speed/wps": 12965.760667857441, "speed/FLOPS": 203645088110928.28, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0457361936569214, "optim/lr": 0.002893113873994637, "optim/total_tokens": 3053977600, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 488613, "loss/out": 4.081629276275635, "created_at": "2025-01-15T05:25:00.583586+00:00"} {"global_step": 11651, "acc_step": 0, "speed/wps": 12956.41175415611, "speed/FLOPS": 203498250574495.22, "speed/curr_iter_time": 1.274, "speed/data_load_time": 0.0009, "optim/grad_norm": 1.1602765321731567, "optim/lr": 0.0028930821080728605, "optim/total_tokens": 3054239744, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 443830, "loss/out": 4.068645477294922, "created_at": "2025-01-15T05:25:05.650765+00:00"} {"global_step": 11652, "acc_step": 0, "speed/wps": 12957.069349443304, "speed/FLOPS": 203508579012111.2, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.1200942993164062, "optim/lr": 0.002893050337605908, "optim/total_tokens": 3054501888, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 354562, "loss/out": 4.1545820236206055, "created_at": "2025-01-15T05:25:10.712151+00:00"} {"global_step": 11653, "acc_step": 0, "speed/wps": 12962.208072890106, "speed/FLOPS": 203589289725189.22, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7365012764930725, "optim/lr": 0.0028930185625938845, "optim/total_tokens": 3054764032, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 451288, "loss/out": 4.041834831237793, "created_at": "2025-01-15T05:25:15.773644+00:00"} {"global_step": 11654, "acc_step": 0, "speed/wps": 12960.409123179079, "speed/FLOPS": 203561034747961.6, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0013, "optim/grad_norm": 0.9100416898727417, "optim/lr": 0.0028929867830368937, "optim/total_tokens": 3055026176, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 4.1022844314575195, "created_at": "2025-01-15T05:25:20.833985+00:00"} {"global_step": 11655, "acc_step": 0, "speed/wps": 12957.673676536859, "speed/FLOPS": 203518070799547.94, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.44970327615737915, "optim/lr": 0.002892954998935038, "optim/total_tokens": 3055288320, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 463485, "loss/out": 4.039928436279297, "created_at": "2025-01-15T05:25:25.896422+00:00"} {"global_step": 11656, "acc_step": 0, "speed/wps": 12961.620736995781, "speed/FLOPS": 203580064807884.75, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.40191754698753357, "optim/lr": 0.002892923210288423, "optim/total_tokens": 3055550464, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 462435, "loss/out": 3.874537706375122, "created_at": "2025-01-15T05:25:30.961133+00:00"} {"global_step": 11657, "acc_step": 0, "speed/wps": 12964.180539152503, "speed/FLOPS": 203620270018292.84, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3920413553714752, "optim/lr": 0.002892891417097151, "optim/total_tokens": 3055812608, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 455808, "loss/out": 4.008316516876221, "created_at": "2025-01-15T05:25:36.020363+00:00"} {"global_step": 11658, "acc_step": 0, "speed/wps": 12966.604426879927, "speed/FLOPS": 203658340505822.8, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 1.0064300298690796, "optim/lr": 0.0028928596193613255, "optim/total_tokens": 3056074752, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 326344, "loss/out": 4.014055252075195, "created_at": "2025-01-15T05:25:41.083400+00:00"} {"global_step": 11659, "acc_step": 0, "speed/wps": 12952.993652904563, "speed/FLOPS": 203444564597376.34, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5258281230926514, "optim/lr": 0.0028928278170810514, "optim/total_tokens": 3056336896, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 401411, "loss/out": 4.016400337219238, "created_at": "2025-01-15T05:25:46.150882+00:00"} {"global_step": 11660, "acc_step": 0, "speed/wps": 12963.202555752949, "speed/FLOPS": 203604909445112.84, "speed/curr_iter_time": 1.2743, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.3264213502407074, "optim/lr": 0.0028927960102564324, "optim/total_tokens": 3056599040, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 446747, "loss/out": 3.9861531257629395, "created_at": "2025-01-15T05:25:51.216528+00:00"} {"global_step": 11661, "acc_step": 0, "speed/wps": 12957.624982801608, "speed/FLOPS": 203517305997523.62, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5516098737716675, "optim/lr": 0.0028927641988875713, "optim/total_tokens": 3056861184, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 473801, "loss/out": 4.015450477600098, "created_at": "2025-01-15T05:25:56.279509+00:00"} {"global_step": 11662, "acc_step": 0, "speed/wps": 12966.51515804851, "speed/FLOPS": 203656938416157.22, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.6389950513839722, "optim/lr": 0.002892732382974572, "optim/total_tokens": 3057123328, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466369, "loss/out": 4.01622200012207, "created_at": "2025-01-15T05:26:01.338344+00:00"} {"global_step": 11663, "acc_step": 0, "speed/wps": 12961.230974335422, "speed/FLOPS": 203573943049713.78, "speed/curr_iter_time": 1.2753, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.49227848649024963, "optim/lr": 0.0028927005625175393, "optim/total_tokens": 3057385472, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 439820, "loss/out": 3.870260000228882, "created_at": "2025-01-15T05:26:06.398202+00:00"} {"global_step": 11664, "acc_step": 0, "speed/wps": 12955.444282911189, "speed/FLOPS": 203483055109148.9, "speed/curr_iter_time": 1.2762, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.39950278401374817, "optim/lr": 0.002892668737516576, "optim/total_tokens": 3057647616, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 487044, "loss/out": 4.034085273742676, "created_at": "2025-01-15T05:26:11.461420+00:00"} {"global_step": 11665, "acc_step": 0, "speed/wps": 12960.227446592327, "speed/FLOPS": 203558181267516.88, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.98316890001297, "optim/lr": 0.002892636907971787, "optim/total_tokens": 3057909760, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441679, "loss/out": 4.024717330932617, "created_at": "2025-01-15T05:26:16.524340+00:00"} {"global_step": 11666, "acc_step": 0, "speed/wps": 12960.910497351191, "speed/FLOPS": 203568909518294.94, "speed/curr_iter_time": 1.2733, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5957499146461487, "optim/lr": 0.002892605073883275, "optim/total_tokens": 3058171904, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 300290, "loss/out": 4.148253917694092, "created_at": "2025-01-15T05:26:21.585990+00:00"} {"global_step": 11667, "acc_step": 0, "speed/wps": 12966.137901977167, "speed/FLOPS": 203651013091152.56, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.9267529845237732, "optim/lr": 0.0028925732352511444, "optim/total_tokens": 3058434048, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 371881, "loss/out": 4.008642196655273, "created_at": "2025-01-15T05:26:26.647574+00:00"} {"global_step": 11668, "acc_step": 0, "speed/wps": 12960.306126100706, "speed/FLOPS": 203559417037312.94, "speed/curr_iter_time": 1.2755, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.7373823523521423, "optim/lr": 0.002892541392075499, "optim/total_tokens": 3058696192, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453666, "loss/out": 4.084011077880859, "created_at": "2025-01-15T05:26:31.708465+00:00"} {"global_step": 11669, "acc_step": 0, "speed/wps": 12962.582970368025, "speed/FLOPS": 203595178005243.06, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0008, "optim/grad_norm": 1.0049735307693481, "optim/lr": 0.002892509544356443, "optim/total_tokens": 3058958336, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 481760, "loss/out": 4.059540748596191, "created_at": "2025-01-15T05:26:36.773480+00:00"} {"global_step": 11670, "acc_step": 0, "speed/wps": 12956.868492179097, "speed/FLOPS": 203505424272770.03, "speed/curr_iter_time": 1.2759, "speed/data_load_time": 0.001, "optim/grad_norm": 0.5746188163757324, "optim/lr": 0.0028924776920940796, "optim/total_tokens": 3059220480, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 466392, "loss/out": 4.136264801025391, "created_at": "2025-01-15T05:26:41.839487+00:00"} {"global_step": 11671, "acc_step": 0, "speed/wps": 12960.079389824963, "speed/FLOPS": 203555855832534.1, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5887628197669983, "optim/lr": 0.0028924458352885137, "optim/total_tokens": 3059482624, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 441679, "loss/out": 3.9948556423187256, "created_at": "2025-01-15T05:26:46.907618+00:00"} {"global_step": 11672, "acc_step": 0, "speed/wps": 12958.998841133232, "speed/FLOPS": 203538884330500.97, "speed/curr_iter_time": 1.2744, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4791939854621887, "optim/lr": 0.002892413973939849, "optim/total_tokens": 3059744768, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 293067, "loss/out": 3.9134628772735596, "created_at": "2025-01-15T05:26:51.968542+00:00"} {"global_step": 11673, "acc_step": 0, "speed/wps": 12964.852335639933, "speed/FLOPS": 203630821505273.06, "speed/curr_iter_time": 1.2741, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.7199958562850952, "optim/lr": 0.0028923821080481882, "optim/total_tokens": 3060006912, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 313309, "loss/out": 3.958453416824341, "created_at": "2025-01-15T05:26:57.029768+00:00"} {"global_step": 11674, "acc_step": 0, "speed/wps": 12959.66874501129, "speed/FLOPS": 203549406091450.25, "speed/curr_iter_time": 1.2745, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5225703716278076, "optim/lr": 0.002892350237613637, "optim/total_tokens": 3060269056, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 325907, "loss/out": 4.085204124450684, "created_at": "2025-01-15T05:27:02.090974+00:00"} {"global_step": 11675, "acc_step": 0, "speed/wps": 12962.791942240288, "speed/FLOPS": 203598460195654.4, "speed/curr_iter_time": 1.2748, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.4903804063796997, "optim/lr": 0.0028923183626362987, "optim/total_tokens": 3060531200, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 378395, "loss/out": 4.000514030456543, "created_at": "2025-01-15T05:27:07.155354+00:00"} {"global_step": 11676, "acc_step": 0, "speed/wps": 12958.371956988058, "speed/FLOPS": 203529038253571.9, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0009, "optim/grad_norm": 0.6487368941307068, "optim/lr": 0.002892286483116276, "optim/total_tokens": 3060793344, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 453131, "loss/out": 3.9443533420562744, "created_at": "2025-01-15T05:27:12.216815+00:00"} {"global_step": 11677, "acc_step": 0, "speed/wps": 12967.459044256073, "speed/FLOPS": 203671763446082.53, "speed/curr_iter_time": 1.2735, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3927764892578125, "optim/lr": 0.002892254599053675, "optim/total_tokens": 3061055488, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 494183, "loss/out": 3.9444222450256348, "created_at": "2025-01-15T05:27:17.274237+00:00"} {"global_step": 11678, "acc_step": 0, "speed/wps": 12961.098891800355, "speed/FLOPS": 203571868512000.16, "speed/curr_iter_time": 1.2751, "speed/data_load_time": 0.0008, "optim/grad_norm": 0.5083263516426086, "optim/lr": 0.002892222710448599, "optim/total_tokens": 3061317632, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 437195, "loss/out": 4.067685127258301, "created_at": "2025-01-15T05:27:22.335362+00:00"} {"global_step": 11679, "acc_step": 0, "speed/wps": 12957.585353796507, "speed/FLOPS": 203516683569542.53, "speed/curr_iter_time": 1.2756, "speed/data_load_time": 0.001, "optim/grad_norm": 1.0205403566360474, "optim/lr": 0.0028921908173011514, "optim/total_tokens": 3061579776, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 483093, "loss/out": 3.8980748653411865, "created_at": "2025-01-15T05:27:27.396611+00:00"} {"global_step": 11680, "acc_step": 0, "speed/wps": 12957.993172637302, "speed/FLOPS": 203523088924839.03, "speed/curr_iter_time": 1.2749, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5995553135871887, "optim/lr": 0.0028921589196114375, "optim/total_tokens": 3061841920, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 471946, "loss/out": 4.047863960266113, "created_at": "2025-01-15T05:27:32.457834+00:00"} {"global_step": 11681, "acc_step": 0, "speed/wps": 12967.082089293066, "speed/FLOPS": 203665842850398.12, "speed/curr_iter_time": 1.2736, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.5784905552864075, "optim/lr": 0.00289212701737956, "optim/total_tokens": 3062104064, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 429470, "loss/out": 3.8178672790527344, "created_at": "2025-01-15T05:27:37.515671+00:00"} {"global_step": 11682, "acc_step": 0, "speed/wps": 12953.640104073917, "speed/FLOPS": 203454718001308.03, "speed/curr_iter_time": 1.2757, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.4846609830856323, "optim/lr": 0.002892095110605624, "optim/total_tokens": 3062366208, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 295894, "loss/out": 4.014578819274902, "created_at": "2025-01-15T05:27:42.583500+00:00"} {"global_step": 11683, "acc_step": 0, "speed/wps": 12960.073176708662, "speed/FLOPS": 203555758247004.88, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.3382597863674164, "optim/lr": 0.002892063199289733, "optim/total_tokens": 3062628352, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 456868, "loss/out": 3.9462761878967285, "created_at": "2025-01-15T05:27:47.674895+00:00"} {"global_step": 11684, "acc_step": 0, "speed/wps": 12964.514718497718, "speed/FLOPS": 203625518764120.0, "speed/curr_iter_time": 1.2739, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.693051278591156, "optim/lr": 0.0028920312834319916, "optim/total_tokens": 3062890496, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 493656, "loss/out": 3.8785595893859863, "created_at": "2025-01-15T05:27:52.748957+00:00"} {"global_step": 11685, "acc_step": 0, "speed/wps": 12957.79472528919, "speed/FLOPS": 203519972036541.72, "speed/curr_iter_time": 1.2754, "speed/data_load_time": 0.0007, "optim/grad_norm": 0.5286481380462646, "optim/lr": 0.002891999363032503, "optim/total_tokens": 3063152640, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 319464, "loss/out": 3.8417413234710693, "created_at": "2025-01-15T05:27:57.816145+00:00"} {"global_step": 11686, "acc_step": 0, "speed/wps": 12962.664785083658, "speed/FLOPS": 203596463017777.28, "speed/curr_iter_time": 1.275, "speed/data_load_time": 0.0006, "optim/grad_norm": 0.6952343583106995, "optim/lr": 0.0028919674380913727, "optim/total_tokens": 3063414784, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 323502, "loss/out": 4.048700332641602, "created_at": "2025-01-15T05:28:02.875455+00:00"} {"global_step": 11687, "acc_step": 0, "speed/wps": 12952.243477039125, "speed/FLOPS": 203432782054559.97, "speed/curr_iter_time": 1.2758, "speed/data_load_time": 0.001, "optim/grad_norm": 1.0295137166976929, "optim/lr": 0.0028919355086087038, "optim/total_tokens": 3063676928, "memory/max_active_gib": 59.213603019714355, "memory/max_active_pct": 74.8109338974098, "memory/max_reserved_gib": 62.310546875, "memory/max_reserved_pct": 78.72363723290228, "memory/num_alloc_retries": 0, "memory/num_ooms": 0, "memory/power_draw": 406484, "loss/out": 4.00034761428833, "created_at": "2025-01-15T05:28:07.939160+00:00"}